Turning incident logs into a forward-looking signal
There’s a quiet assumption baked into most incident reporting tools: you file an incident, you resolve it, and the
record exists primarily for the auditor. Look-back, not look-forward.
That’s not wrong, but it’s incomplete. Your incident history is one of the best forward-looking signals you have. The
patterns are there, if you look.
This post is about how to read them, and what to do about what you find.
What “common problems” looks like in practice
Imagine you’re the operations lead at a 25-person accounting firm. Over the last quarter, you logged 18 incidents.
They’re scattered across categories: vendor non-compliance, audit findings, near-misses, equipment failures. You closed most of them. Some are still open.
If you stop there, the data sits in a folder waiting for someone to ask. If you look at the distribution, three patterns
emerge:
- Six of the 18 incidents were categorized as “vendor non-compliance,” all in the same department.
- Four were classified as “Critical” priority and took an average of 14 days to resolve (compared to the median
- of 4 days for other priorities).
- Three were filed by the same person, all near-misses in the same operational area, none of which led to a
- closed incident with an action taken.
Each pattern is actionable. Six vendor-compliance issues in one department suggests a process problem (or a vendor
selection problem) that needs attention beyond closing individual rows. Four Criticals taking 14 days suggests your
escalation path isn’t pulling in the right people fast enough. Three near-misses with no follow-through suggests
you’re recording but not learning.
These are the conversations that prevent the next incident. They don’t happen unless you actually look at the data.
How GlobalResolve makes this less work
We built a feature called Common Problems & Prevention into the Reports page. It does two things.
First, it does the deterministic part for you. It buckets your filtered incidents by type, counts them, and surfaces the
top three: how many, what percentage, total monetary impact, average resolution time, and the department most
often affected. No AI required. Just math on your data.
Second, on demand, it asks Gemini (with your data context) to suggest common causes and 2-3 concrete prevention
recommendations for each of those top three types. The AI doesn’t decide anything. It synthesizes. The output is
meant to give you a starting hypothesis to test, not a prescription.
The result is something you can take into a team meeting. “These three incident types are the ones we keep seeing.
Here’s what the data says about each. Here are some ideas for prevention, generated from the patterns. Which ones
do we actually want to act on?”
What this looks like in a real meeting
We worked with one of FACFO’s portfolio companies (a 40-person fintech) to test this approach in their monthly
operations review. Before: the meeting was a 90-minute slog through individual incident rows, with the team trying
to figure out what mattered.
After: the team spent 15 minutes on the Common Problems summary (three top types, stats, AI suggestions), 20
minutes deciding which prevention ideas to pursue, and the remaining time on the two or three incidents that
needed individual attention. The meeting got shorter. The actions got more concrete. Vendor non-compliance
dropped by 60% over the next quarter because they took the AI’s first suggestion (consolidate two flaky vendors into
one reliable one) and acted on it.
The honest limits
The AI suggestions aren’t magic. They’re synthesis. If your incident history is small (say, fewer than 20 incidents), the
patterns aren’t statistically meaningful and the AI will reflect that uncertainty. If your data is noisy (categories applied
inconsistently, descriptions sparse), the suggestions will be noisy too.
The deterministic stats are useful regardless of volume. The AI guidance is useful when your data is clean and dense
enough to surface real patterns. If you’re not there yet, the path is to clean up your categories and write better
incident descriptions, and the Common Problems feature will become more useful as your data matures.
A starting checklist
If you want to try this on your own data this week:
- Open the Reports page in GlobalResolve.
- Filter to the last 90 days. Leave other filters off.
- Hover the info icon next to the “Common Problems & Prevention” card title to see the top three types in
your data. - Pick one of the three. Ask: is this a pattern I can act on, or is the categorization too loose?
- If yes, click “Generate guidance” and read the suggestions. Take what’s useful, ignore what isn’t.
- Bring the strongest suggestion to your next ops meeting as a hypothesis to test.
Most of the value is in step 4. The AI is a way to surface possibilities faster. The judgment is yours.
