Planned capability · AI Agent Governance
Let AI agents act.Within limits you set.
Dutira’s longer-term direction as firms introduce more autonomous agents: register every agent, give it an identity and permissions, authorise its tools and actions, require human approval above thresholds, and keep its actions on record. Planned for a later phase.
In development — not part of the initial commercial release. An early version is included in the demonstration workspace so firms can review the design; it is not offered as a production capability.
One inventory
Which agents are live, who owns them, who is accountable, and what systems and data they can reach.
Hard limits
Permissions per tool, financial limits per action and per day, and approval thresholds — checked before every action.
An emergency stop
Suspend an agent in one step. Every later action is blocked and the decision is recorded.
Concept preview
How it is designed to work.
A preview of the planned capability, using a fictional firm’s data. Design, not a production feature.
Payments operations agent
Pays routine supplier invoices to existing, verified payees.
Owner Leila Haddad, Chief Risk Officer · Accountable Richard Hale (SMF1) · Acts within set limits
Approved permissions
- invoices.readread
- payees.readread
- payments.instructexecute
Limits and thresholds
£5,000 per action · £20,000 a day
A person approves from £2,000
0 action(s) tried in this demo
Ask the agent to act — the control decides
Planned capabilities
What AI Agent Governance will include.
In development — not part of the initial commercial release. An early version is included in the demonstration workspace so firms can review the design; it is not offered as a production capability.
- AI agent registration
- Agent identities and permissions
- Tool and action authorisation
- Human approval requirements
- Agent action logs
- Delegated authority limits
- Agent risk monitoring
FCA AI governance expectations
How it relates to the FCA’s October 2026 publication.
The FCA’s AI in financial services: good practice and areas for improvement (19 October 2026) shares examples from industry. In its own words: “They do not create new requirements.” There is no standalone FCA AI rulebook; existing rules apply.
Below, each example it describes, and how this planned capability would support a firm’s own approach.
Registering agents in a central inventory showing which are live, who owns them, what systems and data they can access, and what controls apply
Dutira: Planned: an agent register recording exactly these, linked to each agent’s use case and supplier.
Designing a governance structure before deploying more autonomous agentic systems, with escalation routes and compliance controls
Dutira: Planned: agents start in pilot; going live needs an accountable senior manager and an approved use case.
Clear thresholds for when human intervention and escalation are needed, and how outputs can be challenged or overridden
Dutira: Planned: approval thresholds and approval-required tools that hold actions for a person, whose decision is recorded.
Senior managers accountable for AI risks from the design stage onwards
Dutira: Planned: each agent names its accountable senior manager; agents without one cannot move to production.
The rules behind it
Existing rules continue to apply when firms use AI. Which apply depends on each firm’s permissions and business.
- SYSC 4.1 General organisational requirements Rule
- SYSC 24 Allocation of responsibilities (SM&CR) Rule
- SYSC 15A Operational resilience Rule · scope varies
- SYSC 8 Outsourcing Rule · scope varies
- Rule Binding on the firms it applies to
- Guidance How the FCA expects rules to be met
- Good practice Illustrative examples; no new requirements
Your responsibility. Your firm would set each agent’s permissions, limits and thresholds and stays responsible for what its agents do — including those built on third-party models. Dutira is not approved or endorsed by the FCA and does not guarantee regulatory compliance.
One platform, launched in phases
- Phase 1 · available nowGoverned AI WorkspaceEnable your people to use approved AI tools within company-defined safeguards.Explore
- Phase 1 · available nowAI Policy & AttestationTurn written AI requirements into practical safeguards, workflows and staff acknowledgements.Explore
- Phase 1 · available nowAI Monitoring & ControlsUnderstand how AI is used and identify potential policy risks.Explore
- Phase 1 · available nowSenior Management OversightGive decision-makers the visibility and evidence needed for responsible AI adoption.Explore
- Phase 1 · available nowFCA AI Governance CheckerIdentify potential gaps in your AI governance framework against relevant FCA regulatory expectations — free.Explore
- Phase 3 · plannedAI Lifecycle GovernancePlanned: inventory, registration, risk classification, assessment, approval and review for every AI use case.Explore
Interested in this capability?
Tell us about your plans for AI beyond employee chat; we will share the roadmap and timing.