// verdict · checked 2026-08-11 · Generative orchestration, standard harness (2026 docs)
Copilot Studio agents
Microsoft
SharePoint energy, agent branding. The compliance checklist is genuinely autonomous.
- CAPABILITY
- 5/5 AGENT
- PROOF
- OPERATOR INSPECTABLEYou can run it and read the steps yourself — the record exists for whoever runs it
- TRACE ACCESS
- OPERATOR-OWNEDWhoever runs it holds the record; nothing is published
- LEASH
- Not measuredno execution record scored — measure it yourself
| LOOPSruns until doneyes | LLM planner decides step count per task at runtime; docs cite "dynamic plans" and a plan-execute-adjust loop. Plan complete is a terminal trigger, not evidence of a human-fixed sequence. |
| CHOOSESpicks its own toolsyes | Docs: the agent can choose the best tools, knowledge, topics and other agents from their descriptions. |
| ACTSchanges the worldyes | Tools wrap Power Platform connectors and agent flows, so the agent sends mail and writes Dataverse rows. |
| RECOVERSfixes its own errorsyes | GitHub Copilot harness runtime adjusts when a step fails — docs: 'Retries and finds alternative paths automatically'; computer-use's CUA loop re-perceives and adapts to unexpected screens. |
| UNSUPERVISEDnobody watchingyes | Published event triggers run the agent with no user prompt; with maker credentials on each action a full task completes with no per-action confirmation — confirmation is opt-in. |
Capability decisions combine documented behaviour and execution evidence. The proof label on each row shows which of the two decided it, and Leash is reported separately, only when it was measured from an execution record. Primary source: learn.microsoft.com. Tested against: Generative orchestration, standard harness (2026 docs). Our confidence in this dossier: 5/10.
Can you see it work?
the weakest of the five grades aboveOPERATOR INSPECTABLE — You can run it and read the steps yourself — the record exists for whoever runs it. Trace access: OPERATOR-OWNED — Whoever runs it holds the record; nothing is published.
session chat transcripts downloadable via Dataverse or Studio, 28-day window
Where we looked: Microsoft Learn analytics overview: 'Download conversational transcripts' via Dataverse/Power Apps or the Studio app; session details retained 28 days; trigger-run sessions. · see the evidence →
- 2026-08-113/5 WRAPPER WITH AMBITION → 5/5 AGENTPre-launch adjudication: every criterion the verdict depends on was re-checked by hand against the published rules.editorial correction — no case was filed
- 2026-08-112/5 WRAPPER WITH AMBITION → 3/5 WRAPPER WITH AMBITIONLOOPS corrected to yes: LLM planner decides step count per task at runtime; docs cite "dynamic plans" and a plan-execute-adjust loop. Plan complete is a terminal trigger, not evidence of a human-fixed sequence.editorial correction — no case was filed
Badges
earned, never for sale
These render from the live dataset, so a badge cannot outlive the verdict it claims. If the verdict changes the badge changes with it, and an unearned one returns 409 rather than an image. No sponsor can buy one, at any price.
Think this is wrong?
bring evidence, not opinionName the criterion and link a trace, a doc, or a recorded run that shows the behaviour. A verdict changed by evidence is the best thing that can happen to this site, and the change goes on the public record with your reason attached. Vendors are welcome, and a vendor's own filing is labelled as such rather than buried.