AI Product Case Studies - 4
About Gemini Model training, Agentic Commerce and GPT group Chat Feature
We will be discussing 3 different case studies
What can be the most valuable data source to train Gemini?
Is ChatGPT group Chat feature a good Idea?
Is Agentic commerce is next big thing?
What can be the most valuable data source to train Gemini?
What do you think?
It’s Gmail.
Google can use Gmail Data to train their AI Model. This is an important dataset to teach AI about the Tone Calibration, Human Conversation resolving complex dependencies etc.
Most emails have a clear goal like scheduling, negotiating, buying which is perfect for training models to be next generation Agents that do things rather than just say things.
Emails are full of unsaid context — dates, meeting protocols, and social norms using which the Gemini can take the next leap. It’s basically Human to Human Conversation
Is ChatGPT group Chat feature a good Idea?
Chatgpt recently introduced a group chat feature that allows multiple users (up to 20) and an AI instance to collaborate in a shared space for brainstorming, planning, or decision-making.
What can be the possible reason to launch this?
Viralilty & Acquisition : To use a group chat, you have to share an invite link. This forces you to bring other people into ChatGPT who might not use it often.
Retention (Stickiness): It is easy to delete an app that you use alone. It is very hard to delete an app where you have active groups with friends or coworkers. It locks you in.
Data Goldmine: Training AI on how humans talk to each other is more valuable than how humans talk to a bot. Group chats give them data on social dynamics, arguments, and consensus-building. They way google can utilise the Gmail Data to train their model.
Is it good or bad?
The main point here is how openAI will treat people with different subscription tier like Free user vs Paid User.
If a Paid user generates a high-quality image (DALL-E 3) or uses an advanced reasoning model, everyone can see it. But if a Free user tries to do the same thing in the chat, the bot will likely reject them or give a lower-quality result. It creates a second-class citizen feeling in the group. It breaks the flow of conversation when one person can do magic and the other cannot.
Rather they should focus on releasing the Advertisment Model and that too personalised Ads. The first cut of the model should get out of the door asap like Netflix has recently launched. Chatgpt should cash out from 800 Million active users, they can’t wait for a perfect product for long.
Is Agentic commerce is next big thing?
Agentic commerce might not fly specially for the price sensitive customers.
Price sensitive customer browses 40 tabs , 10 shady website just to get 10% discount on the product they are buying.
This becomes even more difficult when the price of the items is high. In that case, they will spend a lot more time just browisng and researching before placing an order.
Agent however can help in the discovery but not a full fledge shopping. Good think about LLM is that it will be able to make the discovery of tail end category becomes much easier.
Now let’s take the case of low price ( No brainer / Replenishment ) sort of items. Ecommerce companies have already made Replenishment recommendation, Amazon has Subscription feature which has made the purchase of these items very easy and convenient.
For New Year, we are giving EXTRA 60% OFF on our AI PM Flagship Course for very limited Time
Coupon Code — NYE26 , Course Link - Click Here
Shailesh Sharma! I help PMs and business leaders excel in Product, Strategy, and AI using First Principles Thinking. For more, check out my AI Product Management Course, PM Interview Mastery Course, Cracking Strategy, and other Resources
 | A guest post by
|