OpenAI and Amplitude: Wave Links Product Analytics to Coding Agents

Amplitude’s Wave product agent is connecting the decision about what to build with the coding tools that can implement it. An OpenAI customer case study describes a workflow that investigates product usage, prepares changes and measures their effect, while leaving teams control over what reaches users.

In an article published on 12 November 2026, Spenser Skates, CEO, Amplitude said: “OpenAI’s models can investigate a problem, write and run code, and use the results to decide what to try next.” (source).

In the case study published on 12 November, Amplitude estimates that Wave can help teams ship product improvements in one to two days, compared with roughly a month. Read that figure as Amplitude’s reported experience rather than a delivery commitment for a new project.

Teams supply a product area and goals. Wave brings together analytics, session replays, feedback and code to identify opportunities, develop proposed solutions and coordinate implementation. Plans can be sent to coding agents, including Codex, with technical context and acceptance criteria. The system then tracks the outcome.

A homepage experiment measures suggestion-card use

The case study describes a test on Amplitude’s own homepage. Wave proposed personalising suggestion cards using recent charts, dashboards or the page a user had arrived from. The plan retained free typing and generic cards when personalisation context was unavailable. After review, the team handed implementation to a coding agent and ran an A/B test.

According to the experiment readout Amplitude shared with OpenAI, messages sent through those cards increased 160% relative to control after one day. The measured outcome concerns that interaction on Amplitude’s homepage. Readers should not substitute retention, revenue or overall engineering output for the metric the company reported.

OpenAI describes customers using Wave in beta. It says teams can choose their level of automation, with most sending bug fixes straight to coding agents while new features and larger changes receive human review. Those deployment choices matter when interpreting the claimed speed of the process.

In an article published on 12 November 2026, Eric Carlson, Senior Principal AI Engineer at Amplitude said: “It is a great example of how non-groundbreaking things that just get ignored can increase the adoption of features in unexpected ways.” (source).

Evidence and approval travel with the proposed change

Amplitude’s current Wave product page describes how an opportunity carries supporting evidence, proposed designs and a technical plan into existing workflows. A reviewed opportunity can be routed to Linear, Jira, GitHub or a coding agent. Target metrics and experiments connect the resulting change back to its original hypothesis.

The page says customers need an eligible Growth or Enterprise plan, meaningful Product Analytics and Session Replay data, and product and engineering owners who can review recommendations and act on them. Connecting a repository adds implementation context. It also says demonstrating business impact depends on how quickly a team ships a change and gathers enough data to measure it.

That is more specific than a promise to generate code faster. The proposed mechanism links a product problem to evidence, an implementation plan and an outcome measurement. Whether a particular change succeeds remains something to test, rather than a consequence guaranteed by handing it to an agent.

In Amplitude’s November earnings-call transcript, chief executive Spenser Skates cautioned that Wave was not fully automated and described a hands-on approach with a smaller group of customers. Integrating it into existing processes and ways of working remained part of the adoption challenge.

MarketBeat’s coverage of that earnings call also highlighted the setup requirements and selective rollout, attributing those plans to Skates. Use that reporting as context for the commercial approach, and judge the later case study’s speed and engagement figures as company-reported results. The practical distinction is between generating a plausible improvement and proving that the shipped change helped users.

This article was produced by AI agents under Business 2.0 News editorial standards.