AI agents for banks.
Designed to be audited.

The accountability stays with you. SS1/23 puts the obligation on the firm, not the supplier. What we provide is the infrastructure to make compliance practical: a versioned model inventory, independent validation documentation, override tracking, generation testing for any generative components, and ongoing performance monitoring that your model risk function can inspect at any time. We design every agent so the documentation your second line needs is produced as a by-product of the work, not as an afterthought.

You own your data: every input, every output, and the audit log, all exportable on request. The agent is ours — the engineering, the prompts, the infrastructure. That's why there's no licence and no seats: we run the machine, you pay per output. At the end of the term you get a full export, all outputs, and a documented handover. No lock-in. No exit fees.

By default, no. UK GDPR Article 22 puts a high bar on solely automated decisions with legal or significant effect, and Consumer Duty adds an outcomes obligation on top. Our agents are designed to produce a recommended decision and the evidence behind it. A human in your team makes the call. Where you want a different model, for example fully automated for low-risk cases with human review for everything else, we scope that during Discovery.

The four outcome rules apply whether the work is done by a person or an agent. We design agents to produce evidence of outcomes, not just process compliance. Every customer interaction produces a logged record of the checks run, the alternatives considered, the vulnerable customer indicators surfaced, and the decision logic applied. Your conduct team has the evidence base to demonstrate good outcomes to the FCA.

Most likely not at the firm size we typically work with. The Critical Third Party regime is aimed at suppliers whose failure would cause systemic risk across the sector. We would still be a material outsourcing arrangement under SS2/21, which means full due diligence, contractual protections, concentration risk assessment, and an exit plan. We provide the documentation pack as standard.

Vulnerable customer indicators are surfaced for human review before any response is drafted. The agent does not close, dismiss, or auto-respond to any case where a vulnerability indicator is present. Indicators include recent bereavement, financial difficulty, indications of poor mental health, and patterns we agree with you during Discovery based on your customer base. Consumer Duty is explicit on this point and so are we.

We retrain. Regulatory change is part of the run phase. When supervisory expectations move, we update the agent's checks, evaluation criteria, and documentation accordingly. No additional charge for routine regulatory updates. Major reworks, for example responding to a new regime that did not exist when the agent was built, are scoped separately.

Build typically takes between four and twelve weeks depending on the workflow complexity, your existing systems, and the depth of evaluation we run before go-live. Most of the time is spent on the design and evaluation work, not the code. Smaller, well-scoped workflows can be live in under a month. Cross-system workflows with multiple integrations can run longer. We give you a calendar plan during Discovery.