PostHog · AI product · Mid · 45 to 60 minutes

Should PostHog keep its AI chat assistant?

PostHog is a product platform that helps teams build, test, and understand software, with tools such as analytics, feature flags, and session replay. Answer this product management case and get it graded line by line against a weighted rubric.

product strategy
AI UX
prioritization
metrics
tradeoff analysis

The situation

You are the PM responsible for AI experiences at PostHog. The founders want a recommendation before the next planning cycle on where the team should invest next in AI. PostHog already has a shipped assistant that users access like a chat box. It is useful in some cases, but users and teammates increasingly talk about it as "the chatbot," and there is concern that this label may shape how people understand the whole AI effort. One group wants to keep improving the assistant, because it is visible, already in market, and easier to measure. Another group thinks the bigger opportunity is to make AI work through existing product tools and workflows, so users can get value without going to a separate conversation UI. The tension is that the current assistant is a real product surface with momentum, while a more embedded approach may better match PostHog's broader product direction but could be harder to explain, build, and validate. Assume you have only lightweight qualitative feedback and existing product knowledge, not a new research study or fresh usage dataset.

What to produce

Write a 1 to 2 page product strategy memo.

Your answer must

Frame the decision as a product problem, not just a model or feature quality problem, and state 2 to 4 assumptions explicitly.
Compare at least 2 credible options for PostHog's AI experience over the next 1 to 2 quarters, including continued investment in the chat style assistant and a more embedded approach.
Identify the main user segments or workflows affected, and explain how each option changes who can successfully use PostHog.
Recommend one path, with key trade-offs, major risks, and what should happen to the current assistant during any transition period.
Define success metrics, guardrails, and a validation plan with milestones or decision triggers for the next planning cycle.

Constraints

Base your answer only on the prompt and generally known facts about PostHog, do not invent real company data or outcomes.
Keep the recommendation realistic for a product area roadmap over the next 1 to 2 quarters, not a multi year rewrite.
Do not assume a separate human operations team, manual concierge work, or major model breakthroughs are needed to make the plan work.

Format

Use short section headings and bullets where helpful, such as problem, options, recommendation, metrics, and risks.

How PostHog actually played it

PostHog retired its chatbot style assistant because users still saw the AI as just a chat box while the plan was to weave AI through every product. The post covers how the team moved to an agent that builds on top of product tools and what that changed about who can use the product. Useful for seeing how an AI feature gets reframed after launch.

A feature people label as a chatbot can cap the vision of the whole AI strategy
Retiring a working feature can be right when it frames the product too narrowly
Design AI as a system across the product, not a single assistant surface

Built from Why we killed our AI product assistant by Cleo Lant.