Let AI agents move money, with controls your risk team can sign off.
Banks, lenders and payment firms already have agents that can issue refunds, settle disputes and send payments. What they lack is a way to prove each agent stayed within its limits, that a named person approved the big decisions, and that it could be stopped in an instant. Vorim gives you all of that, with a signed record of every action.
The model isn't what's holding agents back.
Most agent projects in regulated finance stall at sign-off, not in engineering. Risk, compliance and security teams ask four questions before anything touches a customer's money, and most teams can't answer them yet.
Which agent did this, and who owns it?
"The AI did it" isn't an answer a regulator accepts. Every agent needs its own identity and a named owner.
What stops it going beyond its job?
A refunds agent has no reason to change a credit limit or add a payee. Its permissions should match the task.
Who approved the big decisions?
Large or unusual actions need a named person to approve them, and a record that shows who did and when.
Can we stop it, and prove what happened?
When something goes wrong you need to stop the agent at once and hand an auditor the full story in one export.
A refunds agent at a UK digital bank
A worked example of how a disputes and refunds agent would run under Vorim. It's an illustration of the controls, not a named customer.
- 01
Give the agent an identity
The refunds agent is registered with its own cryptographic key, an owner in the disputes team, a class and a risk tier. Every action it takes is tied to that identity.
- 02
Grant only what the job needs
It can read transactions and issue refunds. It can't change credit limits, add payees or move money between accounts. Permissions can be revoked in a moment.
- 03
Say what it's about to do
Before it acts, the agent declares its intent, for example "refund up to £42 for complaint #4821". Anything outside that is flagged as off intent and your rules can block it.
- 04
Check every action first
Each refund is checked against your rules before it goes out. The verdict comes back as allow, deny or escalate, and the agent waits for it.
- 05
Put a named person on the big ones
Escalations go to the person your rules name. Each approval or denial is signed with who decided and when. Large refunds can need two approvers, and a request nobody answers times out as a no.
- 06
Stop it instantly
One click suspends the agent and every agent it handed work to. Vorim records and signs how quickly the stop took effect. Suspending is reversible; revoking is permanent.
- 07
Hand over the evidence
Every action, decision and approval is signed and can be checked without Vorim. Export a signed bundle or CSV, Excel and PDF reports, or stream events to Splunk or Datadog in OCSF format.
Evidence for the frameworks your supervisors use
Vorim doesn't make a firm compliant on its own, and it isn't legal advice. It gives you the controls and the records that these frameworks ask you to show.
Where the value shows up
Agents only save money once they're allowed to run. Vorim is how they get past the pilot, and the pilot is where you measure it with your own numbers.
Lower cost per case
Routine, low-value cases are handled by the agent, and people spend their time on the ones above your limits.
Faster to production
Risk and compliance can inspect the controls and the records, so sign-off doesn't stall for months.
Smaller losses
Money limits, velocity limits and circuit breakers cap what a faulty or compromised agent can do.
Cheaper audits and complaints
"What did the agent do on 3 March, and who approved it?" becomes one query instead of a week in the logs.
One agent, one workflow, 30 days
We start in shadow mode, where your rules record what they would have decided without blocking anything. Once you trust what you see, they go live.
Pick one workflow, such as refunds or disputes. We register the agent, set its permissions and write the rules with your risk team.
The agent runs with the rules in shadow mode. You review what would have been allowed, escalated or blocked, and adjust.
The rules go live and named approvers take the escalations. At the end you get the signed record and a short report.
What we measure together
- Share of cases resolved with no person involved
- Median time to resolve a case
- Escalation rate, and how long approvals take
- Value of actions blocked or escalated by the rules
- Time to answer an audit or complaint request
What Vorim is, and what it isn't
Vorim controls what your agents are allowed to do and proves what they did. It doesn't make a model's judgement correct, and it doesn't replace your own model risk work.
We don't have a named financial services customer to show you yet. That's why we start with a pilot measured on your numbers, not ours.
Your agents call Vorim through our TypeScript and Python SDKs or the REST API, so it works with the payment and core banking systems you already run.
Start with one workflow.
A 30-minute call to walk through your agents, your rules and what a pilot would look like.