Agentic identity governance now extends to customer and partner actions

At a glance

What this is: Strivacity is framing agentic identity as a customer and partner governance problem, with controls for authorization, consent, auditability, and revocation.

Why it matters: IAM teams need to extend governance beyond human sign-in flows because delegated AI actions create accountability and authorization gaps across customer, partner, and non-human identity programmes.

By the numbers:

• By 2028, Gartner projects that 90% of B2B buying will be AI-agent intermediated, representing more than $15 trillion in B2B spend flowing through AI agent exchanges.

👉 Read Strivacity's announcement on agentic identity governance for AI agents

Context

Agentic identity is the problem of governing actions taken by AI systems on behalf of another party, especially when those actions include transactions, account access, and decisions that affect customer or partner relationships. Traditional CIAM and IAM patterns were built to authenticate people and authorise their own sessions, not to prove who delegated authority to a machine actor or what that actor may do at runtime.

This release positions customer and partner identity as the control point for agentic activity. That matters because once an AI agent can act outside the enterprise boundary, the governance question is no longer only access control. It becomes consent, delegation, auditability, and the ability to revoke authority without dismantling the identity stack already in place.

Key questions

Q: How should security teams govern AI agents acting on behalf of customers or partners?

A: Security teams should govern delegated agents with explicit authorization boundaries, recorded consent, and revocation paths. The goal is to separate the actor that initiates the session from the actor that is allowed to make decisions, then preserve a verifiable chain of custody for every action the agent takes.

Q: Why do AI agents complicate customer identity governance?

A: AI agents complicate customer identity governance because they can act outside the normal human login model while still affecting accounts, transactions, and data. That breaks the assumption that the authenticated user is also the acting party, so governance must cover delegation, accountability, and proof of intent.

Q: What breaks when organizations rely on standard session-based access for AI agents?

A: Standard session-based access fails when the agent can make multiple decisions within a single session and the organisation cannot show which actions were authorised. The result is weak auditability, unclear accountability, and a higher chance that fraud or misuse will be hard to dispute or contain.

Q: Who is accountable when an AI agent performs an unauthorised action?

A: Accountability should be assigned to the authorising party, the controlling organisation, and the governance process that allowed the scope to exist. If the organisation cannot prove who approved the action and what limits were in force, it does not have sufficient identity governance for agentic access.

How it works in practice

Transactional authorization for AI agents

Transactional authorization limits an agent to specific permitted actions and can require explicit approval for higher-risk steps. In practice, it is closer to scoped delegation than to broad account access, because the system must distinguish between independent actions and actions that still need a human decision point. The architectural challenge is preserving intent at runtime while still allowing machine-speed execution. For customer and partner journeys, this usually means mapping actions to policy conditions rather than granting an open-ended session.

Practical implication: define which agent actions are autonomous and which remain approval-bound before the agent ever touches customer or partner accounts.

Audit trails and chain of custody for delegated actions

Auditability for agentic identity must attribute each action to both the agent and the individual who authorised it. That creates a chain of custody that supports investigation, dispute handling, and compliance evidence. Standard logs that only record the technical session are not enough when the business question is who delegated the authority and whether the action matched that delegation. This is especially important when agents act across multiple systems, because the accountability record must survive orchestration hops and token exchange boundaries.

Practical implication: require logs that bind agent activity to the authorising identity, action scope, and outcome across every system touchpoint.

Know your agent and adaptive access controls

Know Your Agent verification is an onboarding control for external or third-party agents before they are allowed to operate against customer systems. Combined with adaptive access controls, it allows risk to influence what an agent can do at a given moment, rather than treating all agents as equally trusted once a token exists. This pattern fits a world where the enterprise may not control the agent itself, only the permissions and evidence around its use. The important shift is from static trust to continuous verification of agent identity and behaviour.

Practical implication: treat external agents like governed actors, not just applications, and gate access on identity, risk, and purpose.

NHI Mgmt Group analysis

Agentic identity is now a customer governance problem, not just an authentication problem. Once an AI agent can act on behalf of a customer or partner, the identity programme has to govern delegation, consent, and accountability together. That shifts the centre of gravity from login to authorised action, which is where most CIAM stacks were never designed to operate. Practitioners need to treat delegated machine action as a first-class identity event.

Who authorised the agent is now as important as who the agent is. Traditional identity models assume the subject of the transaction is the same party that authenticated. Agentic commerce breaks that assumption because the acting identity and the beneficiary identity can differ, and both matter to governance. This is where customer identity, partner identity, and non-human identity meet in one control plane, and the implication is that delegation evidence must become auditable artefact, not implied context.

Transactional authorization is the right pattern because blanket session trust is too coarse for agentic behaviour. The article points toward a more precise model where authority is scoped to specific actions and approval thresholds. That aligns with OWASP-NHI and zero trust thinking, but the deeper point is that agent behaviour cannot be governed with static account grants alone. Practitioners should expect more policy granularity, not more implicit trust.

Agentic identity exposes a named governance gap: delegation without proof of intent. The old assumption was that if a session existed, authority was already established and durable enough to govern later. That assumption fails when agents execute on behalf of another party and can take actions faster than humans can review them. The implication is that identity governance must prove intent at the moment of delegation, not infer it from the existence of a token or session.

External agents need lifecycle governance, not just access provisioning. The release highlights KYA verification, permission revocation, and auditability because agent access is not a one-time event. The field is moving toward managed agent lifecycles, where onboarding, scope control, monitoring, and offboarding all matter. Practitioners should re-evaluate whether their current lifecycle processes can actually revoke delegated authority cleanly when the acting party is non-human.

From our research:

• 80% of organisations report their AI agents have already performed actions beyond their intended scope, according to AI Agents: The New Attack Surface report.
• 52% of companies can track and audit the data their AI agents access, leaving 48% with a complete blind spot for compliance and breach investigation.
• Ultimate Guide to NHIs explains how to extend lifecycle, audit, and governance controls across machine identities and agentic systems.

What this signals

Agentic commerce will force IAM teams to move from identity proof to authority proof. When delegated AI systems can transact, current customer identity models will not be enough on their own. The next programme gap is proving whether an action was authorised, not merely whether a session was authenticated, and that will affect CIAM, fraud, and partner governance together.

With 80% of organisations already reporting agent scope drift, the operational signal is clear: access review cycles are too slow for delegated machine action. Teams should expect more demand for real-time policy, stronger audit bindings, and lifecycle controls that can revoke authority cleanly when an agent changes role or behaviour.

Delegated AI action creates a new lifecycle category. It is neither classic human access nor standard workload identity. Organisations that already manage NHI lifecycle will be better positioned, because the same questions apply, who owns it, what can it do, how is it monitored, and how is it removed when trust ends.

For practitioners

• Define delegation boundaries for agent actions Map customer and partner agent use cases into explicit allow, deny, and approval-required actions before deployment. Use transaction-level policy rather than broad account entitlements so you can separate low-risk automation from high-risk decisions.
• Bind every agent action to authorising identity Require audit records that capture the agent, the human or system that authorised it, the scope approved, and the outcome observed. Keep those records consistent across orchestration layers so investigations can reconstruct the chain of custody.
• Add agent onboarding and offboarding controls Treat external agents like governed actors with a lifecycle, including verification before access, periodic review of permissions, and a revocation path that removes authority without breaking unrelated customer access.
• Review customer identity flows for machine participation Identify where customer-facing journeys now include AI agents, then decide which steps still require human confirmation, which can be delegated, and which need additional fraud and compliance checks.

Key takeaways

• AI agents acting on behalf of customers and partners create an identity governance problem centered on delegation, not just authentication.
• Evidence from NHIMG research shows the governance gap is already active, with 80% of organisations reporting agent scope drift and 48% lacking full audit visibility.
• Practitioners should respond by scoping agent authority, binding actions to authorising identity, and treating delegated access as a managed lifecycle.

Key terms

• Agentic Identity: Agentic identity is the governance model for an AI system that acts on behalf of another party. It extends identity controls to delegated action, so teams can verify authority, limit scope, and preserve accountability when the actor is a machine rather than a person.
• Transactional Authorization: Transactional authorization is permissioning tied to a specific action rather than to broad session access. It lets teams define what an agent may do, where approval is required, and how to prevent a valid session from becoming an open-ended delegation path.
• Chain of Custody: Chain of custody is the evidence trail that shows who authorised an action, what scope was approved, and what the actor actually did. In agentic identity, it must connect the agent's execution to the delegating identity so disputes and investigations can be resolved.
• Know Your Agent: Know Your Agent is the verification step used to establish trust in an external or third-party agent before granting access. It is the agentic equivalent of onboarding evidence, helping organisations decide whether a machine actor is sufficiently known, controlled, and scoped for the task.

Deepen your knowledge

NHI governance, agentic AI identity, and machine identity security are core topics in our NHI Foundation Level course, the industry's only accredited NHI security programme. If you are responsible for identity strategy, governance, or operational control, it is worth exploring.

This post draws on content published by Strivacity: Strivacity for Agentic AI expands identity governance for AI agents. Read the original.