How would you measure the success of PostHog AI, the in-app AI analyst formerly called Max AI?

Introduction

PostHog AI is the in-app AI analyst that lives inside PostHog. It began as "Max AI" and was renamed to PostHog AI as it moved towards general availability and paid usage, a change PostHog's own handbook cites as an example of PM work on positioning and pricing. Users can ask questions in plain English ("Which plans churn most after their first month?"), and PostHog AI builds insights, writes SQL against the data warehouse, finds session replays, and helps set up flags, experiments and surveys. It is charged through usage credits beyond a free monthly allowance. Measuring its success is tricky: an AI analyst can look busy while giving wrong answers, and chat volume can rise simply because the AI fails to answer the first time. The metrics must capture correct, trusted answers that lead to decisions.

Step 1

Product Context (3 minutes)

Key stakeholders:

• Product engineers and founders: want answers without learning PostHog's query builder or SQL.
• PMs, analysts and marketers: want faster analysis and less time on routine questions.
• PostHog: wants broader team adoption, more product usage and AI credit revenue.
• Customers' end users: their data is queried, so access controls and privacy matter.

User flow:

1. User opens PostHog AI from the sidebar or from inside an insight, replay or error.
2. Asks a question in natural language.
3. PostHog AI plans the analysis, picks events and properties, and generates an insight or SQL query.
4. User reviews the answer, the chart and the query behind it, then saves, edits or asks a follow-up.
5. Optionally, PostHog AI takes an action: creates a cohort, a survey, a feature flag or a dashboard.

Lifecycle stage: early growth. The product is generally available and monetised through credits, but usage habits are still forming and model quality is improving quickly.

Why it matters: PostHog has more than a dozen products. PostHog AI is the layer that lets a new user benefit from all of them without learning each interface. It is also central to the "self-driving product" strategy alongside Replay Vision, Inbox and PostHog Desktop.