The Platform PM

Glossary and sources

Field Guide

AI evals and observability: how teams measure AI quality

10

Glossary and sources

5 min read · Chapter 10 of 10

Words that mean something else here

TermWhat you'd assumeWhat it means here
EvalAn exam or a one-off reviewA repeatable test of AI behaviour: inputs plus a way to grade outputs
JudgeA personUsually a model scoring outputs; also called grader, autorater or LLM-as-judge. Reports should say which
RegressionA statistics techniqueA case that used to pass and now fails; a regression suite is the must-not-break set
TraceA log lineA tree of steps for one request; also a billing unit that each vendor counts differently
Grounded or faithfulTrueSupported by the source the system retrieved, even when that source is wrong or out of date
Online evalAn eval run on the internetScoring a sample of live production traffic, usually by a judge, with no known right answer
RollbackUndoReverting the prompt, flag or model version; doesn't undo actions taken or clear caches
PostmortemA full root-cause analysis with creditsAt labs, a narrative blog post; sometimes a short social-media post

What surprised me

Placeholders in your voice, drafted from the research and the earlier guides. Rewrite each with your own moment.

"Teams without evals ship regressions." In payments, a failing test stopped a release. Here the best-known regressions came from labs that had evals, and the evals passed.

"Scoring is the expensive part." A model judge scores 500 cases for a few dollars. The time went into a support lead deciding what a right answer looks like.

"Monitoring catches what testing misses." In observability an alert usually fired before customers noticed. In the AI incidents I read, outsiders noticed first almost every time.

"Human data is the eval budget." The billion-dollar human-data companies sell mostly to labs, and a product team's graders are usually its own staff.

Sources

Undated entries were read on October 12, 2026; "search result" means seen only as a search snippet or summary. Company figures are self-reported unless they come from a filing or a regulator, vendor surveys cover the vendor's own respondents, and list prices change often.

Model providers' guidance, docs and terms

Postmortems and incidents

Research on judges, evals and benchmarks

Practitioners and surveys

Tracing, tools and standards

Rules and regulators

Market, companies and pricing

Human data

Reused from earlier guides on this site

Field Guides are learning notes, not legal or compliance advice. Rules and fees change; check the cited primary sources before you act on anything here.