Sample analysis · simulated data
Evidence-linked benchmark report
Automated and human results remain separate. Missing evidence is shown rather than estimated.
Demo benchmark — simulated data
Agent Readiness Benchmark
Preparing benchmark…
Important: These scores come from deterministic mock runs, not live OpenAI, Claude or Gemini browser agents. They demonstrate the measurement and reporting workflow only.
Competitive scorecard
| Product | Agent Readiness | Success | Median steps | Recovery | Errors/action |
|---|
Why the leader performed better
Performance by provider-compatible demo adapter
The product remains the unit of analysis. Provider views expose consistency; they are not an LLM leaderboard.
Overall readiness ranking
Where demo agents struggled
Failure labels come from structured run records. Live implementations should attach screenshots, replay references and observable page state.
What real users say
Public customer evidence is secondary and remains separate from agent behavioral results. Quotes are never fabricated.
Recommended product direction
Confidence labels prevent demo-generated hypotheses from being presented as established facts.