Quarterly dip-sampling¶
The Standard is explicit that high-volume AI use should be assured by dip-sampling rather than scrutinising every output — checking every routine email draft is neither proportionate nor realistic. Chronity automates the defensible-sample part; you do the reviewing.
You'll find it as a tab on the Compliance dashboard (Today → Compliance → Dip-sampling), and individual sample batches open from there.
How the sample is chosen¶
On the first day of each quarter, Chronity selects a random sample of the previous period's Track 2 work — routine professional work that used judgement but isn't a formal deliverable. The sample is:
- 10% of Track 2 observations,
- with a minimum of 10 and a maximum of 50, so a quiet quarter still gets a meaningful check and a busy one doesn't become unmanageable,
- chosen with cryptographically secure randomness, so the selection is audit-defensible — you can show the sample was genuinely random, not cherry-picked.
The batch then appears for review.
Reviewing a batch¶
Open the quarter's batch. It shows the selected observations and tracks completed vs outstanding. For each sample you open the observation, check that the classification and the work look right for routine Track 2, and record your review — approve it, or note a concern. Your notes are encrypted and kept as part of the audit trail.
This is the one piece of genuinely periodic work the role involves. It's bounded (at most 50 items a quarter, usually far fewer) and it's exactly the assurance activity the Standard asks for — so it's worth doing properly rather than rubber- stamping.
What to look for¶
You're not re-doing the work. You're checking:
- Did the classification make sense? Track 2 work that should really have been Track 1 is the important catch — it means a formal deliverable didn't get a reliability note. If you see a pattern of it, the taxonomy needs adjusting.
- Does the AI-assisted work look sound for routine professional use — no obvious unchecked errors?
- Anything that, with hindsight, should have raised an alert?
A dip-sample that turns up a misclassification is the system working — that finding feeds the taxonomy and the next quarter is better. A clean dip-sample, recorded, is exactly the evidence an auditor wants to see.
Keep the records
The value of dip-sampling to an audit is the record that it happened — the sample, the reviews, the dates, the reviewer. Chronity keeps all of that; your part is to actually do the reviews each quarter rather than letting the batch sit.
Next: The quarterly risk register.