Our engagements
Four engagements, one for each stage of the system.
Before funding, before committing, before depending, in production: each engagement answers one precise question and ends with a file your leadership can read and your teams can maintain.
Choose the engagement by the decision in front of you
You are funding, signing, already depending or operating: the right engagement is the one that matches that decision.
At a glanceThe question examined and the evidence delivered
Before work starts, the mandate sets the scope, access, examination criteria, timetable and closure conditions.
| Engagement | Question decided | Closure point | Main deliverable |
|---|---|---|---|
| AI portfolio arbitration | Which use cases deserve time, data and budget? | Documented decision for every candidate | Scored portfolio, go / no-go decision per candidate |
| AI systems audit | Is this system built and operated correctly? | Findings decided by a named owner | Findings report ranked by severity, treatment plan |
| Reversibility audit | What do we recover if this supplier stops tomorrow? | Exit executed and result accepted | Dependency register, exit-test record |
| AI production assurance | Does the system still deliver the approved service? | Decision to keep, restrict or stop | Scores against the production baseline, keep / restrict / stop decision |
AI portfolio arbitration
- Question decided
- Which use cases deserve time, data and budget?
- Closure point
- Documented decision for every candidate
- Main deliverable
- Scored portfolio, go / no-go decision per candidate
AI systems audit
- Question decided
- Is this system built and operated correctly?
- Closure point
- Findings decided by a named owner
- Main deliverable
- Findings report ranked by severity, treatment plan
Reversibility audit
- Question decided
- What do we recover if this supplier stops tomorrow?
- Closure point
- Exit executed and result accepted
- Main deliverable
- Dependency register, exit-test record
AI production assurance
- Question decided
- Does the system still deliver the approved service?
- Closure point
- Decision to keep, restrict or stop
- Main deliverable
- Scores against the production baseline, keep / restrict / stop decision
How we rank what we findA finding is only useful if it says who acts, by when, and with what evidence
Every finding carries a criterion, the observed condition, its cause, its effect and a five-level severity: informational, low, medium, high, critical. High and critical findings block release until they are corrected or explicitly accepted by a named owner. A disagreement or a missing piece of evidence never disappears into an average.
Sources and references
- ISO 19011:2018 (new tab)
Guidelines for auditing management systems: principles, programme, audit evidence.
- ISO/IEC 42001:2023 (new tab)
Requirements for an artificial intelligence management system.
- NIST AI RMF 1.0 (new tab)
Govern, map, measure and manage the risks of an AI system.
First conversation
Torn between two engagements?
It happens often: an audit frequently reveals a dependency to measure, and a production review starts by checking the architecture. Describe the decision in front of you; we will tell you which one starts, and whether one is enough.