OpenAI · AI product · Mid · 45 to 60 minutes

Managing a major ChatGPT model change

OpenAI builds AI models and products such as ChatGPT that people use for writing, coding, learning, and other tasks. Answer this product management case and get it graded line by line against a weighted rubric.

product judgment
trade off analysis
user empathy
go to market thinking
metrics design

The situation

A senior leader on the ChatGPT product team asks you for a recommendation before a new foundation model is made the default experience in ChatGPT. The new model is expected to improve some important capabilities, but it may also change how the product feels in day to day use, including response style, consistency, and how well it fits routines that users have already built. ChatGPT now serves a very broad audience, from casual users to people who rely on it repeatedly for work or study. The company wants to keep improving models quickly, but it also knows that a model change is not just a backend upgrade if users notice it and have to adjust their habits. You need to advise how to launch the change in a way that balances product progress, user trust, and operational simplicity. The hard part is that a faster rollout helps standardize the experience and accelerates learning, while a more cautious rollout may reduce confusion but creates complexity and slows iteration.

What to produce

Write a 1 to 2 page product decision memo.

Your answer must

Define the user segments most affected by a noticeable model behavior change, and identify the key jobs or routines at risk for each.
Compare at least three rollout approaches, such as immediate default replacement, staged rollout, opt in access, or temporary fallback access, and recommend one.
Specify success metrics that include product adoption plus user trust or satisfaction signals, and name leading indicators of backlash or confusion.
Describe a launch and communication plan covering before launch, launch week, and post launch follow up.
List the top risks and trade offs of your recommendation, including how you would monitor and mitigate them.

Constraints

Do not assume OpenAI can keep many long term default models running in parallel inside ChatGPT.
Your plan must work within the existing ChatGPT product, not by creating a separate new product.
You cannot rely only on benchmark gains, because users may judge the change by felt quality and workflow fit.
Expect continued model iteration after this launch, so the approach should be repeatable.

Format

Use clear sections with headings: user impact, options, recommendation, launch plan, metrics, risks.

How OpenAI actually played it

Uses the GPT-5 launch reaction to discuss how people adapt to AI quickly and then feel the loss when a familiar model changes. The author draws product lessons about change management and expectations for AI products.

People adapt fast to new capabilities and then resist losing them
Handle model or behavior changes as a user change, not just an upgrade
Expect the honeymoon phase of AI to end

Built from 3 lessons from ChatGPT's launch (And what your AI product can learn from it) by Kristen Berman (Product Teardowns, Irrational Labs).