What an AI inventory should contain, and who owns it
RBI's Governor has asked banks to keep a complete inventory of AI systems, including those inside vendor software. A useful inventory records what each model decides, what data it reads, who owns it and how to stop it.
Intelligent systems2 min read
An AI inventory is a register of every model your organization uses to make or shape a decision, whether you built it or a vendor shipped it inside a product. For each one it records what it decides, what data it reads, who owns the outcome, how it is monitored and how it can be stopped. Without that list, it is hard to show any other AI control works.
In India the expectation is now explicit for banks. At FIBAC on August 11, RBI Governor Sanjay Malhotra named a complete inventory of AI systems, including AI embedded in vendor products, as an immediate priority. Across Indian enterprises, the gap is wide: ServiceNow's 2026 index found that only 22% of Indian enterprises have AI testing, auditing and risk-assessment processes in place.
22%
of Indian enterprises have AI testing, auditing and risk-assessment processes in place (ServiceNow, September 2026)
What each entry should record
- The decision it makes or influences, in business terms, and the system where that decision is taken.
- Whether it acts on its own, recommends to a person, or drafts for approval.
- The data it reads, where that data comes from, and the purpose under which it was collected.
- The model's origin: built in-house, fine-tuned from a public model, or embedded in a vendor product.
- The business owner who answers for its results, and the technical owner who keeps it running.
- How it is monitored, when it was last validated, and the measure that triggers a review.
- How to switch it off, and what the process falls back to when it is off.
Finding the models you did not build
The models your data team built are already known. The ones that surprise you sit in vendor software: scoring in a lending package, routing in a contact-center platform, anomaly detection in a security tool, a language model added to a ticketing system in a routine upgrade. Ask every vendor, in writing, which features use a trained model and what data they read. Add the answer to the contract renewal checklist so the inventory stays current.
Who owns it
The inventory itself needs one owner, usually in risk or the office of the chief data officer, with authority to require entries. Each model in it needs a business owner, not only an IT owner. The business owner is the person who would be asked why a customer was declined or a claim was flagged, and who can decide to stop the model.
Keep it alive
An inventory built once for an audit goes out of date quickly. Tie entries to the change process: a new model, a retrained model or a vendor release that adds AI cannot go live without an updated entry. Review the list against what is running at least once a year, and after any incident.
Questions
- What is an AI inventory?
- A register of every AI model an organization uses to make or influence decisions, including models inside vendor products, with each model's purpose, data, owner, monitoring and shutdown procedure.
- Does RBI require banks to keep an AI inventory?
- RBI Governor Sanjay Malhotra named a complete inventory of AI systems, including AI embedded in vendor products, as an immediate priority for banks at FIBAC on August 11, 2026.
- Who should own the AI inventory?
- One function, often risk or the chief data officer's office, owns the register. Each model in it needs a named business owner who answers for its decisions and can stop it.
Sources
- RBI Governor Sanjay Malhotra Sets Out AI Playbook for Indian Banks, The Kanal, August 13, 2026
- India's enterprise AI investment surges 119% but, only 22% of Indian enterprises have the governance to match their AI ambition, ServiceNow, September 8, 2026
More on intelligent systems
All writingNovember 2026
CERT-In's AI security audit guidelines: what an auditor will check
CERT-In's AI Security Auditing Guidelines, dated November 4, 2026, sort AI systems into four risk tiers and audit them across nine domains. For AI agents, auditors will test whether an agent can exceed its permissions or bypass approval. Here is what to prepare.
October 2026
How do you test an LLM before it goes live in a regulated business?
Test a language model the way an auditor will question it: on your own cases, against written pass criteria, with attacks on its permissions and data, and with a record of every result. RBI's Governor has asked banks to red-team and stress-test AI before deployment and periodically after; five test sets are where that starts.
October 2026
Multilingual voice AI for Indian call centers: a build checklist
A voice AI line in Indian languages succeeds when it resolves the calls that make up most of the volume, hands the rest to a person with context, and logs every call in the core systems. A build checklist, with results from a line we built that answers in 11 languages.
Tell us which system the AI has to run inside
If that system is not ready for AI yet, tell us that too. An engineer will reply and say plainly whether we are the right firm for it.