ChatGPT as a Service: What does it really mean to businesses ?

30th November goes back into history as a date when the world witnessed the true power of Generative & Conversational AI.

A record million users in a week, fastest to 100 million users to date and a monthly active user base of over 10 million is probably the best way to represent this innovation and the perceived value/impact.

This free research preview by OpenAI garnered world’s attention and took significant airtime coverage over cable & media networks. ChatGPT has become a board room conversation and top of mind for chief executives. When a neighbour who is not connected to IT and who has never spoken a word of technology, starts talking about ChatGPT as a cool innovation, it is undoubtedly an evidence on the virality of this innovation.?

With all the goodness that this research preview demonstrated of the true power of Conversational AI, the question remained of its usage within the context of a business and for key business scenarios. Enough of playtime/discovery, and we now have ChatGPT offered as a service in Azure OpenAI Service.

What does it really mean to businesses ?

ChatGPT is a brain (I rather call it a brain than a model) trained on a massive corpus of world’s data. As I write this post, the brain is trained on data that represents the world till Sep 2021. This trained brain interacts with the user in a conversational manner (dialogue format), which makes it possible to answer follow-up questions, understand complex intent, reason with data, challenge incorrect premises, and reject inappropriate requests. Truly profound in how this works and the right word is ‘magical’.

ChatGPT as a service on Azure is classic evidence of an innovation being completely democratized for ease of adoption.

While the public variant hosted at chat.openai.com is a closed system with no means to influence behaviour of the brain, the service on Azure OpenAI gives a window to instruct the brain (via a prompt) to operate with certain guardrails. This means you now have a mechanism to tell the brain ‘What to do and What not to do’, and then infuse this capability in business scenarios.

Let us take a quick example to better understand how a business can leverage this innovation and position them as key conversation end points for various personas.

Scenario: An automotive firm ABC’s (fictitious name) priority is to address user attrition and bring in stickiness to customer conversations at their website/app. They believe that providing right information in a human like conversation will bring in the much-required stickiness leading to customer delight, and fast track sales.

How can ChatGPT help here assuming that the brain is aware of ‘ABC’ automotive firm?

Prompt for ChatGPT Setup:

You are an ABC customer support agent whose primary goal is to function like an Automotive advisor. You are friendly and concise. Provide short responses. You only provide factual answers to queries, and do not provide answers that are not related to automotive industry. You also do not respond to questions that lead to competitive comparisons. However, you can compare within the brands under ABC. Restrict answer to brands under ABC. Do not answer anything outside of context of automobile as a subject.

Basically, we are instructing the brain what to do and what not to do. That’s it and you are all set to use this service in your applications. Isn’t this simple??

I am sure you have a few Qs but the one that I get asked often is:

Q: ‘Is there a way to present recency or custom data as a prompt as part of the setup of ChatGPT service?’

A: The answer is yes, and you can do so during query time and append to the Prompt for ChatGPT.?

For more on What and How, I highly recommend the following resources.

• ChatGPT is now available in Azure OpenAI Service
• The ChatGPT model (gpt-35-turbo)
• Quickstart - Using the ChatGPT API