Amplitude · AI product · Senior · 45 to 60 minutes

Should Amplitude recommend what teams build next?

Amplitude is a digital analytics company that helps teams understand product usage and customer behavior. Answer this product management case and get it graded line by line against a weighted rubric.

Strategic framing
AI product design
Prioritization
Measurement
Risk management

The situation

Amplitude is exploring its next AI product bet. The chief product officer and chief AI architect ask you to recommend whether Amplitude should move beyond dashboards and copilots into a proactive system that scans customer product data, identifies likely growth or retention opportunities, and helps teams act on them. Why now: many customers already have more analytics than they can operationalize, and new AI tooling is making execution faster, which may shift customer pain from building changes to deciding which changes matter most. The hard part is that bad suggestions could waste roadmap capacity, erode trust, or push teams toward local metric gains that hurt the broader business. The opportunity is also technically messy, because product usage data alone may not be enough to judge what is worth building, and Amplitude would need to decide how much context and workflow to own. Assume Amplitude can fund a two quarter pilot with a small cross functional team. You need to propose the product direction, where humans stay in the loop, how success would be measured, and what to build first.

What to produce

Write a 2 page product strategy memo.

Your answer must

Define the primary target user, the core job to be done, and why this problem is becoming more urgent now.
Recommend the product scope for a first release, including what signals or data sources it should use and what output it should produce for users.
Describe the end to end workflow from surfaced opportunity to team action, including where review, approval, and explanation are required.
Propose a phased plan for the first two quarters, with what you would launch first and what you would defer.
Specify success metrics and a learning loop that connects recommendations to actions taken, experiments run, and measured outcomes.
Identify the top risks and tradeoffs, including trust, false positives, privacy, and feasibility, and explain how you would mitigate them.

Constraints

Keep the proposal credible for Amplitude's existing analytics customers and current data footprint.
Enterprise customers will expect explainability, permissioning, and an audit trail for AI generated recommendations.
The pilot must show value within two quarters and cannot rely on direct write access into customer production systems.
Data quality across product, business, and code context will be uneven, so the design must work under partial coverage.

Format

Use headings: user and problem, product strategy, workflow, metrics, risks and sequencing.

How Amplitude actually played it

Amplitude's chief AI architect describes Wave, an agent that watches product data for signals, proposes improvement opportunities that teams approve and ship, and then tracks the experiments and outcomes. The post argues the bottleneck has moved from building to deciding what is worth building. It is a long look at designing a proactive AI product that closes the loop from signal to measured result.

When building gets cheap, the hard problem becomes choosing what to build.
Keep a human approval step between a surfaced opportunity and shipped work.
Track downstream experiments so the agent learns from outcomes.
Combine product data, business data and code context for better suggestions.

Built from How we built a product that tells us what to build next: Inside Amplitude Wave by Eric Carlson, Amplitude.