Case studies
What a verification gate looks like when it is actually installed.
Three engagement patterns, one per kind of team: a studio delivering to clients, an engineering team merging agent-written code, and a publisher handling generated imagery.
Four done, one to go.
Independent product studio · Design and build, eight people
Signing off AI-assisted client work without hiring QA
Twelve to fifteen pages a month, half of them drafted with an assistant, one senior designer reading copy between other work. The studio did not want a QA hire; it wanted a step that took twenty minutes and caught the things a client would notice.
Read the engagement- 31
- Findings across the first five deliverables
- 6
- Blocking, all caught before client review
- 20 min
- Added per deliverable
B2B SaaS team · Series A, fourteen engineers
Reviewing agent-written pull requests without stalling the queue
Roughly sixty percent of pull requests arrived from an agent. Static analysis passed on all of them, because static analysis reads code without running it. Two of the four incidents that quarter came from errors that had been caught and silently discarded.
Read the engagement- 60%
- Of pull requests written by an agent
- 4
- Signature defect classes prioritised
- 1st pass
- Where blocking findings were caught
Digital publisher · News and features, thirty staff
Publishing AI images with provenance that survives the CMS
The image desk was receiving generated art and stock alternates several times a day. Editors resized, cropped and re-exported, which stripped the provenance metadata the desk had been relying on to answer the only question that mattered: where did this come from?
Read the engagement- 3x
- Re-exports before a typical hero image publishes
- 0
- Manifests surviving the old export path
- Image desk
- Where the gate now sits
Client names are withheld and details generalised. Figures describe the pattern of these engagements, not audited outcomes.
Your team
Every one of these started with a single report.
Run one file through the console, and see whether the findings match what your process already catches.