Agentic Payments
Agentic payments advisory helps institutions govern payments in which software, not a person, makes the decision at the moment of payment, within limits set in advance. Payment Labs advises banks, payment providers and corporates on delegated authority, control architecture, evidence and liability, and on which payment flows justify an agent.
Four issues shaping Agentic Payments
Authorisation shifts from a moment to a mandate
Customer authentication was designed around explicit human action at the point of payment. With agents, authority is granted in advance as a mandate, so institutions must verify both the agent's identity and the scope of the principal's authority. The IMF identifies authorisation and traceability as central design questions for agentic payments.
Probabilistic agents require deterministic controls
The main risk is not the use of probabilistic reasoning itself but allowing adaptive systems to make irreversible payments without proper controls, checks or accountability, and it argues that the payment core should remain deterministic. In practice, limits, payee checks and approvals must be enforced in the payment path, outside the agent's own reasoning.
Liability and accountability remain unsettled
Existing rules on consent, authentication and unauthorised transactions were written around human behaviour. Until rules settle, institutions need records of authority, context and decisions to support disputes and audit.
Correlated behaviour and cyber exposure raise resilience risk
Opacity, correlated agent behaviour and cybersecurity among the risk channels, alongside legal uncertainty. Many agents acting on similar signals can produce concentrated, simultaneous payment activity, and manipulated instructions can redirect an agent's spending. Institutions should test for both before volumes grow.
What our Agentic Payments advisory
delivers
01
Use-case triage and delegation ladder
02
Agent authority and mandate model
03
Agent identity and authentication design
04
Evidence and dispute framework
05
Liability and regulatory mapping
06
Fraud and risk-control design
07
Protocol and rail roadmap
08
Limit-enforcement architecture
How an engagement works
Assess
Triage flows and review current controls and liability exposure.
Design
Define authority, limits, evidence and protocol approach.
Deliver
Support build, testing and controlled launch.