BookmarkSubscribeRSS Feed

SAS GPT - A Personalized AI Data Analytics Chatbot

Started ‎02-08-2023 by
Modified ‎04-27-2023 by
Views 5,188
Team Name SAS GPT
Track AI, IoT
Use Case

Develop a platform where businesses can interact with their data by using prompts in plain English language and get the data analytics done at fingertips.


Watch this final video submitted for Hackathon

Technology SAS, Python
Region EMEA - Norway
Team lead Frode Karoliussen
Team members Mayur Jadhav (@MayurJadhav ) , Frode Karoliussen, Mitesh Bhosale ( @mit7 )
Social media handles

Mayur Jadhav: 

Mitesh Bhosale:

Is your team interested in participating in an interview? Y
Optional: Expand on your technology expertise  


Jury Video



Pitch Video


How do you envision it to be different than the current AI assisted data analysis tool?

I'd have more precise and clear while writing use case description but anyways I've just updated it now.

I'm not aware about any existing AI assisted data analysis tool in SAS that supports prompts or instructions in plain English. I would love to learn more about it if it already exists.


The simple objective here is, you would be able to ask questions like "How many employees are on leave today and who are they?"

The SAS GPT chatbot should be able to process the given prompts, generate sas code, execute it in back-end and present the desired results immediately. 

Welcome back Mayur! 

You will see a list of tools by Google "AI data analytics". I think I've seen a demo or two, but don't remember from which company. May not be what you have in mind.

Thanks @Peter7  

I'm glad you still remember my name! 🙂



I am looking for a team. DO you need other team members. My skills: Python, sql , Background : data science

Great topic

can you achieve this by combining SAS conversation designer and ChatGPT API?


SAS Conversation Designer | SAS

Chatbot SDK, out of the box KPIs | SAS Viya 2021.1.1 - YouTube

10 Innovative Use Cases of chatGPT | by SM Raiyyan | Jan, 2023 | Medium



number 1 is your project!

Hi, I am a Python developer and have a good knowledge of SQL. I have
knowledge of data science too. I want to join in your team for GPT project.

Hi @Atabarut ,


Thanks for sharing those resources on SAS conversation designer which is impressive chatbot and can be enabled in eCommerce or any website. This is definitely helpful for the end users. "SAS GPT" is an idea which is solely focused on generating heavy analytics specifically crafted for business users within the organizations who knows SAS and does little bit of SAS coding, use SAS VA reports, runs SAS EG projects, etc. 


Enabling chatGPT is a great idea but at this moment chatGPT sometimes write wrong SAS code. Also chatGPT doesn't have functionality to access lib, databases, connect to SAS, execute generated SAS code by given prompt and returns the results. There would be possibility in the future with API's but why we should dependent on chatGPT when we can build our own sasGPT.


What I'm aiming to build here is an independent SAS GPT tool exactly similar as chatGPT but for data analytics. 

Developing a language model in that sense and quality 

1. requires huge code set repositories with inputs and outputs

2. SAS - Python converted code matches

3. SAS - R

4. SAS - Lua examples etc

5. And very heavy computation on GPUs 

which equals to millions of dollars and seems not feasible right now, If you have the money that why not...


And the essence of a hackathon is first you need to build an MVP, quick and cheap...

It was great working on this use case. You can find our final video here! 


Any feedback, comments, or suggestions are always welcome! 🙌

Have to say that I think this was a really good MVP! The Jury video had enough technical details in the end to show how it was done. Do not see the SAS REST API used too often which was great. From a technical side, it does look like the MVP depends more on predetermined SAS code uploaded to a table that is retrieved based on pattern matching SAS functions for the input text and less around AI-generative code derived from NLP-trained data such as ChatGPT. May be a bit of a misnomer to call it SAS GPT. There could be a sense of concern regarding the ability to scale the application as a database becomes more complex and user inputs vary more often from the expected inputs. HOWEVER, given that this was a hackathon with limited time and resources, I would consider this very successful. Great job!

Thanks @CSmith_DECA  for your comments! 

Version history
Last update:
‎04-27-2023 04:30 AM
Updated by:


Available on demand!

Missed SAS Innovate Las Vegas? Watch all the action for free! View the keynotes, general sessions and 22 breakouts on demand.


Register now!

Article Tags