Agentic AI in financial services needs delegation, not impersonation

At a glance

What this is: NHI Mgmt Group's analysis, drawing on Raidiam's blog post: Agentic AI in financial services needs delegation, not impersonation, and agent-level identity becomes the trust boundary.

Why it matters: IAM, IGA and PAM teams need to treat autonomous agents as governed actors with explicit authority, because human-centric identity patterns do not preserve auditability or accountability at agent speed.

By the numbers:

• Gartner forecasts that by 2028, autonomous agents will drive 20% of digital storefront interactions and make 15% of daily work decisions in financial services.

Context

Agentic AI in financial services refers to software that can make and execute decisions with limited or no human intervention. The governance problem is that many banking, payments and fintech platforms still assume a human actor at the centre of every authorisation flow.

Raidiam's panel discussion frames a familiar IAM failure mode in a new setting: impersonation works poorly once an agent must act repeatedly, transparently and at scale. The article argues that delegation and agent-level identity are the practical response when trust, accountability and compliance must survive autonomous execution.

The implication is not that all automation becomes autonomous identity. It is that once an AI system independently selects actions in a live financial workflow, the identity model must be designed around the agent rather than the person it imitates.

Key questions

Q: What breaks when agentic AI is treated like a human user in financial services?

A: Human-style impersonation hides the real actor, blurs accountability and weakens auditability. In regulated finance, that becomes a governance failure because permissions, logs and reviews are tied to a person while the autonomous system is the one making decisions. Delegation is the cleaner model because it gives the agent explicit scope and traceable authority.

Q: Why does delegation reduce risk for autonomous agents in banking and payments?

A: Delegation reduces risk because it limits authority to a defined purpose instead of letting an agent inherit broad human permissions. That creates clearer logs, better compliance evidence and less ambiguity when decisions affect customer journeys, payments or risk controls. The control objective is not just access, but provable accountability for each action.

Q: How do security teams know whether an agent identity is actually governed?

A: An agent identity is governed only when teams can identify the owner, locate the credentials, define the allowed scope, and revoke access without hunting across endpoints or backup files. If any of those pieces are missing, the identity is partially shadowed. The practical signal is whether access can be answered in minutes, not days.

Q: What role does digital public infrastructure play in agentic AI trust?

A: Digital public infrastructure provides shared, standards-based trust rails so agent identity and delegation work consistently across services and organisations. Without it, every integration becomes a local exception, which increases fragmentation and weakens auditability. For practitioners, the value is interoperability with governance rather than interoperability alone.

Technical breakdown

Why impersonation breaks in agentic finance

Impersonation means an AI system acts as if it were a human user, reusing that user's identity context to reach downstream systems. That may be workable for narrow automation, but it becomes brittle when the system makes repeated decisions, changes pace dynamically, or needs provable accountability across multiple actions. In financial services, that erodes auditability because the system's real decision-maker is obscured by the borrowed identity. It also weakens access control because the permissions were granted to a person, not to an autonomous runtime actor.

Practical implication: stop treating human impersonation as a durable authorisation pattern for autonomous workflows.

What delegation changes for authorisation and audit

Delegation gives an agent its own bounded authority for a specific purpose, rather than borrowing a person's session or credentials. That matters because authorisation can then be expressed as agent-scoped permissions, unique identity records and transaction logs that show what the agent was allowed to do and why. This is closer to NHI governance than traditional consumer identity, even when the use case begins in customer journeys. The real control shift is from user authentication to agent accountability.

Practical implication: define agent-scoped permissions, logs and approval boundaries before letting agents touch regulated workflows.

Why digital public infrastructure matters for agent identity

Digital public infrastructure provides the shared rails for interoperable identity, trust and policy enforcement across organisations. In this article's context, that means financial institutions can avoid inventing isolated agent identity models that break once systems cross organisational or sector boundaries. Interoperability is not a nice-to-have here. It is what allows an agent to be recognised, constrained and audited consistently across multiple services without collapsing back into implicit trust.

Practical implication: align agent identity design with interoperable trust infrastructure instead of isolated platform-specific controls.

Threat narrative

Attacker objective: The objective is to exploit opaque impersonation paths so autonomous actions occur without clear accountability, auditability or authority boundaries.

1. The article describes a trust failure in which an AI system can be made to act through human-style impersonation rather than explicit delegation.
2. Once the agent operates under borrowed identity, accountability and authorisation become opaque because the permissions belong to the person, not the autonomous actor.
3. That opacity can lead to unintended financial exposure when agent actions are not bounded by unique identity and auditable scope.

Breaches seen in the wild

• Entra ID actor token flaw (CVE-2025-55241): Hidden Actor tokens plus an Azure AD Graph validation flaw could have let attackers become Global Admin in any Entra ID tenant (CVE-2025-55241).

Read and download The State of NHI & AI Agent Breach Report 2026, covering 200+ breaches impacting Non-Human Identities including AI Agents.

NHI Mgmt Group analysis

Delegation is the first real identity boundary for agentic finance: once an AI system can choose actions at runtime, the security question is no longer who it copies but what authority it receives. Human impersonation hides that boundary and leaves audit trails attached to the wrong actor. Financial services teams should therefore treat agent identity as a governed control plane, not an implementation detail.

The assumption that identity can be provisioned around a person collapses under agentic execution: human-centric IAM was designed for users whose authority is stable enough to authenticate, review and certify. That assumption fails when an agent acts independently and repeatedly on behalf of a business process. The implication is that governance has to move from user substitution to explicit machine-authorised delegation.

Know Your Agent becomes the practical analogue to knowing the customer or the employee: the article's KYA idea reflects a broader shift from identity proofing to actor accountability. In financial services, that means organisations need to know not only who approved the agent, but what scope, purpose and oversight model govern its actions. Without that, compliance artefacts describe the human sponsor while the autonomous actor remains opaque.

Digital public infrastructure is a governance enabler, not just an interoperability story: open standards matter because agent identity is only useful if it can survive across services and counterparties. Closed, ad hoc controls create local trust islands that do not scale to banking, payments or cross-sector workflows. Practitioners should read DPI as a signal that agent governance will increasingly depend on shared trust rails, not isolated platform logic.

From our research library:

• 53% of security leaders expect AI to run major portions of their infrastructure autonomously within the next three years, according to the 2026 Infrastructure Identity Survey.
• 40 percent of financial and software companies have already deployed agentic AI systems, and deployments are expected to double by 2028.
• Read next: AI Agent Authorisation Guide

What this signals

Agentic finance needs a delegation-first operating model: when an AI system can initiate and complete actions independently, the control problem shifts from authenticating a user to constraining a machine actor. Teams should expect authorisation design, logging and approval flows to be rewritten around agent scope rather than session reuse.

Identity proofing alone will not solve autonomous workflows: the article points toward a broader KYA model where the question is not only whether an agent is real, but whether its authority, purpose and oversight are explicit. That changes how IAM, PAM and compliance teams evidence control over regulated actions.

Financial services leaders should treat interoperable trust infrastructure as a prerequisite for scaling agentic AI across products and counterparties. The governance challenge is not one platform feature but the ability to preserve consistent delegation boundaries as agents move through banking, payments and customer operations.

For practitioners

• Define agent-scoped delegation Map each autonomous use case to a specific purpose, maximum authority and explicit decision boundary, rather than extending a human session into machine execution.
• Separate impersonation from identity Prohibit production patterns where an agent reuses a user's credentials or session for ongoing actions, and replace them with a unique auditable agent identity.
• Instrument audit trails for agent actions Log the agent, the delegated scope, the triggering context and the downstream system touched so reviewers can reconstruct who acted and under what authority.
• Align governance to regulated workflows Review payment, onboarding and customer-service flows for places where autonomy changes the required approval model, accountability chain or evidentiary record.

Key takeaways

• Agentic AI in finance exposes the limits of human impersonation because autonomous systems need explicit authority, not borrowed identity.
• Delegation, unique agent identity and auditable scope are the controls that make autonomous financial workflows governable.
• Digital public infrastructure matters because agent governance has to survive across systems, counterparties and regulated processes.

Key terms

• Agent Identity: An agent identity is the set of attributes, credentials and permissions assigned to an autonomous software entity. It is treated as a non-human identity because it can authenticate, act on systems and accumulate access over time, which creates governance, audit and lifecycle obligations similar to other production identities.
• Delegation authority: Delegation authority is the right for one identity to act on behalf of another within a defined scope. For agents, it must be explicit, time-bound, and auditable because the system may initiate actions independently once granted access.
• Know Your Agent: A governance pattern for verifying that an AI agent is the intended actor, not a hijacked or impersonating one. It extends identity assurance beyond login by focusing on runtime behaviour, delegated authority, and the integrity of the agent's actions across connected systems.
• Digital Infrastructure: Digital infrastructure is the underlying technology and process foundation that supports online services, transactions, and identity workflows. It includes the systems, standards, and operational capabilities needed for secure digital interactions. For onboarding and KYC, it determines how reliably organisations can verify users, exchange data, and scale services across channels.

Deepen your knowledge

NHI governance, agentic AI identity, and machine identity lifecycle are core topics in our NHI Foundation Level course, the industry's only accredited NHI security programme. If you are building or maturing an IAM programme, it is worth exploring.