Recompute a decision chain
Six published records from a demonstration tenant. Recomputing the chain checks internal consistency; it does not establish who produced it.
Inspect the records and methodIntelXview AI Control Plane
Your AI agent is ready to work. Can you prove it’s under control?
Before an AI agent handles a customer request, accesses sensitive data or takes an action, you need to know what it’s allowed to do, when someone must approve it and what evidence it leaves behind.
Put those controls to the test on one of your AI workflows.
Discuss your workflow. The PoC is paid; scope and price are agreed before work starts. To change or cancel, reply to your calendar invitation. Open AI Control Plane opens the product sign-in page.
A proof run, with its limits visible
ACP public evidence screen · 4 October 2026
What a PoC tests
Four controls, tested on one agreed workflow.
Each test is agreed with you before any paid engineering starts, and limited to scenarios ACP supports today.
Authority
Restrict the agent’s actions to the permission you agreed.
Human approval
Hold sensitive actions until someone approves them.
Evidence
Inspect the records of allowed, denied and approved outcomes.
Failure
Test what happens on expiry, revocation and when the control is unavailable.
How it works
Start with one workflow.
Agree scope, success criteria, technical dependencies and price before paid engineering. Receive the agreed test results and evidence, with no obligation to buy the ongoing service.
The first stage may be a controlled demonstration with synthetic data. Installing ACP in your environment, cloud permissions and integration are scoped and priced as separate stages.
- 01
Free scoping call
Tell us about one AI workflow and who owns it. No long questionnaire up front.
- 02
Written scope and quote
We set out feasibility, proposed tests, your dependencies, acceptance criteria and the price.
- 03
Paid PoC
Engineering starts only after you agree the scope and price.
- 04
Results and next step
You receive the agreed results, evidence and limits, then decide whether to discuss a longer-term service.
Before an AI agent acts, what is it authorised to do?
Follow one governed action.
Start with the agent’s identity and authority. Follow the policy decision, then inspect the record it leaves.
- 01
Identity
Which agent or person is making the request?
- 02
Authority
What are they allowed to do?
- 03
Mandate
What is the permitted purpose and scope of this job?
- 04
Policy decision
Does this request meet the configured rules?
- 05
Enforcement
Apply the decision on the route through ACP.
- 06
Evidence
Keep the decision record available for review.
This is an explanation of the control path. Enforcement depends on the configured integration: activity that bypasses ACP is outside this path. Read the product boundary.
Evidence you can inspect
What would you show for one governed action?
Inspect the example record and the control it demonstrates. Each published artefact states its source, verification method and limits.
Check a signed evidence pack
Download the published pack, verification key and script. Read the signing and provenance limits alongside the result.
Verify the signed packCan you prove your AI agents are under control?
Book a free scoping call about one AI workflow. Bring a short description of the workflow and who owns it. We come back with feasibility, proposed tests, a written quote and a deliverable plan.
Discuss your workflow. The PoC is paid; scope and price are agreed before work starts. To change or cancel, reply to your calendar invitation.
Assess workplace AI exposure
If your immediate question is where sensitive data can enter workplace AI tools, the existing exposure check and assessment remain available.
Explore the assessmentAbout IntelXview
IntelXview builds AI Control Plane to connect policy decisions with inspectable evidence.
Read about the companyPractical notes on AI governance
All insightsAI Governance
When Human Approval Becomes the Bottleneck for AI Agents
Requiring a person to approve every agent action feels safe until the queue grows and people start working round it. The fix is deciding which actions need approval, and recording each decision.
Read insightAI Governance
When an AI Agent Takes an Exceptional Action, Who Answers for It?
An agent does something unusual. Someone has to explain whether it was authorised, who allowed it and what happened next. Logs that show it ran are not the same as evidence that it was allowed.
Read insightAI Governance
When a Role Changes, Does the AI Agent's Permission Change With It?
Access granted for one task or role often outlives it. When an AI agent acts on that access, the old permission keeps working. Here is the control question to ask, and what a record should show.
Read insight