Why a system nobody revisits gets worse
Microsoft shipped two changes to this in September 2026 alone, neither announced far in advance. Meanwhile your teams add slides and find new uses for Copilot. A system nobody revisits is materially worse a year later, and the drift is invisible until someone sends the deck. Oversight is what keeps the setup working as the platform changes and your teams drift.
What happens every quarter
Re-run the tests
We run the same fixed set of prompts against your template that we used at the build, so each quarter’s results compare with the last.
Review what your people generated
We look at real decks your teams produced with Copilot, not only our test prompts.
Refresh and update
We refresh the example slides and update the rules file where the output has drifted.
Report in writing
You get a written report of what we found and what we changed.
Your part in an Oversight quarter
About 1½–2 hours a quarter, mostly your brand manager’s.
- Week 1: share 10–20 recent Copilot decks and tell us about any brand changes (about 25 minutes).
- Weeks 2–3: approve the refreshed example slides (about 20 minutes; Standard only).
- Week 3: upload the updated template and rules to your Brand Kit (about 15 minutes). Only you can: it is your Microsoft 365 tenant.
- Week 4: the review call (45–60 minutes).
What the written report shows
- A score for every test output, on a 0–4 scale of the human work it still needs: 0 usable as generated, 1 light cleanup, 2 skilled correction, 3 senior judgment needed, 4 rebuild it.
- A cause for every failure: the template, the examples, the prompt, the brand rules, or Copilot itself. That is what tells us what to fix.
- Structural checks: which layouts Copilot chose, whether it filled placeholders or added loose text boxes, and any colors or fonts outside your theme.
- What we changed: the examples we refreshed and the rules we updated.
- The comparison: this quarter’s results against earlier quarters, on the same prompts.
- Anything we could not review: if a step is skipped, the report says so plainly, for example “Not reviewed: no decks received this quarter.”

What we commit to
No one can guarantee what a generative tool will produce, and we do not. What we commit to is conformance with Microsoft’s published guidance, testing with real prompts against your own template, and correcting what drifts. Oversight is how we correct what drifts, every quarter.
Read how the setup is built in the first place, what we have tested so far, where oversight sits in the timeline, and how oversight is priced. Back to the overview.
Start with the audit
Oversight follows a build. The first step is a scored review of the template you have today.