OpenAI + HR Platforms: 11 ideas to integrate OpenAI & ChatGPT in your SaaS HR Platform

a happy cat wearing glasses thinking new ideas, highly expressive oil painting (created with Dalle2)

Integrating OpenAI into your SaaS platform can bring a wealth of benefits for your customers. Imagine automating repetitive tasks, improving decision-making, and even forecasting talent needs. In this article, inspired by a discussion with a Polish HR-Platform SaaS vendor, we’ll dive into how OpenAI can be leveraged by software development companies, startups, scaleups, SaaS vendors to revolutionize the HR operations.

Note: If you want to have a look at the general capabilities,limitations,pricing of OpenAI on Azure, ChatGPT, GPT-3, you can read this previous post: Unlocking the Potential of OpenAI and ChatGPT on Azure: A Guide for Startups and Digital Natives Companies –

First, let’s remind ourselves what are the most common HR Systems Functionality:

  • Recruitment and Employee Screening
  • Company Benefits and Compensation
  • Rewards
  • Performance Evaluation
  • Employee Relations
  • Employee Records
  • Learning and Development
  • Career Planning/Succession Planning
  • Competency management

So, let’s get into the ideas of what OpenAI can do for your HR Platform. I am trying to test the ideas with ChatGPT wherever possible, and stress it a little bit, to see if it can really bring good results.

11 Ideas for OpenAI + HR SaaS platforms

1.Resume screening: OpenAI can be trained on a large dataset of resumes and job descriptions to automatically screen and identify resumes that are the best fit for a particular job opening.

Example: I ask from ChatGPT to read my Linkedin resume and identify the roles I would be a better fit for.

Here is the prompt I give:

And then, ChatGPT gives me an ok-ish reply, mentioning more or less the information I have put in my resume, but in a list. Not impressed but still valid info.

Here is what I get:

Now, let’s give it something more difficult as a test.

I am NOT a developer and I would be a terrible fit for a Chief Developer position. However, since I do have a technical background and I have been working with devs for many years, I thought I would be sneaky and try to fool ChatGPT, into believing that I am a fantastic fit for a dev position.

And the results are….

So, I couldn’t fool ChatGPT and OpenAI. Indeed, it correctly pointed out that I give no evidence of real hands-on experience in developing software. Well done ChatGPT, thnx for destroying my developer career…

2.Interview assistance: OpenAI can assist in the interview process by generating interview questions based on a job description, or by analyzing the candidate’s responses and providing feedback to the interviewer.

Example: Give me 10 interview questions for someone applying for a Chief Information Security Officer Position in a Bank.

I like the questions and I also like the fact that it created a few ones that are domain-related (i.e the one with the regulatory requirements for PCI-DSS). I would happily use these questions as inspiration in my next interview process, to help me create better and more targeted questions to candidates.

3.Candidate matching: OpenAI can be trained on a dataset of successful hires and their job descriptions to predict which candidates are the best match for a particular job opening.

Example: So, as said earlier, I am not a dev. But my brother is. So, I am getting both our resumes from Linkedin and give them to ChatGPT. I then ask from it to analyze both resumes, and let me know which one of the two candidates is a better fit for a developer position.

And the results below are again correct. ChatGPT found out that my brother is the correct candidate and also justifies this. It also explains why I am NOT a good fit, despite having experience related to developers. Good job again ChatGPT; at least the job stays within the family..

4.Chatbot for candidate support: OpenAI can be used to build a chatbot that can answer common questions from candidates, such as information about the company culture, benefits, etc.

Example: I ask ChatGPT for a list of the main benefits for employees at Microsoft

I then follow up with a question regarding what is ESPP and Matching 401k contributions. And here are the results…

Now, you could go even further and try to create a full-fledged functionality HR Chatbot, which ideally could have the following as functionality:

  1. Employee Self-Service: Employees should be able to access information about their benefits, pay stubs, vacation time, and other HR-related information through the chatbot.
  2. Onboarding: New employees should be able to use the chatbot to complete onboarding tasks, such as filling out paperwork and completing compliance training.
  3. Time off and Leave Management: Employees should be able to request time off and check the status of their leave requests through the chatbot.
  4. Benefits Enrollment: Employees should be able to view their benefits options and enroll in coverage through the chatbot.
  5. Employee Directory: Employees should be able to search for contact information for other employees in the organization through the chatbot.
  6. Employee Feedback: Employees should be able to provide feedback and suggestions to HR through the chatbot, with the option to make suggestions anonymously.
  7. HR Policies and Procedures: Employees should be able to access HR policies and procedures through the chatbot, such as company code of conduct, diversity and inclusion policies and emergency procedures.
  8. Employee Recognition: Employees should be able to access information and submit nominations for employee recognition programs through the chatbot.
  9. Chatbot should be able to understand natural language and answer questions related to employee benefits, policies and procedures, and other HR-related topics.
  10. Chatbot should be able to integrate with other HR systems to provide accurate and up-to-date information, and also should be able to route complex queries to the appropriate HR representative.

5.PR and Internal Communication Messages: OpenAI can help you to provide your customers with custom-made AI-generated messages, in the tone you prefer.

Example: I ask from ChatGPT to create the message that a Bank CEO would send to all employees, announcing layoffs to the organization. Certainly the toughest message to send in an organization.

Here is what ChatGPT got:

6. Business English: OpenAI can help with writing sentences related to HR procedures, in a good business english language.

Example: Write 5 different sentences to inform a candidate on withdrawal via email

7. Communication with Candidates: Writing emails to communicate a specific message to candidates.

Example: Rewriting in proper business english a complete email message about informing a candidate on a non-selection.

8. Translation

Example: Please translate the previous message in Polish

9. Create AI-generated Job Descriptions: OpenAI can create job descriptions for your HR people, that can be posted on Linkedin.

Example: I ask from ChatGPT to create a job description for the Chief Information Security Officer position in a bank.

Here are the results, pretty solid to become a full job ad after a few edits from the hiring manager:

10.Interview scheduling: OpenAI can be used to automate scheduling interviews with pre-selected candidates, by using natural language processing to understand the availability of both the candidate and the interviewer.

There is no built-in outlook/calendar integration provided and you have to build it. You can then use the NLP to find the availability from both sides (HR and candidate)

One could argue that there are easier ways to find common calendar availability though, that might work faster than the NLP one. Not sure on what works faster, need to see and try a real NLP solution on that, and haven’t seen anything similar yet in the market.

11. Competency Based Management Support

A few HR systems, provide support for a competency-based approach for the organization, and in that case a competency dictionary is necessary.

In the below example, I ask from ChatGPT to write a behavior-based competency description for “Leadership”.

And then, I ask for a questionnaire that can be used for self-assessing the leadership competency level. I get open-ended questions but then I am trying to get them into a multiple-choice type of questionnaire.

And the multiple-choice results:

I hope these ideas provide a spark of inspiration to start looking into OpenAI/ChatGPT and how to integrate it in your SaaS HR Platform.

Unlocking the Potential of OpenAI and ChatGPT on Azure: A Guide for Startups and Digital Natives Companies

OpenAI and ChatGPT are the new cool kids on the block, so let’s have a look how startups and software development companies can leverage them to create cutting-edge applications. OpenAI is already available as a managed service on Azure and ChatGPT is coming soon.

In this post we’ll have a look at:

  • What is OpenAI
  • Dall-E
  • GPT-3
  • ChatGPT
  • The 4 GPT-3 language models that startups can use
  • The Codex models
  • Potential Use Cases for Startups (overall and per language model)
  • What apps are startups already developing with OpenAI
  • Open AI on Azure: How to get access to it, integrations, advantages and pricing models + cost considerations
  • Open AI competition
  • Limitations of OpenAI
  • Resources and further information

(Note: If you are a startup, you can get for free $2.500 USD credits to OpenAI, through applying to the Microsoft Foundershub program).

Introduction to OpenAI, DallE, GPT3 and ChatGPT: Capabilities and Potential use cases For new Apps

OpenAI is a research company that was founded in 2015, in San Francisco, by Elon Musk, Sam Altman (former president of YCombinator), Peter Thiel and others, with the goal of developing and promoting friendly AI in order to benefit humanity as a whole. The initial funding was $1billion USD and Microsoft invested an additional $1B USD in 2019.

OpenAI has released 3 AI solutions that have become super popular in the past 12-24 months: GPT-3, Dall-E and ChatGPT.


GPT-3 (Generative Pre-training Transformer 3), is a language model trained on trillions of words on the internet. It can generate human-like text, translate text, summarize text and answer questions. It can write poems, science fiction, and create chatbots and virtual assistants that can hold natural conversations with humans.

GPT-3 in a dialog can do empathetic math, keep the context and “remember” and copy your style of communication. Here are three examples, from startup Replika presentation (link) that explain these three capabilities with 3 neat examples.


Dall-E is a deep learning model, that can generate digital images from a text. It can also create “similar” images to an image you give it, or it can “continue and extend” your images.

In the below example, I asked from Dall-E to create the interior design for a small living room. It didn’t get it 100% correct but it wasn’t too bad for a start.

One of the following 4 paintings belongs to my 6-years young daughter. I gave it as input to Dall-E and it created the other three paintings, based on my daughter’s style. My daughter couldn’t actually identify which one was hers, when I asked her after a couple of weeks have passed.

Which one is the original? And which one is Dall-E made?

And getting more creative, I asked from Dall-e to act as if it was the famous painter Basquiat, and paint the below for me. I could see this as a painting hanging in a room..

Tip for Music Lovers: I am more into creating music and there is an open source AI project for auto-creating loops from text prompts: Try it at – I’m very much looking forward to the era when AI will be helping us writing Cubase songs with just text prompts such as “Write a drums midi rhythm in Cubase, similar to the intro from Paradise city, and add a bass line in A minor, with the funky sound of bass of RHCP

But enough with the artistic endeavors; let’s get back to business. Let’s have a look at ChatGPT.


ChatGPT is the conversational version of GPT-3 that can be used to create chatbots, voice assistants, and other conversational AI applications. It can have as an output text, or even code (e.g. a python script).

Please note that ChatGPT is coming soon to Azure (announcement from Satya Nadella on January 16 2023)

If you have played with ChatGPT, just skip this section, otherwise here are a few examples of what it can do:

I ask Chat GPT to create a one-day travel itinerary

Can ChatGPT create a poem dedicated to all of us musicians trapped in the body of an electrical engineer?

That’s at last a song I can relate to!

Then I ask from ChatGPT to write the code for a simple education game, I can give to my daughter to help her learn multiplication tables. Here is what I got as a code:

I tried the code in to see if it actually works, and indeed it works as expected:

Obviously, you should not blindly trust the code produced; since I am not a developer, I asked my brother (who is a ninja-guru dev) to make a few tests of the produced code on text I would write (he would compile it just by seeing it..pure magic to my eyes), and he agreed that it’s in a pretty good shape for some general purpose code (obviously far from substituting the work of devs but pretty good to save them a few hours per week).

Going back to the capabilities of ChatGPT I asked from it to create a VC pitch for a startup idea I gave to it as input: A drone delivery service for sailboats outside of Mykonos island in Greece. even got a nice name for my startup “Saildrone” is the pitch

ChatGPT remembers the context of the previous questions and builds on that knowledge. Here is a follow-up questions and its answer:

It can get creative with providing ideas for ads on social media and the internet.

And obviously, translation in French is an easy task:

I also asked from it to suggest a TV ad scenario and here is what I got:

I also try to find the intersection of ChatGPT with Dalle, so I ask from ChatGPT to create a text prompt for an ad I can use with Dall-e.

And here is what I get from Dall-e for the above prompt.

Not there yet but I am sure I can get a better result if I try a few more times.

One of the main advantages of GPT-3 and ChatGPT is that they can be fine-tuned for specific tasks and use cases. For example, a chatbot built using ChatGPT can be fine-tuned to understand and respond to specific domain-specific phrases. This allows developers to create chatbots that are tailored to their specific business or industry, providing a more personalized experience for users.

So in the above example, if you are the “Saildrone” startup, you could easily create a chatbot that takes orders or replies to customer questions, that is more targeted to the sailing tribe (e.g the specific phrases that are used to describe ports, docks, and other areas of potential delivery).

Another advantage of OpenAI and ChatGPT is that they can be used to create applications with minimal development time. Because GPT-3 and ChatGPT are pre-trained on large datasets, developers can use them as a starting point for their own applications, rather than having to train their own models from scratch. This saves a significant amount of time and resources and allows developers to focus on creating the user interface and other aspects of their applications.

So from all the above examples and discussions, we can think of a few potential use cases for OpenAI and ChatGPT (and here are more of them:

  • To create chatbots for providing customer service for an e-commerce website
  • For helping e-banking websites customers understand in simple terms the different terms of accounts/finance jargon etc, or navigate the website
  • To answer questions of citizens in gov portals on how to do e-citizen tasks
  • To summarize medical reports
  • To summarize news
  • To create articles, essays, blogposts, content
  • To scan profiles of employees and provide feedback to HR
  • To translate in friendly or business style, the content of a website
  • to write replies to emails
  • to write simple code that can help us build an application
  • to do financial analysis and forecasting
  • do sentiment analysis to the call center calls and find the VIP customers who are unhappy with your service
  • Generate educational content, games, quizes, stories for children
  • And many more…

The Four Language Models of GPT-3

Open AI GPT-3 is providing four different language models, called Ada, Babbage, Curie, and Davinci.

How are they different and what type of different apps can you develop?


Ada is the smallest and most cost-effective of the four models, designed for simpler language tasks such as sentiment analysis and intent classification. Ada can understand and respond to text input in a basic way, but it does not have the fine-tuning capabilities that the other models have. The pricing for Ada is per API call, making it a good option for startups who need a simple and cost-effective solution.

Use for: Parsing text, simple classification, address correction, keywords


Babbage is the next step up from Ada, and it is designed for more complex language tasks such as language translation and text summarization.

Use for: Moderate classification, semantic search classification


Curie is the next step up from Babbage, and it is designed for even more complex language tasks such as question answering and conversation generation.

Use for: Language translation, complex classification, text sentiment, summarization


Davinci is the most powerful of the four models and designed for the most complex natural language tasks such as creative writing, poetry, and fiction generation. It is also the most expensive of the four models.

Use for: Complex intent, cause and effect, summarization for audience

Here are some example of how to integrate each one of them in your solution:

A startup that specializes in language learning could use Ada to create a chatbot that can hold basic conversations with users in different languages, providing a simple and cost-effective solution. Babbage could be used to create a chatbot that can translate phrases and idioms for users in different languages, providing a more advanced solution. Curie could be used to create a chatbot that can answer questions about grammar and other language-specific topics. Davinci could be used to create a chatbot that can generate creative writing such as poetry.

If you ask about what is the pricing for each one of them, here is the table with the pricing of the different OpenAI language models on Azure.

Pricing of Ada, Babbage, Curie and Davinci

The Codex Models

Codex is a fine-tuned version of the fully-trained GPT-3 model. These are models that can understand and generate code. Their training data contains both natural language and billions of lines of public code from GitHub (159GB of public data repositories were used to train the model)

They’re most capable in Python and proficient in over a dozen languages, including C#, JavaScript, Go, Perl, PHP, Ruby, Swift, TypeScript, SQL, and even Shell.

Similar to GPT-3, the Davinci model is the strongest one in analyzing complicated tasks and the Cushman model is the fast one in code generation tasks (and it is cheaper than Davinci).

You can use Codex via the Gihub Copilot:

Here is the Open AI Codex demo on youtube:

OpenAI Codex Live Demo

If you are into reading research papers, this is the one for Codex: and PDF Download of the paper here. The paper says that Codex has solved close to 30% of all the problems that the scientists gave them, according to the HumanEval evaluation set they created.

Going into the potential applications of Codex, here are a few ones from / :

  • Github copilot is the most famous integration for the moment.
  • Pygma aims to turn Figma designs into high-quality code.
  • Replit leverages Codex to describe what a selection of code is doing in simple language so everyone can get quality explanation and learning tools. Users can highlight selections of code and click “Explain Code” to use Codex to understand its functionality.
  • Machinet helps professional Java developers write quality code by using Codex to generate intelligent unit test templates.

I think, that what matters the most here to startups, is to save time from their dev team. The dev cost per hour ranges from 20 USD to even 200 USD, so every hour saved is meaningful. And Codex & Github Copilot seem to be delivering on that frontier.

OpenAI, ChatGPT and Azure: Integration, How to get Access, Advantages, Cost


Both OpenAI and ChatGPT are integrated with Azure using the OpenAI GPT-3 API. The API can be accessed via Python, Java, C#, JavaScript etc, using Azure’s Cognitive Services, which is a set of pre-built APIs for natural language processing, computer vision, and other AI-related tasks. Here is also a chatGPT for VS Code

How to Get Access To Azure OpenAI

If you want to try OpenAI and ChatGPT on Azure, just create an Azure account and sign-up for the OpenAI GPT-E API. If you are a startup, you can get for free $2.500 USD credits to OpenAI, through applying to the Foundershub program.

If you are not a startup, then you have to apply for access, here

Note: Microsoft announced on 16th of January the General Availability of the Azure OpenAI service:

Advantages of OpenAI on Azure

One of the main benefits of using OpenAI and ChatGPT on Azure is scalability. Because the GPT-3 model is hosted on Azure’s cloud, developers can easily scale their applications to handle a large number of requests without having to worry about managing their own infrastructure. Additionally, using Azure’s Cognitive Services allows developers to add more functionality to their applications without having to build it from scratch.

Azure Cognitive Services is a collection of pre-built APIs for natural language processing, computer vision, and other AI-related tasks. These APIs can be used to add additional functionality to applications built with OpenAI and ChatGPT, such as natural language understanding, sentiment analysis, and image recognition.

Some of the functionality of Azure Cognitive Services that developers can use include:

  1. Language Understanding (LUIS): This API allows developers to add natural language understanding to their applications. It can be used to identify and extract entities, intents, and other information from text input.
  2. Text Analytics: This API allows developers to extract insights from unstructured text data, such as sentiment analysis, key phrase extraction, and language detection.
  3. Speech to Text and Text to Speech: These APIs allow developers to add speech recognition and text-to-speech capabilities to their applications.
  4. Computer Vision: This API allows developers to analyze images and videos to extract information such as objects, faces, and text. It can be used for tasks such as image recognition, object detection, and OCR.
  5. Personalizer: This API allows developers to personalize the user experience by providing recommendations and content tailored to individual users.
  6. Anomaly Detector: This API allows developers to detect anomalies in time series data, such as network traffic or sales data.
  7. Translator: This API allows developers to add language translation functionality to their applications, supporting more than 60 languages.
  8. News Search: This API provides access to a search index of news articles, allowing developers to search for news by keyword, category, or location.

So, in simple terms, by using all these APIs, your developers don’t have to build everything from scratch and can focus on the UI, Security and other aspects of their apps. You are also getting all the benefits of a managed service (scalability, elasticity, performance, security etc)

Pricing of the Azure OpenAI service

Pricing of the OpenAI on Azure is on a pay-as-you-go model, so developers only pay for the requests they make to the API. How much does it cost? You can check the pricing details here and at the table below.

Since the API pricing is based on tokens, it’s important to understand them.

A token is a bundle of a few letters. E.g the word hamburger is 3 tokens (ham – bur – ger). The complete Shakespeare literature is 900.000 words, which is around 1.2 million tokens. So, consider that 750 words worth 1000 tokens, or a token to word ratio of 1.4.

Another pricing consideration is that, in the API Call, you get charged both for the prompt you are sending and for the predicted text.

So, if you are using the best and most expensive model of OpenAI, the “DaVinci” model, it is 2 cents for every 1000 tokens (or 750 words produced via the API calls). This can easily add up to large numbers if you are building apps that will be heavily used in terms of text by a large customer base. Make sure you take this into account when building your apps. There are startups that had to move away from OpenAI on a later stage due to costs (e.g see the Replika presentation on that here)

Below a couple of examples on how to calculate the total cost of using the OpenAI service on Azure; basically you need to factor:

  • Which language models you are going to use (DaVince?Curie?Both?)
  • The Tokens you will need per day
  • The hours you are going to deploy the model
  • The Fine-tuning cost

Let’s look into a simple example of a pricing calculation for a new solution.

Let’s suppose that you want to create an application that connects to the website of a popular e-commerce store customer feedback chatbot and to the call-center to do voice-to-text of customer feedback calls. It then analyzes the feedback and tries to classify complaints, do sentiment analysis.

Let’s assume that you have 60 calls per hour from 8:00 in the morning to 20:00, 12 hours per day, for 6 working days. And that you also have 20 customers per hour providing written feedback via the chatbot in the website, for the same 12 hours per day. Let’s assume 250 words text from each customer for both channels (website chatbot and call center)


  • 80 touchpoints per hour, with 250 words per touchpoint, 12 hours per day
  • This is a total of 80x12x250=240.000 words per day
  • 1 word is 1.4 token, so we need 240.000×1.4=336.000 tokens per day
  • Using the Davinci model, costs 0.02 USD per 1000 tokens, this is 6.72 USD per day, or around 161USD for 24 days of the month (running the service from Monday to Saturday).

Now that doesn’t seem much, but the calls per hour might be 10 to 100 times higher and the same could go for the feedback gathered from the website, depending on the popularity of the store and the market it operates in.

Finally, bear in mind that for the moment the OpenAI service on Azure is availble in the South Central US region, which you may want to test for latency-related issues in your application.

Use Cases and Examples of Startups that are currently using OpenAI

Here are a few examples of use cases and startups using Open AI, for your inspiration.

Carmax: Used car retailer CarMax has used Azure OpenAI Service to help summarize 100,000 customer reviews into short descriptions that surface key takeaways for each make, model and year of vehicle in its inventory. Read about the case study here.

In the healthcare industry, GPT-3 is being used to generate medical reports and summaries. A startup called Medopad uses GPT-3 to generate medical reports from patient data, allowing doctors to spend more time with patients and less time on paperwork. Another company called Freenome uses GPT-3 to analyze genetic data and identify potential health risks, helping doctors make more informed decisions about patient care.

In the finance industry, GPT-3 is being used for financial analysis and forecasting. A startup called Blue River uses GPT-3 to analyze financial news and generate reports on market trends, helping traders make more informed decisions. Another company, called Alpaca uses GPT-3 to analyze financial data and generate trading strategies, helping traders identify profitable opportunities in the market.

In the e-commerce industry, GPT-3 is being used to generate product descriptions, improve search results and create chatbots that can answer customer questions. A company called Verbling uses GPT-3 to generate product descriptions for their language learning platform and another company called, Sift Science uses GPT-3 to improve search results on their e-commerce platform.

The next four are part of the OpenAI funded startups:

Descript is a powerful video editor reshaping the way creators engage with content by using AI to make video editing as simple as editing a text document.

Harvey is developing an intuitive interface for all legal workflows through powerful generative language models. Its technology expands a lawyer’s capabilities by leveraging AI to make tedious tasks such as research, drafting, analysis, and communication easier and more efficient. This saves lawyers time, ultimately allowing them to deliver a higher quality service to more clients.

Mem is building the world’s first self-organizing workspace. Starting with personal notes, Mem uses advanced AI to organize, make sense of, and predict which information will be most relevant to a user at any given moment or in any given context. Mem’s mission is to build products that inspire humans to create more, think better, and spend less time searching and organizing.

Speak is on a mission to help more people become fluent in new languages, starting with English. The company initially launched in East Asia with a focus on South Korea, and has nearly 100,000 paying subscribers. Speak is creating an AI tutor that can have open-ended conversations with learners on any number of topics, providing real-time feedback on pronunciation, grammar, vocabulary, and more.

There are more apps for your inspiration at and as more companies discover the benefits of using OpenAI and ChatGPT on Azure, we could expect to see even more innovative use cases and success stories in the future. You can also have a look at and for a list with apps that use GPT-3

Open AI Competition

Here are the main competitive solutions to the Open AI GPT-3 one:

  1. Anthropic AI: Its language model performance seem to be on the top 3 at the moment
  2. Google: Google has several AI models, such as BERT, T5 , which are similar to Open AI’s GPT-3
  3. Microsoft: Microsoft has the Azure Cognitive Services, which include LUIS (Language Understanding) and Text Analytics
  4. Amazon: Amazon has Amazon Lex and Amazon Transcribe which are similar to Open AI’s GPT-3
  5. Facebook: Facebook has an AI platform which includes RoBERTa, which is similar to Open AI’s GPT-3
  6. IBM: IBM watson is the relevant solution here, which offers natural language processing, computer vision and other tasks
  7. Baidu: Baidu is a Chinese company with several research teams on AI, and the model similar to Open AI’s GPT-3 is called ERNIE.

A summary of all the language models can be found here and an assessment of language models from Stanford university can be found here

State of Language Models as of 2023

Limitations of OpenAI & ChatGPT

OpenAI and ChatGPT are powerful AI tools, but like any technology, they have some limitations. Some of the main limitations of OpenAI and ChatGPT include:

1. Data bias: OpenAI and ChatGPT are trained on large datasets of internet text, which can introduce bias into the model’s predictions and outputs. This can be particularly problematic for sensitive applications, such as healthcare or financial analysis.

For example, if a chatbot that uses GPT-3 is trained on a dataset that is mostly written by men, it may not be able to understand or respond well to text written by women, or it may generate responses that are insensitive or offensive to women. Another example of bias could be in a language model that is trained on a dataset that is mostly written in English, it may not be able to understand or respond well to text written in other languages, or it may generate responses that are not accurate or appropriate for that language.

2. Lack of control: OpenAI and ChatGPT are pre-trained models, which means that users have limited control over the specific parameters and settings of the model. This can make it difficult for users to fine-tune the model for specific use cases.

For example, a startup that wants to create a chatbot that can hold natural conversations with users in a specific industry, such as finance, may not be able to fine-tune the model to understand and respond to specific industry-specific phrases and jargon. This can make it difficult for the chatbot to hold natural conversations with users in that industry and make the experience less personalized for the users.

Another example could be if a startup wants to create a language model that can generate text in a specific tone or style, such as a formal tone, it may not be able to fine-tune the model to generate text in that specific tone. This can make it difficult for the model to generate text that is appropriate for the desired use case.

It’s important to note that OpenAI has released a number of tools and guidelines for fine-tuning GPT-3 for specific use cases, such as GPT-3 fine-tune, which allows developers to fine-tune the model on their own dataset, making the model more suitable for their specific task. Additionally, OpenAI also offers several options for controlling the temperature and other parameters of the model to make it more suitable for specific use cases.

3. Limited interpretability: OpenAI and ChatGPT are neural networks, which means that they can be difficult to interpret and understand. This can make it difficult for users to understand how the model is making its predictions and to identify and correct errors.

4.High cost: OpenAI and ChatGPT are commercial products, which means that they can be quite expensive to use, depending ofcourse on the use case. This can be a barrier for some startups and small businesses that may not have the budget to use them.

5. History of Data Training: The model is trained with data until 2021, so it misses a couple of years of new events and information.

6. Prompt Size: The OpenAI maximum length of text prompt size is 2048 tokens which is around 1500 words.

Resources and Further Reading

  1. OpenAI’s documentation:
  2. Azure’s documentation: and
  3. Community forums: OpenAI API Community Forum / and Azure Cognitive Services Forums
  4. Webinars and training sessions: and
  5. Partner program:
  6. Blogs and articles: and
  7. Case studies:

Innovation in Retail from China and around the Globe

Retail is an amazing area for tech innovation: AI, Big Data, IoT, Drones, Autonomous Delivery Vehicles, Social Media, Influencers…It’s all there.

This week we had a very interesting retail industry session on the latest trends and innovations by Noah Herschman, our Online and Offline Retail Innovation Leader, and as we are working with many ISV partners developing solutions for retail customers, I thought I’d share my notes.

#1 Complex to Simple

The first trend is that retail per category goes from being very complex to become very simple. There are several brands capitalizing on this trend. As an example, if you search Amazon for mattresses you will get 30.000 results but Casper has simplified this to only 4 types, making it much easier for the customer to go through the different options.

Casper capitalizes on the retail category simplification trend. Only 4 choices, with easy-to-understand differences.

The same goes with shoes or with Harry’s for razors. Harry’s decided that they are not going to compete in the game of “more blades are better”, standardized on 4 blades and instead tried to offer a subscription service for shaving.

#2 Private Labels vs Direct to Consumer (D2C)

The store brands are growing; The private label brands of retailers are growing.
At the same time brands are increasingly selling directly to consumers. As an example, Nike is selling online directly to consumers and it even has products on its website which are not available through their channel/resellers. No one can predict where this will settle but at its extreme, we could see in the future well-established brands selling only D2C through their websites, to their customer base.

#3 Gamification Trend

Retail is trying to figure out how to use fun stuff and games to increase customer engagement and sales. Two examples in the gamification trends are Drest and Pinduoduo.

Drest, combines gaming with e-commerce. It allows its users to play the game of being a pro stylist via virtual goods of 200 luxury brands, including Gucci and Bottega Veneta. Users can purchase any of these items at the end. This trend is called RVR, which stands for Real Goods turned into Virtual and then purchased in Reality and you can read more about it here.

With Drest, digital clothing is one step closer to mainstream | Vogue  Business
Drest allows users to become pro-stylists using virtual luxury brand items.

Pinduoduo is the second biggest marketplace in China after Alibaba. Pinduoduo has created a game that helped the platform to sell agricultural products during the pandemic. You can play the Duoduo orchard game and have the chance to win 1kg of fruits.

The Pinduoduo food game lets you grow plants and gives you the opportunity to get free food in return.

#4 Visual AI Use Cases

Another key trend in retail, is visual search, with the use of AI technology. As an example, Syte, from Israel, has developed a Visual search solution.
It can find all the fashion products in a picture and the user can then purchase them online. Syte is focusing on home accessories and fashion items; see how it works here.

Syte’s visual search AI tool can discover all the fashion items from a photo.

Apparel Sizing, is a key focus area for innovative solutions, as in fashion there is a 35% return rate due to improper fitting. Here you can find several startups trying to solve this problem, such as Presize, Virtusize, and Bigthinx.

Facial recognition: The innovation in facial recognition for retail focuses on finding the buyer sentiment, finding shoplifters, creating a better checkout process. As an example, in China, there is a “Smile to Pay” solution, in which you just have to look in front of a camera to pay for your order in a retail shop, as your face is linked to your bank account.

KFC in China tests letting people pay by smiling - The Verge
KFC in China using facial recognition technology to let customers pay. [Image source]

Inventory tracking: There are solutions that let you use the phone to scan barcodes and find which products to restock in your store, prices, etc.

#5 Unstaffed Stores

In China, the unstaffed stores sector is taking off with China’s autonomous retail sector estimated to grow to 7.1 billion USD this year. There are mainly three categories of these: unstaffed shelves, vending machines, and unstaffed stores.

An example here is the startup F5 Future store, which raised 14.5M USD to build fully autonomous convenience stores in China. These stores are open 24×7 and are operated entirely by robotic arms.

An F5 future store in China
A robotic arm delivering food in an F5 Future store in China.

I would also add here the autonomous delivery vehicles, that have been used in China, especially since the Covid start, such as the Meituan’s one.

China's e-commerce giants deploy robots to deliver orders amid coronavirus  outbreak | South China Morning Post
Meituan’s autonomous delivery vehicle (Image Source:SCMP)

#6 Innovative Loyalty Programs

In retail, there are multiple types of Rich Customer data that can be collected:

  • Personally identifiable information (PII)
  • Demographic information (gender/age/income/ethnicity/wealth)
  • Psychographic information (hobbies/interests/music/movies)
  • Shopping Behaviors (categories, frequency, spend, paring)
  • Connections (social media connections of customers)

The two big problems here are:
How do you get this data? how do you convince customers to hand over their PII data?
– How do you entice customers to read your marketing messages (email/SMS/Viber/FB/WhatsApp etc)?

The 8 Types of Loyalty Programs

There are eight kinds of loyalty programs:

  1. Points Programs: E.g., the banks’ loyalty points program that give you 1 point for 1 USD spent.
  2. Tiering Programs: E.g., the airline companies loyalty programs with gold/silver cards for flyers.
  3. Charity Programs: Every time you purchase, a certain amount goes to a charity.
  4. Coalition Programs: Different partners under the same loyalty scheme,
  5. Game/Gamification Programs: E.g. the Pinduoduo’s game example above.
  6. Paid Programs (e.g. the Amazon Prime)
  7. Hybrid Programs (they offer multiple types of benefits, e.g. a 2% on gas, a 5% discount at a hotel, etc)
  8. Value-Added Programs (like Amazon Prime which offers free shipping but no discount deals, or the Nike program [see below])
  9. No loyalty programs: E.g., Apple and Microsoft don’t offer loyalty programs

The truth is that loyalty programs are saturated, so you really need to stand out to make an impact on your business. In the US, 14.8% are using loyalty memberships and 6.7% are active members.

Some interesting examples of loyalty programs approaches are the following:

Nike offers zero discounting. Instead, they have created a value-added loyalty program, the NIke Plus Membership, which offers: Free shipping, events, audio-guided runs, home workouts, training plans, and other perks they are giving for free. Nike tries to add value to its customers without discounting its products.

The Nike membership adds value to its customers without product discounts

2.Taylor Stich
Another interesting approach to creating loyalty, the crowd-sourcing loyalty. Taylor Stitch is doing crowd-funding for some of its new products and pre-funds various clothing styles.

Taylor Stitch crowdfunds some of its new product designs.

Klarna has launched a “Buy now, Pay Later“, loyalty program, which tries to undercut the high credit card interest fees for shoppers.

The Buy-Now-Pay-Later loyalty program of Klarna, called Vibe

4.Hong Kong MTR: (metro station of Hong Kong)
This is a loyalty system that Microsoft Consulting Services have built for the metro station of Hong Kong. It is a points program that covers transportation, the shopping malls that exist underway, the in-station shopping experience, and others. As an example, you can get a 30% discount at a Starbucks coffee, while waiting for your next train to arrive.

The MTR mobile loyalty program in Hong Kong

You can also watch the following video for a preview of how the MTR mobile works:

The MTR mobile loyalty program

#8 KOL – Online Influencers

KOL stands for key opinion leader, and it is mostly a term used in China. The term refers to the youtube/live-streaming/social media influencers who direct consumer behavior and sales.

The influencers in the US are driving/influencing around 10 billion USD of sales, while in China this is much higher, at around 340 billion USD until 2022. Alibaba has even a KOL platform, which lets influencers connect with brands.

Alibaba’s KOL platform connects influencers with Chinese brands

An example is the 11.11 global shopping festival. According to Forbes, live streaming drove 6 Billion USD in sales during the festival. 300 million Taobao users watched live streams.

Viya, is the most famous e-commerce influencer, and according to Bloomberg, the e-commerce live streaming queen.

Another top influencer is Austin Li. Here is what Forbes has to say for Viya’s and Austin Li’s performance during the double eleven shopping period.

A chart with 11.11 pre-sale results for Alibaba’s two top live streamers.
Austin Li drove sales of 100 million USD in one day through live streaming. Source: Forbes

The top KOLs operate multi-employee businesses to achieve these results. Have a look at the crew of Viya:

Simsim, is an example of a video e-commerce platform coming from India. Simsim sells fashion, beauty, and electronics products as well as health and wellness items. Simsim works with influencers known as community opinion leaders.

Fanjoy, is an e-commerce platform only driven by influencers. At Fanjoy, the influencers can create products based on their real-life experiences, interests, and expertise. E.g. fashion items, toys, and DIY items. Then, they send them as a subscription, surprise box to the customers.

Fanjoy lets influencers sell and send surprise boxes to their followers

Yunji, is another example of this area, coming from China. It has a business model that lies somewhere in between Costco’s and Amway’s.

#9 Farm-to-Table Supply Chain

The farm-to-table innovation trend in retail tries to solve the problem of how to get fresh food on the tables of consumers, with the use of technology.

Freshippo, the grocery chain store of Alibaba, is an example in China. You can trace all the history of the fresh products. You scan the barcode and you know the farm that the food was produced at, when it is harvested, the reputation of the farmer, and other info. Here is a video:

Pinduoduo is also an example. Pinduoduo digitalizes the agricultural products supply chain in various ways, such as visual AR for growing plants, or presell the products before they are planted.

Jutudi, in China, is a Farm-to-fridge service. It allows consumers to buy a plot of farmland and have the crops shipped to them after the harvest.

#10 Last-mile innovation

A key innovation trend here are the autonomous robots for delivery. E.g., Suning has autonomous delivery robots that can run for 10 hours, go door to door and deliver products. has a drone delivery service. delivers books using drones in Indonesia during pilot run -  FreightWaves

Kroger and Walmart are testing self-driving robots.

Nuro Kroger self-driving delivery grocery
Kroger’s self-driving robot delivers groceries in Arizona.

CVS tested prescription delivery in Houston.

Nuro's Delivery Bots Adding CVS Prescriptions To Grocery And Pizza Runs
Image Source: Forbes

Smart Lockers

Albertsons has launched two different takes on unattended pickup: A temperature-controlled locker pickup service and a robotic pickup kiosk.

A pickup kiosk from Albertson. Image Source:
Albertsons Pilots Temperature-Controlled Lockers to Expand Pickup Options
Albertons’ pickup temperature-controlled lockers

Visibility Tools for the last-mile delivery

The idea here is to track where your product is, at its last stage of delivery. There are several companies offering a visual map with the location of the delivery driver. FedEx, has launched the Fedex Insight web tool to allow delivery tracking. And Pitney Bowes does the same for parcel tracking.

#11 Big Data and Trend Prediction

Retailers have been gathering a huge amount of data. Just a small sampling of available retail data from the past 20 years: Product 360/customer 360/store 360/employee 360/supplier 360

Now, retailers gather even more data: structured/unstructured data, internal data ( pos transaction data, customer PII, website clickstream, promotions, and elasticity), and external data (online image data, market research, brand data, social media data, in-store image data.

With the new technologies and innovations, you can put all of these data on a datalake and find trends, predictions, and correlations. Potential outputs could be used for demand forecasting, private brand design, etc.

Another interesting use of data is the Social media photo scraping and classification. As an example, you can find what products influencers are using, and with AI image classification you can predict some trends. E.g., what is the next trendy color?

Actually, Microsoft has implemented such a project for a top FMCG brand worldwide, to predict the shapes in lipsticks in various countries of the world.

Service as a Software Business Model

You probably know the Software as a Service business model (SaaS). But there is a new kid in town! The Service as a Software business model is a new trend that emerges thanks to the advances of AI and Machine learning. Here are some thoughts on what opportunities this might bring in the next years.

What is Service as a Software?

It is the idea of finding a service executed by humans and then using software (mainly Machine Learning and AI), to automate it. Once it is automated, you can use Software (an AI bot) or Hardware (a Robot) to perform the task.

A couple of examples:

Example 1: Have Software deliver an Accounting Service

A team of accountants gets all the Expense Report receipts from 500 employees of a company. They do the data entry of the amounts, then decide on the category of the expense and then input the information in the accounting system.

In the Service as a Software model, you have to place a software “watching” the actions of the accountants and getting trained on what it should do. Once the model is trained, you can have the software bots doing the work and keep a couple of accountants to supervise them.

Image result for ui path expense report
UI Path’s software recognizing the fields in an Expense Report.

Some examples of software companies working on this are UIPath, Automation Anywhere and (our Greek) Softomotive.

Example 2: Have software deliver a Call Center Service

A bank uses a call center with hundreds of agents replying to customer issues on credit cards, accounts, bonus points, and others.

Image result for omilia demo
Omilia’s Conversation Intelligence platform, recognizing the intent of the spoken words.

In the Service as a Software model, you build a bot that uses Speech to Text and a domain ontology that understands the concepts that your customers are talking about. Then you have the software answer the calls and provide customer support. Omilia (another great example from Greece) is providing this conversation intelligence using AI very effectively, as recognized by Gartner and Forrester.

Example 3: Have a Robot execute a delivery service

Image result for starship estonia
The Starship robot delivering food in the streets of Esthonia
Image result for amazon drone

You place an order in an e-food restaurant and instead of having a human delivering the food to your place, you have a sweet robot doing the job. Starship is building such robots in Esthonia. And as you know, Amazon is experimenting with drones delivering parcels.

Example 4: Have a robot make coffees

Image result for cafex
The CafeX barista gets instructions via an app, on how to prepare your perfect coffee.

CafeX is disrupting the baristas’ service by having a robotic arm making and delivering coffees, with full customization on the preferences of the customer.

Can Software and AI be better than humans in service delivery?

One might think that nothing can replicate the effectiveness of a human in service delivery. But is this true? Let’s have a look at where we stand regarding the accuracy of humans and AI.

Speech Technology: The word error rate for speech to text software is now at 5.1%, the same as a human. This means that software and humans misinterpret around 5 words in every 100 words they listen to.

Image Recognition Technology: Since 2015, the error rate for image recognition in specific domains for software is 4.5%, while humans are at 5%.

Examining Possible Outcomes and Decision Making: Software wins Chess champions and Go champions. Given a specific set of rules in which an environment operates, it is very tough for a human to win software.

Will AI and Robots substitute humans?

In some areas yes, AI and robots are and will be able to completely substitute a human. In most areas, we will live together and humans will be the supervisors of the services delivered by AI software and robots. And there will always be some areas in which humans will do better (e.g, movie creation services, lawyer services, industry research services etc).

How Can Startups Capitalize on the trend of Service-as-a-Software?

There are two main ways here to disrupt a service area delivered by humans.

The first way is to create the AI/Machine Learning software technology that will deliver the service in a quality-to-price ratio that is better than humans and then sell this technology to businesses.

E.g., develop an AI system that uses image recognition and drones to paint a house. You can then go and sell this technology to every business that paints houses, so that they can become more efficient and deliver the service with a better margin, e.g. because they will not have to create 30-meter scaffolds to paint the exterior of a 10-floor building

Image result for drones painting buildings
Airbrush drone painting a house.

The second way is to use this technology yourself, set up a painting franchise and disrupt all the existing painting service companies. This could prove to scale faster for your business, rather than having to convince and educate every existing supplier of a service to use AI/ML software and hardware.

Ideas for areas for service-as-a-software disruption

Obviously it is hard to predict which service areas will be the next ones to be disrupted by AI software+robots but here are a few ideas, with some of them under development at the moment. These are in random industries and come both from the AI Software and AI+Robots world, just to make the point of the range of areas for disruption.

  • Cleaning the Windows of Skyscrapers and Hotels.
  • Automating Accounting Services in Companies (Expense Reports, Data Entries etc): Mostly solved by RPA companies.
  • Cooking and Food Preparation Services for restaurants: Actually, a Robot Pizza Startup worth now $2 billion USD. Let’s build a Greek Spinach Pie robot please…
  • Coffee Preparation: CafeX underway.
The Mark Skakr robot preparing your bio-juice.
  • Call-Centers: Disruption currently underway with Omilia and other vendors.
  • Holiday Planning: A travel bot could learn all your preferences and suggest the perfect weekend getaway with your family: Book you a flight after 10:00 am because you hate early morning flights, get you a 4-star hotel with average rate on of +9.0 next to a metro, buy 4 metro cards for 3 days, the tickets to the top 3 museums for your children and save you from 20 hours of searching for information on the internet to create the itinerary.
  • Online Exams and Certification Tests: Having an AI system to do the online-proctoring to check via a camera a student who takes an exam for certification.
  • Traffic Police Services (cameras for speed tickets, license plate recognition, insurance fees avoidance checks, etc): Already in use.
  • Security Services: AI software and anti-drone technology will be used to protect critical infrastructure better than the security guards can at the moment. Rada and Kasperky are working on it.
Kaspersky Antidrone technology
  • Trading: High-frequency trading algos using AI are now broadly used and account for the most transactions in the US stock market. They have already disrupted to a large extent the service provisioned by daily desk traders.
  • Music Creation Service: I would not see it far fetched in a few years to have an AI software replacing the job of a music composer and a music audio engineer/producer. It could learn from the patterns of the top hits in a country (most popular rhythms, scales, mix of genres) and produce the next country hit. It could even create a synthetic song to sing, that could be a mix of the 3 most popular singers in the genre. AI music creation is much easier done in electronic music at first. Here is an AI produced album.

Let me know what you think!

12 Reasons to Move your Software Solution to the cloud

In my discussions with software houses that already have an on-premise software solution, the question of “Why should we move to the cloud?”, often comes up.

I know it is 2019 and in many parts of the world the discussion is not “Why Cloud?” but “Which Cloud?“, but a lot of software companies founded before 2005 are still having their solution on-premise.

The more “traditional” the software house, the more years in the business, and the more customers it has on an on-premise solution, the bigger the resistance to move to the cloud world.

And I cannot blame the resistance to change.

On-premise software has a higher profit margin than a cloud, software-as-a-service solution. The developers have to build new skills to develop optimally for the cloud. The IT people supporting and developing the solution will have to learn about the cloud infrastructure and DevOps world. The sellers have to change from calling and meeting customers face-to-face, to delivering online demos and following up leads that come from the web campaigns. And the marketing has to transform to a seller-centric department that creates sales funnels and measures web visitors, customer acquisition costs, lifetime value, campaign costs, A/B testing of websites, A/B testing of online campaigns and so much more.

It’s not an easy task. But given all the benefits and the latest trends, it’s an almost inevitable one.

So, here are the top ten reasons that software houses choose to move their solutions to the cloud world. It is a mix of reasons that are valid for IaaS and/or SaaS cases.

#1 Predictability in Revenues

Selling on-premise software, even with a software assurance or a maintenance contract, makes it difficult to predict when you are going to get paid again in the future. New users may use the product without declaring the licenses, and the new upgraded versions are not easily bought by the customers. Moreover, you cannot predict safely the quarter in which the sales will land; and it’s pretty easy for the deals to slip for some months and affect your forecast predictability.

When selling your software as a service, via the cloud, you usually have monthly payments and you build an “Annuity Business”. Or the famous MRR, which stands for the Monthly Recurring Revenues. The equivalent of collecting a monthly “rent”. Knowing that I have 1000 users paying 19 euros per month for my SaaS solution, and my churn rate and monthly growth, makes it pretty easy to predict the revenues in the forthcoming months and quarters. This affects the cash flows and the financial health of the company, as it “protects” it from the effects of a bad sales forecast predictability.

#2 Shorter Sales Cycles thanks to the free trial

A typical sales cycle of an on-premise software solution to a private sector customer is around four to six months, starting from the first meeting to pitch the solution, until the moment that you see the first euro. Yes, I know that many times it can be 12 to 18 months (especially with enterprise solutions) and in the Public Sector case, we count years instead of months. But let’s keep the six months example.

Most SAAS solutions offer a free trial for some days, weeks or months. Usually a 15 day or 30-day free trial. This makes it easy for a potential customer to try the software and decide if it fits his needs or not. And the customer is psychologically driven to make a go or no-go decision after thirty days, especially if there is a discounted price.

We all know the effects of the free-trial but let’s think about it for a moment with a different example. Imagine that you want to buy a new TV and Samsung brings you the TV for free for 30 days in your living room and gives you a 20% discount to purchase it. Otherwise, you can just say no and Samsung will come and pick it up from your house at no cost.

Compare this to the state of “I am thinking about buying a new TV“, or “talking to sellers in retail shops about TVs“. Which one would create a faster sales cycle?

This is what is happening with the free-trial effect on your sales cycle.

#3 Easier Up-Selling of the more expensive versions

Up-selling is the sales tactic of selling the more expensive version of your product. You go to MacDonalds for a single burger and you end up buying a double one. You get the basic low-definition Netflix subscription for 7.99 USD/month and then you get sold the HD one for 11.99USD/month.

In the on-premise world, upselling means that you have to explain the additional features to the customer, set up a new contract, negotiate discounts, get new internal approvals, deal with technical IT issues. The customers will be waiting for the renewal date of the previous contract to decide on new products. And when you are trying to upsell your new version, you often get quotes such as “let’s wait for others to use it first, maybe it’s too soon and it has bugs,” or “I need one feature but not of all the new ones, and I am doing my job fine with the old version”.

It’s more of a psychological game. If you are used to buying a TV every 3 to 5 years, you will not easily “upgrade” after 12 months. On the other side, if you were paying 15 euros per month to own your TV and someone gave you a new bigger and better one after 12 months for 18 euros/month, you would most probably get the deal.

#4 Easier Cross-Selling of other solutions

Cross-selling is the sales tactic of getting sold a complimentary product. You go to McDonald’s for a burger and you always get asked: “do you want fries with it”?

Cross-selling in the online saas world is pretty straightforward. You go to Amazon to buy a book and you are getting sold a Kindle, an Amazon Music subscription, and more relevant books. The online process of cross-selling is all automated and non-dependent on sellers’ incentives and quotas.

In the on-premise world, there are some obstacles. E.g., a seller who sells Product A might not be incentivized to sell Product B, if it is not in his quota and sales targets. Moreover, the seller may not want to spend the time to get to know Product B, C, D of the company. Even if the seller knows B,C,D and they are in his sales targets, if they are not easily sold and they risk Product A, which is 80% of his revenue, he might not even bring them into the discussion with the customer.

#5 Global Reach

Just try to sell on-premise software to another country. You will most probably need to set up an office, create a partner channel and find system integrators to work with your products. This means that you end up paying up-front, a big cost, for taking the risk to expand your business in a new country. And if you fail to set up a business in Country A, you have less power to go to Country B and then to Country C.

Enter the SAAS world. You localize the saas solution, the sales funnel and the ads and you try your luck in Country A. You fail fast with a small cost. You go to Country B. And C. And D. Until you find success. Ok, it’s never as simple as that but you get the idea. Trial and error in going global costs much less in the cloud world.

#6 Easier way to create a sales funnel and measure results. Marketing is the New Sales.

In the on-premise world, your sellers (internal or partner sellers), keep the keys to your customers. Are you sure that if all your sellers disappear today, you would continue to sell in the coming days? Most probably not.

In the cloud world, and in SaaS applications, you can build a predictable sales engine. You can build a marketing funnel, which you feed with a specific input (money for ads, keywords, content marketing, social media campaigns, adwords), and you get a predictable output (number of web visitors, number of trials, conversion rates from visitors to trial and to paid customer). You can optimize this by A/B testing and trial-error mechanisms.

In the cloud, you have a smaller dependency on your sales team and a bigger dependency on your online marketing team.

Actually, Marketing is the New Sales in the cloud world.

#7 Lower adoption cost for the end customer

A software solution can be deployed in 4 ways:

  1. As an on-premise solution to the software solution provider’s own infrastructure
  2. As an on-premise solution to the infrastructure of the customer
  3. As a cloud solution to the cloud tenant of the software provider (ISV)
  4. As a cloud solution to the cloud tenant of the customer

Let’s suppose we are trying to sell a CRM solution to a bank. If it is an on-premise solution to the bank’s infrastructure, then the bank will have to find available servers or purchase new ones, set up the environment, deploy the solution with the help of the ISV and monitor the infrastructure as the solution operates.

This demands time, energy, money, and human resources from the bank.

Now, compare this to using an online CRM solution, from the cloud tenant of the ISV. The IT of the bank doesn’t have to run/install/deploy the infrastructure, and the bank doesn’t have to invest new resources in hardware.

This leads to a lower adoption cost for the customer in almost all cases if the TCO analysis exercise is done correctly.

#8 Lower TCO for the customer

Buying hardware to host your solution is a CAPEX (capital expenses) move. You pay all the costs upfront for at least the next three to five years.

CAPEX costs include: Server Costs, Storage Costs, Network Costs, Backup and Archive Costs, Disaster Recovery Costs, Datacenter Infrastructure Costs, and Technical IT people Costs

Going to the cloud is an OPEX (operating expenses) move. You pay each month for what you use.

OPEX Costs Include: Leasing of Software and Custom Solutions, Scaling charges based on usage and demand.

In almost all cases, when you do the exercise of the Total Cost of Ownership, for having an on-premise environment VS a cloud environment, the return on investment is better in the cloud world after two-three years.

Below you can have a look at the challenge of the Capex model, especially when there are fluctuations of demand over time.

A graph showing how expected demand can be different from real demand and how CapEx infrastructure can be exceeded by demand.

Examples of extreme fluctuations in demand are:

  • Black Friday for retail websites
  • A local news site with a video that goes worldwide viral
  • National Exams Results for University entry (Ministry of Education website)
  • Teacher allocation/recruitment results (Ministry of Education Website)

#9 Improve your product by aggregating data on your users’ behavior. Shorten the product development cycle.

In the cloud world, it is pretty easy to get data on how users are interacting with your solution. This speeds up the product development cycle, even up to 33%.

This means less dev time needed and lower development costs.

#10 Attract more easily external investors

If I was to invest in a software solution that is claiming that it can scale to worldwide levels, I would find it very strange if it was an on-premise solution. VCs and tech due diligence consultants show a heavy preference for cloud-based solutions vis-a-vis the on-premise ones.

Moreover, going to the cloud will help you move to a recurring revenue business model. And there are companies such as Pipe, which let you sell in advance your recurring revenue to investors, and fuel your growth without VCs.

Recurring revenue on cloud solutions is the new sought-after currency

#11 ISO and other Certifications

A lot of software solution vendors have to comply with ISO certifications and there are audits to pass every now and then. Being on the cloud, vs being on-premise makes it much easier to comply with the requirements of the audits. Usually, auditors just pass through to the next steps when they hear that your data are on the cloud, secured by a trustworthy vendor.

#12 Security

With the on-premise solutions, you have to take care of all the security work to keep your data and applications safe. In the cloud world, this becomes less troublesome.

What is AI (for non-technical people) and how to use it in your company

Below you can find my notes from a very interesting online course on AI, from Andrew NG, an AI team leader at some of the most prominent AI projects in the world, which is available here.

So, let’s have a look at what AI is all about and what AI can do for your company.

Let’s start with defining what is Machine Learning, the most common tool of AI today.

Machine Learning

What is Machine Learning? It is just a tool of AI. And what does this tool do? The most common use of this tool (ML), is what is called “Supervised Learning”. This means learning how to go from a Point A to a Point B. Or, from an input to an output.

Here are some examples of Supervised Learning applications:

  • Giving to a spam filtering application an email (input) and deciding if this a spam or not (output).
  • Giving to a speech recognition an audio file (input) and getting the text transcript (output).
  • Giving to a translation application an English text (input) and getting the Chinese translation (output).
  • Giving to a self-driving car image and radar information (input) and getting the location of other vehicles (output).

The trick here is obviously how to get the best possible output from my input. To do that, I need to train the system so that it develops its own brain (neural network). And the more data I have to train my application, the better the output will be. If I use 50 images of dogs to train a dog image recognition application to decide if an animal is a dog or a wolf, the system will not behave as well as if I use 10.000 images of dogs (and wolves) to train it.

And this is why everyone says that big data is the new oil. If I know how to use my large datasets correctly, I can develop several “brains” in my company, to perform tasks for me.

How do I get the data I need to train my application?

Let’s suppose that you want to build an AI application that decides if an animal is a cat or not. A cat detector. There are three ways to get the data to train your application.

Method 1: Manual Labeling of the Data

This means that you take 1000 pictures of cats, dogs and other animals and then you label each one whether it is a cat or not. You then feed this information to your AI application. Obviously, this method needs a lot of work.

Method 2: Observing behaviors

I might have a video recorder that monitors the animal and observes the behavior. Does the animal jump 3 meters up to a tree? Does it sleep on the top of your refrigerator? Does it sleep many hours close to the fireplace?

Method 3: Downloading the data from the web or getting them from a partner

For many applications, it is often easy to find large data sets available for free on the web. E.g., stock market prices, real estate prices, images, temperature data, seismic data etc.

How to make good use of your data for AI applications

A lot of companies think “I have a lot of collected data, so that means I can throw them to an AI team and they can create something AI-fantastic for me”. That’s not always true. More data is better than less data, but a lot of data doesn’t mean necessarily that you can build a useful AI application.

As an example, I may have a factory, and an IOT sensor that gathers data from my engine every 10 minutes. This is obviously a lot of data after some years of operation. But if I want to build a predictive maintenance application that tries to forecast when my engine has the risk of failure and needs to be services, I might need data from my engine every 1 minute. Or every one second.

So, the recommendation here, is that once your IT team has some data, you give them to your AI team, and then the AI team comes back with recommendations about how to collect data that will be useful for what they are trying to build.

Moreover, the data that you want to feed to your AI system have to be cleaned. There must be no wrong labels, or missing information because the AI system will learn the wrong things then. It’s like training my kid to learn a new language and teaching to spell wrong the word “apple”. Or by mistake, teaching her to see an apple and calling it an “orange”.

The AI applications work with two types of data: The structured and the unstructured data.

Structured data is whatever can be put in an excel table: Prices of houses, square meters, number of bedrooms.

Unstructured data is images, videos, audio and text. Videos of houses, images of furniture, text descriptions of houses for rent.

Terminology of AI

The most common terms you will hear in AI, are “Deep learning”, “Neural Networks”, “Machine Learning”, “Data Science”.

Machine Learning vs Data Science

Let’s start with the difference of Machine Learning vs Data Science projects. Let’s suppose that you have an excel with data about houses: Number of bedrooms, square meters, year built, year of being renovated and prices.

A machine learning project would create a SOFTWARE that helps go from point A to point B. E.g., a software in which you input the data of number of bedrooms, square meters, year built and it suggests the right price.

A data science project would help you create a POWERPOINT with insights about the data. E.g., you may look into data and find out that if you have renovated your house in the past 5 years, you win a 15% premium in the price. So, data science projects help you create insights that drive business decisions.

Neural Network vs Deep Learning.

Actually, this is the same thing. It’s just that Deep Learning is the new way to call a neural network. Maybe it sounds more impressive to say that I deal with “Deep Learning” than “Neural Networks” as a scientist.

And what is a neural network or a deep learning system? It is actually a big mathematical equation, that tries to help us go from point A to point B. E.g., if I have the input of house data (number of bedrooms, square meters, year renovated) as Point A and I would like to predict the price (Point B), the system that does the calculations and tries to do that is called a “Neural network” or a “deep learning system”.

The term “neural” comes from the “neurons” that we have in our brains. So, as our brains have neurons that connect with each other (well, at least usually….), and try to create an output based on specific input, the same is the task with an Artificial Neural Network. Of course, this is just a metaphor and the actual way of operation of a human neural network has nothing to do with the artificial neural network way of operation.

Other buzzwords in AI are “Unsupervised Learning”, “Reinforcement Learning”, “Knowledge graphs” etc. These are just other tools to make computers think smarter. But the two most important AI tools are “Machine Learning” and “Neural Networks”.

Summarizing, Neural Network (or deep learning) is a subset of Machine Learning, which is a subset of AI.

What Makes an AI company?

The previous technology era has been the “Internet era”. And now we are into the “AI era”. In the internet era, if I had let’s say a shopping mall and threw in a website, this wouldn’t mean that I have become instantly an internet company. An internet company in the retail (such as Amazon for example), would actually use all the good staff that internet was made to provide. This would be A/B testing for products in my website, short iteration times in launching solutions and products on the web, and push of decision making to the engineers and the product managers, instead of having the CEO make all important business decisions.

Similarly, in the “AI era”, throwing a Deep Learning system in my company doesn’t make me an AI company. To get closer to becoming an AI company, I will need to:

  • Have a strategy on how to acquire data. I might even have to launch products and solutions that don’t make money, so that I collect the strategic data I need for other solutions that I can monetize.
  • Build a unified data warehouse: If I have several different databases, that are supervised by many different owners, an engineer could never built a solid AI system. All the data need to be brought into a single place.
  • Automate whatever can be done by a computer vs a human.
  • Create new roles in my company, such as the Machine Learning Engineer.

The multi-year process to become a good AI company for Microsoft, Google, Baidu and other companies has five steps:

Step 1: Create a few small AI projects. Have your teams create them, so that everyone gets an idea what the company could do with AI. This could be done by an internal team of engineers or you can outsource it to an external team.

Step 2: Build an in house AI team.

Step 3: Provide AI training to a broader set of employees, including engineers, managers, business decision makers.

Step 4: Develop AI strategy

Step 5: Develop internal and external AI communications

What Machine Learning Can and Cannot Do

The rule of thumb is that whatever a human can do with giving it one second to think, we can automate it with AI supervised learning. For example, if a human can recognize a scratch in a phone in a second, we can automate it. If a driver can find the position of other cars in a second, we can automatize it. If I can recognize the speech in a couple of seconds, we can automate it with AI.

AI cannot successfully do things that take a lot of hours for a human to conduct. As an example, a human would need many hours of thinking to write a 50-page report with a market research analysis. It would be impossible for an AI system to do this.

In another example, if I am building a self-driving car, AI can help me build a system where I put as input the data (images, radar information) from a car in front of me and the output would be the answer to “where is the car”. What AI could not do, is to understand the intention behind a gesture. If I see a police man raising his hand, a human being could easily understand that this is a “stop sign” but AI could not figure out the intention. The same goes if I see a bicyclist raising her hand to the left, signaling that she wants to turn left. An AI system could not understand the intention behind this. The problem here for AI is that there is no parity in the interpretation of human gestures. Gestures might mean different things in different cultures and there are many variables of gestures for an AI system to work effectively.

Obviously, if I narrow down this system to the level that a Kinect game could understand that when I raise my hand up it means “up”, this is possible for an AI system to handle.

Building AI Projects

Workflow of a Machine Learning Project

The key steps of an ML project are three: Collect the data, Train the Model and Deploy the model. Let’s see the example of trying to build a speech recognition ML for Amazon’s echo, to recognize the “Hey Alexa” initiation command.

The first step would be to collect the data. That means, I would have to ask 1000 people to say “Hello”, or “Hi”, and “Alexa” and get the audio clips.

The second step would be to train the model. I want my system to understand if someone said indeed “Hello Alexa”, so as to initiate the speaker system. This means that I will need to build a supervised learning system, that has as input the audio files I collected from people saying “Hello”, Hi”, “Alexa” and has as an output whether the user said indeed “Hello Alexa”.

The third system would be to deploy the system. So, this would be in our case to deploy the software in a speaker and ship it to the users. At this point, I might have problems with my application. As an example, I might have used American accent audio files to train my model, and ship it to Greece, where people speak English with a different accent. In this case, I would need to collect the new data again (from Greeks speaking English), update my model and relaunch it.

Specific AI Examples per Industry or Job Function

Here are some examples on how ML could help several job functions or industries:

  • Manufacturing: ML can help with visual recognition of defective products in the production line.
  • Recruiting: ML can help with deciding which CVs to look at.
  • Marketing: ML can help with A/B testing of websites
  • Agriculture: ML can help with precision agriculture. E.g, recognizing a specific spot in the field that needs to be sprayed. Moreover, ML can help with crop analytics.
  • Sales: ML can help with prioritizing leads to contact in an organization. Or with sales forecasting.
  • Travel: ML can help with supporting travel chatbots to provide answers about travel destinations.

How to Choose an AI project for your Business

AI can do a lot of things. But what is relevant to your own business? Getting ideas about AI projects doesn’t happen by luck so you need to work for it. You will need to brainstorm, and to do it effectively you will need two teams: An AI team and a Domain Experts team. E.g., if you are brainstorming ideas for AI in your marketing, you will need the AI team that knows what is possible to do with AI and the marketing people (product managers, online sales managers, etc) in the same room.

The suggested brainstorming framework for an AI project, has the 3 following axis:

  1. Think about Automating Tasks and not automating jobs. For example, if you are brainstorming on AI projects for your call center, think about the several tasks that take place (picking up the phone, searching the CRM, updating the CRM, opening a support ticket, sending emails, issuing refunds, call routing, email routing etc) and which of these can be automated. In this case, call routing or email routing would be the most appropriate for an AI project.
  2. Ask the question:  “What are the main drivers of business value for my company?”. If I can find an AI project that supports one of these drivers (e.g. customer satisfaction, technical support, speed of delivery), the value will be tremendous.
  3. Ask the question: “What are the main pain points of my company?”.

It is important to note here that you don’t necessarily need big data to start an AI project. Even with small datasets you could start seeing meaningful results. The amount of data you need for an AI project is problem-specific and you will need to ask your AI expert about how much data you need to train the model.

How to decide if it makes sense to start a specific AI project

An AI project might take some days or some months of work to get completed. Before jumping on it, you will need to do your due diligence on a technical and business level.

-Technical Diligence: Talk to your AI experts to understand whether it is possible on a technical level to achieve the desired outcome. As an example, the Word Error Rate for humans is 4%. Meaning, that out of the 100 words that a person listens to, 4 of them might be misinterpreted. AI has helped systems to achieve the same error rate. But if you expect that you can do 0.01% Word Error Rate, you need to break the world record many times to do that and on a technical level it would be unachievable for your company. The second question here is “How much data do I need?”, and “Do I have a way to get this amount of data?”. The third question to answer is “How much time do I need and how many engineers to build this AI solution?”.

-Business Diligence: You will have to quantify with real KPIs, what would be the business ROI of your AI investment. So, if I indeed managed to succeed with my AI marketing project, and I can drive 10% more website customers to the shopping cart, how much additional money would that mean for my business?

-Ethical Diligence: AI can do a lot of things, but would I want to use AI to make gamblers who have an addiction problem, visiting more often my betting site?

Should I do my AI Project in-house or Outsource it?

Both approaches could work. With outsourced ML projects you get faster access to talent and if you run successfully a couple of AI projects, then you can start building your own internal AI team. Data Science projects on the other hand are usually built in-house. This is due to the very close business input and insights you need.

What Data will an AI team ask from you as a business owner?

If you start working with an AI team, you will be asked for two types of data. A specific dataset to be used as a “Training Dataset”, and another dataset to be used as a “Test Dataset”.

Should I Expect 100% Accuracy in my AI solution?

No. It’s a common mistake that many expect AI to be 100% accurate, e.g. in understanding language, images, automating tasks, finding defective products, email and call routing and other applications. This is not the case as there are limitations to AI, due to data mislabeling, ambiguity in data, wrong data input and other. Instead, you should focus on an accepted level of accuracy that would provide huge business value for you. Yes, you would love to have 100% accuracy, but maybe 95% will hit all your business targets as well.

AI Technical Tools

Some of the most common open-source frameworks used for ML, are TensorFlow, PyTorch, Keras, MXNet, CNTK, Caffe, 

PaddlePaddle, Scikit-learn, R or Weka. A lot of research publications are available at Many teams publish their code on Github (

Moreover, an AI project can be deployed on-premise, on the cloud, or on Edge. On-premise means that you install it on your own servers at your company, Cloud means that you use the infrastructure of a cloud provider (e.g. Microsoft Azure) and Edge deployment means that you put a processor where the data is (e.g. if I have a smart speaker at my home, I put a processor inside the speaker instead of sending the files via the cloud, getting the results and resending them back to the speaker). Generally, the world is going onto cloud solutions for AI and to Edge deployments where cloud doesn’t make sense (e.g. self-driving cars, smart speakers etc).

Roles in an AI Team

Usually, in a large AII team you need Software Engineers, Machine Learning Engineers, Machine Learning Researchers, Data Scientists, Data Engineers, AI Product Manager. Of course, you can always start with just a Machine Learning Engineer.

Further Reading

AI Customer Stories

Is it possible to startup if I already work as an employee?

Often, the startup ideas come from employees that see a gap in the market they are working with. E.g., an employee at a corporation who finds an opportunity to create a mobile app in the tourism area and gets on board a couple of team members to work together on the idea, on a part-time basis.


According to my observations, there are limited possibilities for a part-time team to have a successful launch of their project. What happens usually is that the part-time group starts with a lot of enthusiasm and devotes weekends and late after-work nights to the project. As time goes on, the team discovers the problem of the different rhythms within the team. One co-founder has to devote the weekends to the family, another one might be too pressured with work for a particular month and so on. This creates the problem of one team member delaying the rest of the team. And instead of hitting the first plan of creating a minimum viable product in 4-5 months, this seems to go far in the future.


Another common pattern is that a part-time member of the initial team starts to lose interest in the project as he sees that it’s much more difficult and time-consuming than what he thought in the beginning. If this person continues to be in the team but with devoting less and less time on the project, then the team psychology starts to change: “All three of us have 33% in this idea and the two of us are devoting so much time, sacrificing weekends with our friends and family, while the third doesn’t seem to be that committed”. This often leads to difficult discussions inside the team which could lead to having a team member leaving the team. Obviously, this has the cost of having to onboard a new part-time member and lose even more time.

I am not saying that part-time teams cannot succeed. But you should be three times more careful when joining such an effort. Here are some things you can do to mitigate some risks when you start with a part-time team:


1. Get the approval of families: This is especially important when having team members who have families and kids. Each co-founder has to explain to his significant other that joining such an effort, might require that he devotes many weekends and late nights to the idea and not to the family. And that there might be no revenue coming out of this at the end. Family support has to be granted with a clear view of the risks and the opportunities since day one.


2. If it’s something simple, pay it and don’t build it: If you want to build e.g. a mobile app for finding the best parties in the town in real time, you may think that you have to recruit a developer and a web designer in the part-time team to build it. Well, do your research. If you can buy the first prototype to test your idea with 2-3k USD through or a freelance developer, why should you bring additional part-time members in the beginning? Just bite the bullet, spend some money and get this off the ground as soon as possible. If you see traction, you can later recruit more members in the team.


Of course, if you want to do a sophisticated machine learning algorithm or a new social network or a very demanding software as a service, this is not possible, yet many times the startup ideas are not based on the excellence in technology but in the market connections or another unfair advantage that the team might have.


3. Create Rules of Disengagement in the team: Part-time teams usually don’t discuss what are the rules of leaving the startup, in the beginning. They feel that this will ruin the enthusiasm and the positive spirit of the team. It would be advisable to get in such an agreement with other team members as soon as possible. As an example, agree that you are all going to spend the next 12 weekends to create the first version of your mobile app and recruit 100 beta users. And that at the end of these 12 weekends you will meet to discuss if someone wants to leave, or if the team wants someone to leave due to low commitment. In that case, you could agree that the person that has to leave will let all his equity go, without requesting compensation for the hours worked. Or that he could keep a 2% in the company without doing anything in the future, instead of the 33% that possessed on day 1.


4. Create Sprints. Rest. Create Some More Sprints: It’s not easy for 3 persons who work 9 hours per day as employees to be on high performance all the time. Agree to the rhythm of the startup. As an example, agree that you will spend this week doing a sprint to gather 200 questionnaires for your market research. Let 3-4 days pass without doing anything. Create a second 4-day sprint of setting up the landing page, connecting analytics, creating online ads, your first blog posts and the press media kit. You get the idea. Create a healthy tension in the system with tight commitment and deadlines and execute on that as your whole life depends on it.


5. Manage the Emotional Flow: Cash flow is important, yet emotional flow might be even more important in the beginning. Try to find some ways to have fun as a part-time team. Celebrate a small victory, such as getting 100 beta testers with a night out with the team and everyone’s families. Get together as a team and work on some ideas that can help you keep the spirit high when the tough times come.


6. Have complimentary skills in the part-time team: If you are a business expert great in representing your startup and getting it off the ground but you don’t have industry connections, don’t recruit a friend of yours in the team just because you like him and he likes the idea. Get an industry expert, who could bring the connections that you don’t have.

What is a fair share of equity that each startup member should get?

A common question between co-founders when they start a new venture, is how to split their shares in the startup.

Let’s suppose that you have an idea for a startup and you decide to co-fund your startup with another two persons. So, you are three co-founders. Should you split equally the equity in your startup and get 33.3% each?

Actually, you should not do it. Although this is the way most startups choose to start, especially if the other two are friends of yours, you should take under consideration several factors to determine what is the fair way to split the equity in your startup.

Here are some questions to take under consideration, when deciding about the split of equity among co-founders:

  • Who had the original idea?
  • Who is the CEO?
  • Who pitches to investors?
  • Who writes code?
  • Who is working part-time, with the promise to turn into full time when the startup takes off?
  • Who runs the online digital marketing campaigns and does content marketing with blog posting?
  • Who is paying at the moment for the business expenses?
  • Who has the most connections in your targeted industry?
  • Who could become the CTO of the company if you hire many developers in the future?
  • Who is the co-founder that if he leaves, you will have much fewer possibilities to raise funds?
  • Are you going to give a small percentage to advisors? (Typically advisors get 1% in the company)
  • What is the market value of each co-founder at the moment? Meaning, what would be his/her salary if he was employed by a company?
  • How many startups has each founder founded in the past? What was the biggest achievement that they had in the previous startups?
  • Has any co-founder participated in an accelerator in the past?
  • How many hours has each co-founder contributed to the project up to now? And what would be the cost per hour for his/her expertise?


Answering all these questions will provide to the co-founders team with a better understanding of what should be a fair split of equity. There is no formula that works perfectly to define the equity split. You will have to make up yours. Having said that, there are several online tools that could be used to help you understand the gravity and the “special weight” of each of these factors.

Here are some of them: