Amplitude says its Wave agent, a proactive product assistant that turns usage data into build plans, is roughly 80% powered by OpenAI models and can help ship fixes in 1 to 2 days, per OpenAI News.
Can analytics skip the backlog and write the ticket itself?
On its own homepage Amplitude says yes, with card-driven sends up 160% versus control after one day.
Wave takes a product area and a goal, then combs session replays, analytics, feedback and code to rank fixes. It drafts a mockup and a technical plan with acceptance criteria for coding agents including Codex, then measures results. "AI can now find useful work we never thought to assign it," said Spenser Skates, chief executive of Amplitude, which built Wave to close its own loop. Jacob Newman, principal AI product manager at Amplitude, said the team tested it where 80 of 100 replays showed users typing instead of clicking suggestions, so Wave proposed personalized cards while keeping free typing.
Senior principal AI engineer Eric Carlson called it proof that small ignored fixes can lift adoption, according to the company.
Amplitude assigns GPT-6.1 Sol to deep investigations, GPT-6 Luna to high-volume scans and checks, and GPT-Image-2.5 Flare to wireframes from replay screenshots. Sol scored 4.06 out of 5 on the Opportunity Investigator test against 3.62, 3.52 and 2.94 for alternatives with a 54-minute median time, according to Amplitude. On the Data Science Agent test Sol scored 67.5 against 69.3 for a rival but cost $1.28 per dispatch versus $2.63, according to Amplitude. Flare scored 80.3 against 63.9 for another image model across 15 opportunities, according to Amplitude.
Most teams send bug fixes straight to agents while bigger bets get human review, and among customer launches one in four went live same day, half in under four days and two in three within a week, according to Amplitude. Internally teams have also used Wave to save text answers, clarify setup rules, change how agents investigate requests and fix docs friction, according to Amplitude's Wave blog. The shared chart is illustrative rather than individual observations and the 160% lift is a one-day readout from the company's own test. Product leaders should watch whether that speed holds outside Amplitude's homepage with human review still in place. Not financial advice. Markets move fast.
