Prove your AI controls work, with evidence from how your AI actually runs.
Policies say every agent has an owner, a budget and safe data handling. Customers and auditors now ask for proof. Bodh tests each control against operational data and keeps the evidence.
We test three controls against one month of operational data and send back a result like this one, in about a week.
Control check, October 2026
| Control | Evidence | Result |
|---|---|---|
| Every production agent has an owner | Agent register, 14 agents | 12 of 14 pass |
| Agents run within approved budgets | Spend logs, October | 1 exception |
| Sensitive fields are redacted in logs | Sample of 500 runs | Pass |
Mapped to ISO/IEC 42001, NIST AI RMF, the EU AI Act and India's DPDP. This is evidence mapping, not certification.
Operational data becomes evidence that controls ran
You share exports from the tools you already use. Nothing to install.
What you get
Controls tested from data
Each control is checked against real operational data, not a questionnaire.
Exceptions with owners
Every failed check becomes an exception, assigned to a person, with a record of how it was fixed.
Framework mapping
Evidence mapped to the frameworks your customers and auditors ask about, kept separate from legal obligations.
Audit-ready history
A dated record of every test, result and fix.
Could you prove your AI controls work?
Five questions, about two minutes. Your answers stay in your browser; nothing is sent.
Do you keep an inventory of the AI systems in production?
Does every system have a named owner?
Do you have evidence, not just documents, that controls ran last month?
Are exceptions recorded along with how they were fixed?
Could you answer a customer's AI questionnaire with evidence within a week?
How it works
- Pick three controlsWe start with the ones your customers or auditors ask about most.
- Send the dataAgent lists, spend logs and a sample of telemetry. No prompts or customer content.
- Get your control checkResults, exceptions and the evidence trail, in about a week.
What we ask for, and what we never touch
We ask for
- Usage and billing exports from your AI providers or gateway
- A list of your agents or AI features, if you keep one
- Thirty minutes with the person who owns the numbers
We never ask for
- Prompts, outputs or customer content
- Access to your systems or credentials
- Anything installed in your environment
A mutual NDA is available before you share anything. We delete your exports within 30 days of the diagnostic unless you ask us to keep them.
Who you'll talk to
Laukesh Singh
Partner at Bodh. Laukesh takes the first call and runs every diagnostic with your team.
Laukesh on LinkedInQuestions
What does it cost?
The control check is free. If you want controls tested every month, we agree the terms with you as a design partner.
Is this legal advice or certification?
No. It shows whether your own controls are working and maps the evidence to frameworks. Your legal and audit advisers decide what it means.
Which frameworks?
ISO/IEC 42001, NIST AI RMF, the EU AI Act and India's DPDP, with others on request.
Who is this for?
Companies whose customers or auditors have started asking how they govern AI.
Start with the free diagnostic. Decide after you've seen your own numbers.
No commitment, no installation, about a week.