Insurance · PMO

From 30 to 50 hours of review to 4 to 8 hours per cycle.

Ryan built an AI agent for a Canadian insurer's Program Management Office. It reads every project closure report, finds the failures that keep coming back, and drafts the "repeat offenders" report in the PMO's own format.

Shipped to a client · built in the client's Microsoft 365

Built byRyan Rezeika

DomainInsurance · PMO

PlatformMicrosoft 365 · Copilot Studio

CadenceQuarterly, was annual

30–50h → 4–8hReview time per cycle

4Reports a year instead of one

EveryClosure report read, not a sample

0New platforms: it runs in the client's existing Microsoft 365

The problem

A report that mattered, produced too late to act on.

The PMO produces a "repeat offenders" report: a lessons-learned analysis of the failures that recur across completed projects. It drives corrective action and accountability for problems that are systemic, not one-off.

A thorough cycle took a senior PM 30 to 50 hours of reading closure reports, cross-referencing issues, sorting them by theme and severity, and writing the result in the corporate format. So the reports were sampled, not read in full, and the analysis ran once a year. By the time a pattern was visible, the next projects had already repeated it.

Pain pointWhat it costWhat the build does

Reading every report by hand

30 to 50 hours of senior PM time per cycle.

The agent reads every closure report and extracts each improvement issue.

Sampling instead of reading

A repeat failure can hide in the reports nobody read.

Every report is read in full, every cycle.

Once a year

Patterns surface after the next projects have repeated them.

Quarterly reporting, so there is still time to act.

Format work

Hours spent putting findings into the corporate template.

The report is drafted in the PMO's own format, ready for review.

The approach

Use the platform the client already has. Keep the PM in charge.

There were no new platforms and no outside tools. The agent runs inside the client's existing Microsoft 365: Teams for the conversation, SharePoint for the closure report library, and Copilot Studio for the agent.

The AI reads everything. The PM decides what it means.

The work started with mapping the manual workflow and the structure of the closure reports, then confirming the repeat-offenders report as the highest-value first build. Prompts, the issue taxonomy and the agent design came next, followed by testing, deployment and team training.

The AI does the reading and the first draft in minutes. The PM still reviews the findings, which is where the 4 to 8 hours go, and owns what goes to leadership.

How it works

Three AI stages, then a person.

Each stage feeds the next, from raw closure reports to a report the PM signs off. Click a stage.

Interactive

Try the agent

A simulation of the agent, with sample projects.

Repeat Offenders Agent simulation · sample projects

I read project closure reports and find the issues that keep coming back. What would you like to do?

What it shows

What this work shows.

The win is not only hours. It is reading everything, four times a year, so a pattern gets caught while it can still be fixed.

Read in full, not sampled

A repeat failure cannot hide in the part of the portfolio nobody had time to read.

Frequency is the real gain

Moving from annual to quarterly turns a retrospective into an early warning.

Use what the client owns

Built inside Microsoft 365, so there was nothing new to buy, approve or secure.

The PM keeps the judgment

The agent drafts. A person decides what the findings mean and who owns the fix.

See it working

Try something like it.

Run the same workflow yourselfThe "Inside the PMO" tab of the AI Workflow Playground runs a lessons-learned review with agents you can watch.Open the playground →What a PMO should automate firstA short, practical answer on where PMO automation pays back.Read the answer →

Microsoft TeamsCopilot StudioSharePointPower AutomatePowerPoint output

Is your PMO reading reports nobody has time to read?

The same pattern fits status packs, stage-gate reviews and lessons learned. Start with a conversation about the report your PMO produces too late.

📅 Book a call