Tier 2 · fixed-price engagement
Agent governance review
What your agent is allowed to do without a human, how you stop it, and whether the audit trail would survive being asked for.
Who this is for
An organization running agents that take actions rather than only producing text, where the question has become who is accountable for what the agent did.
Fixed price
$7,900
- Duration
- 2 weeks
- Deposit
- $1,400
- Other currencies
- €7,200 · £6,090 · S$10,600
- Catalog
- T2-08
Thirty minutes with the engineer who would run the work, before any paperwork.
The work, stage by stage
In order, with what happens at each stage and what it produces. The timings are the ones the catalog prices, not an estimate written for this page.
Enumerate the agent's authority
Days 1 to 3
Every tool, API and credential the agent can reach, and what each one can change. Written as a capability list rather than an architecture diagram, because the governance question is what it can do, not how it is wired.
Find the human decision points
Days 4 to 5
Where a human approves, and whether that approval is a real control or a dialog everyone clicks through. An approval nobody can refuse under deadline is not a control.
Exercise the stop control
Week 2, first half
Whether a kill switch exists, who can use it, how long it takes to take effect, and what happens to work already in flight. Exercised rather than read about, because an untested stop is an assumption.
Read the audit trail as an auditor would
Week 2, first half
Whether you can reconstruct, for one specific action on one specific day, what the agent did, what it was given, which model version produced it, and who was accountable. Tested by attempting the reconstruction rather than by reading the schema.
Check model change control
Week 2, second half
What happens when the provider updates the model underneath you: whether anyone is told, whether anything is re-tested, and whether the previous version can be pinned.
Write the findings with owners
On delivery
Each finding with the control that closes it and the person who would own that control.
Everything that arrives, listed
The contents of the document, and the artifacts that come with it. If something here would not be useful to you, say so before we start and the scope changes.
- The capability list: every tool, API and credential the agent can reach, and what each one can change
- A map of human decision points, each marked a real control or a formality, with the reasoning
- Stop-control findings from exercising it, including time to take effect and the fate of in-flight work
- An audit-trail assessment written against a specific reconstruction attempt
- Model change-control findings: notification, re-testing, and version pinning
- Each finding with its closing control and a named owner
What this covers, and who takes the rest
Stated up front, because scope is what derails an engagement in week two rather than at the proposal.
Inside this engagement
- The agent's authority, and the credentials that carry it
- Human-in-the-loop boundaries as practiced rather than as documented
- Stop controls, audit trail, and model change control
Handled separately
Named so you can plan and budget for it, whether we do it or somebody else does.
- Prompt injection and exfiltration testing, which the AI security assessment covers in depth
- Building the controls the review calls for, scoped once the findings name them
Start with a call, not a proposal.
Thirty minutes, straight to the engineer who would run the work. We will tell you if a smaller engagement answers your question, or if nothing here does.
Typical reply within one business day · Engagements start at $2,500
Fixed price · 2 weeks