Independent technical audit
I inspect architecture, code, security, agent environments and actual practices. I separate observations from unknowns.
Deliverable
An evidence-based assessment and a prioritised remediation plan.
AI audit and roadmap
Find where work gets stuck, what is worth funding and which decisions your teams need to make first.
01 / The method
How the work runs
A roadmap grounded in how your team works
We start with a decision you need to make and a real request. I follow it with the people who handle it and cross-check their observations against the repositories, CI and available traces.
Select a level in the funnel to see an example.
-
Set the scope
Choose the decisions, systems and user journeys to examine. Agree on access and boundaries.
Example We follow a bug fix from report to release in one repository, using read-only access.
-
Map the work
Compare repositories, documentation and operational data with how the team actually works.
Example We compare the workflow in the wiki with the tickets, pull requests and approvals the team actually uses.
-
Verify findings
Try the checklist on a sample. Agents help gather evidence. I verify findings against sources or repeatable tests and flag what remains uncertain.
Example We rerun a check on a sample change to see whether it catches the defect it is meant to block.
-
Set priorities
Connect each risk to its impact and decide which actions to take first.
Example roadmap item
- Owner
- Tech lead
- First trial
- Test an approval workflow in 1 repository.
- Baseline
- Current time to an accepted change, including review and rework.
- Decision
- Continue if lead time drops and review still catches the expected defects.
-
Hand over
Give leadership the summary and roadmap. Give the team the evidence and steps to reproduce the findings.
Example Leadership gets the roadmap. The team gets the repository, revision and steps needed to rerun the check.
What you receive at handover
- For leadership
- An audit report and a roadmap prioritised by impact and effort, with the rationale for the first trial.
- For the team
- Notes for each repository with the sources, the branch and revision reviewed, and the steps to rerun the checks.
What your team will be able to do on its own
Trace a finding to the right revision, rerun the documented check and verify whether the fix resolves the problem.
If the audit Brain is in scope, the team can use an assistant to query the findings and retrieve the evidence and steps to rerun the checks.
02 / Anonymised audit example
AI instructions on an unmerged branch
During a follow-up audit, AI instructions appeared to have disappeared after a migration. Restoring them could have become a priority.
- Signal
- The file used in the previous assessment is missing from the default branch.
- Git evidence
- Git history shows the file lived on a branch that was never merged into the default branch.
Whether the instructions were used in practice remains to be checked.
- Decision
- Revisit the regression finding and set priorities based on the configuration in place.
What we can work on
Processes & bottlenecks
I follow work from request to delivery, with the people doing it, to identify queues, rework and handover gaps.
Deliverable
A map of the bottlenecks and changes to test first.
Vision & AI roadmap
I align business goals, use cases and constraints with leadership and teams.
Deliverable
A prioritised roadmap, decision criteria and clear ownership.
Two decisions for teams using coding agents
Assess the coding-agent budget
I compare subscriptions, API usage and infrastructure for your workloads. We agree how to count accepted work and which costs to include, such as retries and review time.
Deliverable
A cost model with its assumptions, investment options and a plan to track cost per accepted task.
Review agent governance
I examine who can authorise a tool or data access, which actions require human approval and who verifies the result. We review the rules against examples of how teams actually work.
Deliverable
A map of responsibilities and approval points, with gaps to address first.
Frequently asked questions
How can we use the Brain after the audit?
If the Brain is in scope, your team can look up the evidence behind a finding and the reasoning behind a recommendation when picking up the work.
How can we verify a finding after handover?
The team gets the evidence and steps to reproduce the finding. The handover separates what we verified from what still needs checking.