Speak-up Case Trends & Audit-Committee Report: Program Health Without Exposing Anyone

The problem this kills

Every quarter (or before every board meeting), the ethics lead has to tell the audit committee how the speak-up program is doing. So they open the case tracker, dump everything into a spreadsheet, and start counting by hand: how many reports this period, broken down by category, how many were substantiated, how long cases took to close, and whether anything is trending up. Then they hand-write the narrative, double-check that nothing in the deck could accidentally point to a specific reporter or subject, and rebuild the whole thing from scratch next quarter.

It's slow, it's error-prone, and the scariest part is the anonymization. One small number in the wrong cell — "1 substantiated harassment case in the Denver office" — can effectively name a person, and that's exactly the kind of mistake a manual deck makes under deadline pressure. You don't need to be a developer to fix this, and you should not be exposing case identities to do your job.

What you'll build

A simple internal web tool. You import case metadata only — category, severity, status, the key dates, and the outcome — with no names, no narratives, no identifying details. The tool computes the numbers the committee actually wants: report volumes by category, substantiation rates, time-to-close, and emerging hotspots, and it compares this period to prior periods so you can see what's moving. Crucially, it enforces anonymization: it suppresses small cells (any group too small to be safe) so the metrics can never single someone out. AI then drafts a plain-English narrative of what the numbers say. The ethics lead reviews the metrics and the draft, edits the wording, and clicks Approve. Only then does the tool assemble the committee pack and email it via Resend, and export the aggregated metrics CSV.

What's inside the Implementation Plan

The downloadable plan is a step-by-step file you paste into an AI coding agent. It opens by interviewing you about your program — which case tracker you export from, exactly how your categories and severities are named, what dates you record, how you define "substantiated" and "closed," your reporting cadence, and your suppression threshold — and then it tailors the data model, the metrics, and every later step to your answers. This is not a generic template; the agent reflects a short spec back to you and waits for your thumbs-up before it builds anything. From there it walks the agent through the de-identified import, the metric and trend calculations, the small-cell suppression rule, the AI narrative draft, the ethics-lead review-and-approve screen, the committee-pack email, and the aggregated CSV export — each step with a ready-to-copy prompt. There's a fallback so you can build and run the whole thing today from a CSV or Google Sheet, even with no integration to your case tracker.

The governance it includes (this is the point)

This is sensitive program data, so the tool ships with the controls a compliance function needs: login so only your ethics team can use it, row-level security so you only ever see your own organization's data, a complete audit trail of who computed, edited, approved, and sent which report and when, a hard human-approval gate so no committee pack is emailed and no narrative ships until the ethics lead signs off, and duplicate guards keyed on case ID so the same case can't be counted twice. On top of the standard controls, this tool adds anonymization by design: only aggregated metrics ever leave the tool, small cells are automatically suppressed so a count can't identify a reporter or subject, and the narrative stays human-owned — the AI proposes the words, a person decides them.

Who it's for

Ethics and compliance officers, chief compliance officers, and program managers who report speak-up program health to the audit committee or board and want a faster, safer way to produce the quarterly pack. If you can describe how you categorize and close cases, you can build this.

You've got this — start with the plan, paste the first prompt, answer the interview, and you'll see your anonymized trends pack take shape the same afternoon.