Compass - the automated baseline

Know where you stand before you improve anything.

Where you build software, step one of the method is automated. Compass reads your software delivery - repositories, workflows, tooling - and turns it into an evidence-based baseline. It shows how work ships, how well quality is guarded, where AI is actually used, and where improvement pays first.

"We build software; I want AI in the delivery itself, without losing control." The Compass report shows where AI already runs in your delivery and what it changes - and the two ends keep every change under your sign-off.

What you receive

Four things you receive - not dashboards to learn.

Per-repository report

Delivery health, engineering health, real AI adoption and ranked opportunities - one report per repository, in plain words.

Portfolio view

The same picture across every repository you point it at - where attention goes first.

Ranked opportunities

Where improvement pays back first, every recommendation anchored in the measurements behind it.

Re-run comparison

The same baseline, run again after a change - the delta, not a feeling.

What the report looks like

Dummy data - illustrative, not a result
Compass report - Example Parts Ltd (a made-up company) - repository "orders-api" dummy data
Overview AI adoption Delivery Engineering health Repo hygiene Automation Opportunities
ThroughputSteady
PredictabilityNeeds attention
Quality gatesStrong
AI adoptionSet up, unused
Repo hygieneMixed
Review wait (median)2.4 daysdummy reading
Changes shipped weekly14dummy reading
Checks green on merge8 of 10 runsdummy reading
AI adoption - set up vs actually used (dummy)
Code assistant - set up and used daily; its traces appear in the recent work. Review agent - set up, unused for weeks; paid for, changing nothing. Chat tool - used by the team, with no shared rule on what may leave the building.
Engineering health and hygiene (dummy)
Tests cover the main flows - thinner at the edges. The release branch accepts changes without review - open gap. Two core libraries are a year behind on updates.
Automation that actually runs (dummy)
Six checks run on every change - build, tests, style, safety. Two more are set up but have never triggered.
Ranked opportunities and risks (dummy)
1. Reduce review wait - the largest measured friction in this dummy picture; every change queues behind it. 2. Retire or adopt the unused review agent - visible cost, no visible change; adopt it with a shared rule, or stop paying. 3. Protect the release branch - the one gap that can undo everything the other checks guard. Risk noted: chat tool in daily use with no shared rule - covered by the two ends when augmentation lands here.

Every reading above is invented for illustration. Your report carries your numbers, each with its source - the seven views above, one report per repository, plus the portfolio view across all of them, re-runnable after every change.

Why you can trust the numbers

The mechanical checks are deterministic and re-runnable - same input, same answer. The interpretation built on them is a consultation, not a score.
The re-runnable baseline is the answer to AI's different-answer-every-run problem: the measurement does not depend on the day.
Read-only by design: Compass never writes to the systems it reads.
Repositories are never ranked against each other; each is measured against its own evidence and its own history.

Not building software? The same step one exists for any workflow - an automatic answer where automation is possible, the workflow baseline everywhere else. See the method.

Want your baseline?

Tell us what you are trying to improve. We will tell you what the baseline covers, and whether you need more than that.

Book a free working session