What is the second line of defence?

The second line of defence is the risk management and compliance function that sets policy, monitors how the business manages risk and challenges it independently. It does not run the business's processes; it checks them. Once both the business and the second line use AI agents, keeping them separate becomes a design question, not just an organisation chart.

Why does the second line matter once agents act?

The second line is valuable because it is independent. Its people report through a separate route, do not own the processes they review and can say no. If a monitoring agent shares its brief, its sources, its owner or its write access with the agent doing the work, that independence quietly disappears.

The risk is practical. A shared agent platform makes it easy to reuse an agent built for the business to "also check" its own output. That is the first line marking its own homework, with an agent in the middle. We argue firms should keep the second line apart from the first, even when both use agents.

What does the second line do in practice?

Its usual responsibilities are:

• Writing risk and compliance policy and setting risk appetite with the board.
• Advising the business on regulatory requirements.
• Monitoring and testing whether controls work.
• Challenging risk assessments and decisions.
• Reporting on risk to senior management and the board.

Agents help most with monitoring and testing: sampling work, reading alerts, comparing actions to policy and drafting findings. That is the role we describe in AI compliance agents: second line, never the last word. The challenge, the judgement and the report to the board stay with people.

An example from a lender

Take a lender whose first-line affordability agent drafts credit decisions for underwriters to approve. The second line wants to test whether affordability checks follow policy. It deploys its own testing agent with its own brief, owned by the head of compliance monitoring, with read-only access to completed applications and the filed record of the first-line agent's actions.

The testing agent finds that in a set of applications, income verified from payslips was accepted where policy required bank statements. It drafts a finding. A compliance analyst confirms it and raises it with the head of lending. The testing agent never had access to change an application, and the first-line team could not alter the testing agent's brief. A second-line separation checklist for compliance agents lists the conditions to verify.

How is the second line different from the third line?

The third line is internal audit. It gives the board independent assurance over the whole system of risk management, including the second line itself. It reports to the audit committee and does not set policy.

The second line is part of management. It sets standards, monitors and challenges the business continuously. The third line examines periodically whether both the first and second lines are doing their jobs. If the second line's testing agent is poorly certified, it is the third line that should find out. The three lines of defence model describes how they fit together.

Which terms sit next to the second line?

• First line: the business teams that own and run risk.
• Independence: the separation of reporting lines, access and ownership that gives oversight its value.
• A three lines of defence matrix for AI agents assigns each agent to one line.