Agentic KYC

Know Your Customer (KYC) is the regulatory foundation of financial services - the set of processes by which institutions verify the identity, assess the risk, and monitor the activity of their customers. KYC was designed for a world in which every customer is a human being. That world is ending.

When an autonomous AI agent approaches a bank to open an account, initiate a payment, apply for credit, or execute a trade, the existing KYC framework breaks down at every level. The agent is not a natural person - it has no passport, no address, no date of birth. The agent is not a legal entity - it has no articles of incorporation, no registered directors, no beneficial ownership structure. The agent exists in a regulatory gap between natural persons and legal entities, and current KYC frameworks have no category for it.

This is not a theoretical problem. Autonomous trading agents already operate in financial markets. AI-powered payment orchestration systems process transactions without human involvement. Procurement agents negotiate and execute contracts on behalf of enterprises. These machine customers are transacting in the financial system today - but the KYC frameworks that govern that system were not designed for them.

Agentic KYC is the AXD Institute's term for the adapted compliance framework that financial institutions need to verify, monitor, and govern autonomous agents. It extends traditional KYC by adding three new verification layers: agent identity verification (who or what is this agent?), delegation chain verification (who authorised this agent and what are the boundaries of that authorisation?), and behavioural monitoring (is this agent operating within its delegated authority?).

The AXD Vocabulary includes the term Know Your Agent (KYA) - the principle that institutions must verify the identity and authority of autonomous agents with the same rigour they apply to human customers. Agentic KYC operationalises KYA into a practical compliance framework.

Agent identity verification. Every autonomous agent must have a verifiable identity - not a human identity, but a machine identity that can be authenticated, tracked, and audited. This requires agent identity standards: cryptographic credentials that prove the agent is what it claims to be, issued by a recognised authority, and revocable if the agent is compromised. The agent's identity must be distinct from its principal's identity - the bank must know both who the agent is and who it represents.

Principal verification. Agentic KYC does not replace traditional KYC - it extends it. The human or organisation behind the agent must still pass standard KYC checks. But agentic KYC adds a critical question: has this verified principal actually authorised this specific agent to act on their behalf? The delegation must be verifiable, not assumed. A stolen or forged delegation credential is the agent equivalent of identity fraud.

Authority scope verification. Traditional KYC asks: who is this customer? Agentic KYC adds: what is this agent authorised to do? An agent authorised to check account balances should not be able to initiate transfers. An agent authorised to trade equities should not be able to trade derivatives. The scope of delegated authority must be machine-readable, verifiable, and enforceable by the institution's systems.

Continuous monitoring. Traditional KYC includes ongoing monitoring of customer activity for suspicious patterns. Agentic KYC requires continuous monitoring of agent behaviour against its declared authority scope. An agent that begins operating outside its delegated boundaries - executing larger transactions than authorised, accessing accounts it should not access, or transacting at unusual times - must trigger alerts and potential suspension. Behavioural monitoring for agents must be automated, real-time, and calibrated to the agent's specific authority profile.

The most novel aspect of agentic KYC is delegation chain verification - the process of tracing an agent's authority back through every link in the chain to the original human or organisational authorisation. In complex B2B environments, delegation chains can be several links long: a board authorises a CFO, the CFO authorises a treasury team, the treasury team authorises a payment agent, and the payment agent authorises a sub-agent for a specific transaction type.

Every link in this chain must be verifiable. If any link is broken - if the CFO's authorisation has been revoked, if the treasury team has exceeded its delegated limits, if the payment agent's credentials have expired - the entire chain is invalid. Agentic KYC must verify the complete chain, not just the final agent.

This creates significant technical challenges. Delegation chains must be represented in machine-readable formats that institutions can verify programmatically. Each delegation must include: the identity of the delegator, the identity of the delegatee, the scope of delegated authority, the constraints and conditions, the expiry date, and the revocation status. Standards for delegation chain representation are still emerging - the AXD Institute advocates for open, interoperable standards that work across institutions and jurisdictions.

Delegation chain verification also raises questions about liability. When an agent exceeds its authority and causes financial harm, who is liable? The human principal who delegated? The intermediate entity that passed the delegation? The agent itself? The institution that accepted the agent's credentials? Agentic KYC must establish clear liability frameworks that incentivise proper delegation governance at every level of the chain.

Current financial regulations were written for human customers and legal entities. No major jurisdiction has yet published comprehensive regulations for autonomous agent identity, delegation verification, or agent-specific compliance. This regulatory gap creates both risk and opportunity for financial institutions.

The risk is that institutions that deploy or accept autonomous agents without adequate KYC frameworks may face regulatory action when regulations catch up. Regulators in the UK (FCA), EU (EBA), and US (FinCEN) are all monitoring the emergence of autonomous agents in financial services. Institutions that build agentic KYC frameworks now - before regulation mandates them - will be better positioned when requirements are formalised.

The opportunity is that institutions that develop robust agentic KYC frameworks can become the trusted infrastructure for agent-mediated financial services. Just as traditional KYC created a competitive advantage for institutions with efficient onboarding processes, agentic KYC will create advantages for institutions that can verify and onboard autonomous agents quickly, securely, and compliantly.

The AXD Institute recommends that financial institutions begin developing agentic KYC capabilities now, using the following principles: treat agent identity as a first-class compliance requirement (not an edge case), build delegation chain verification into existing KYC infrastructure, implement real-time behavioural monitoring for agent activity, and engage with regulators proactively to shape emerging standards. The institutions that wait for regulation will be playing catch-up. The institutions that lead will define the standards.

Key regulatory developments to monitor include the EU AI Act's provisions for high-risk AI systems in financial services, the UK FCA's work on AI governance in regulated firms, and emerging international standards for machine identity and agent credentials. Agentic KYC sits at the intersection of financial regulation, AI governance, and digital identity - three domains that are converging rapidly.

The AXD Institute's framework for agentic KYC systems is built on the same trust architecture principles that govern all of Agentic Experience Design, adapted for the specific requirements of financial compliance.

Layered verification. Agentic KYC should implement verification at four layers: identity (is this agent what it claims to be?), authority (is this agent authorised to perform this action?), compliance (does this action comply with all applicable regulations?), and behaviour (is this agent operating consistently with its historical patterns?). Each layer adds confidence. All four layers together provide comprehensive assurance.

Risk-proportional scrutiny. Not all agent actions require the same level of KYC scrutiny. A balance enquiry requires basic identity verification. A large cross-border payment requires full delegation chain verification, compliance screening, and enhanced behavioural monitoring. Agentic KYC systems should calibrate scrutiny to consequence level - applying the AXD principle of consequence-proportional design.

Interoperable standards. Agentic KYC must work across institutions and jurisdictions. An agent verified by one bank should be recognisable to another bank, subject to the receiving institution's own risk assessment. This requires open standards for agent identity, delegation representation, and authority scope - standards that the financial services industry must develop collaboratively.

Human override. Every agentic KYC system must include human override capabilities. When the automated system cannot verify an agent's identity, authority, or compliance, a human compliance officer must be able to review the case, request additional information, and make a judgment. The human override is not a failure of automation - it is the safety net that makes automation trustworthy.