Access management in 2026 fails when governance stays centralized

At a glance

What this is: This is an analysis of why access management fails at scale and how distributed decision-making, lifecycle automation, and RBAC reduce risk.

Why it matters: It matters because IAM, IGA, PAM, and NHI programmes all depend on the same control model, and that model breaks when access grows faster than governance.

By the numbers:

• 80% of organisations report their AI agents have already performed actions beyond their intended scope, including accessing unauthorised systems, inappropriately sharing sensitive data, and revealing access credentials.
• 96% of technology professionals identify AI agents as a growing security threat, and 66% believe this risk is immediate.

👉 Read Cakewalk's analysis of access management, RBAC, and AI agent governance

Context

Access management is the discipline that decides who or what can reach systems, data, and privileges, and this article argues that the discipline fails when access is created at the edge while governance stays centralised. In fast-growing environments, that mismatch turns requests, approvals, and reviews into bottlenecks that cannot keep pace with the business.

The primary identity issue here is not policy quality but operating model design. Human access, service accounts, and AI agents all accumulate risk when lifecycle controls are manual, role clarity is weak, and exceptions become the default way work gets done. That is a common scale pattern, not an edge case.

The article also treats AI agents as a growing non-human identity problem because they expand access faster than human teams can review, document, and revoke it. Once agents are attached to internal and external systems with standing permissions, the same governance gap that affects people becomes harder to see and faster to exploit.

Key questions

Q: How should security teams govern access when growth outpaces the IT team?

A: They should move from manual approval queues to lifecycle-driven governance. Joiner, mover, and leaver triggers should drive access changes, while roles and policy determine what access is granted. That keeps decisions close to the business event and prevents ticket backlogs from becoming the de facto control.

Q: Why do AI agents create more risk than traditional automation?

A: AI agents create more risk because they can interpret context, choose actions, and invoke tools autonomously. Traditional automation follows fixed rules, but an agent can be manipulated into using its own authority in unintended ways. That makes permission scope, tool boundaries, and monitoring more important than model accuracy alone.

Q: What breaks when access reviews are the main control for fast-growing companies?

A: Reviews become retrospective paperwork if permissions are changing faster than the review cycle. By the time the review happens, access may already be stale, duplicated, or invisible across ticket systems and spreadsheets. Lifecycle automation is what keeps reviews meaningful.

Q: Who should own access decisions in a distributed access model?

A: Policy should stay central, but decisions should sit with the people who understand the business need, such as managers or app owners. The security team should define guardrails, enforce evidence capture, and review exceptions that exceed policy.

Technical breakdown

Why centralised access governance fails at scale

Centralised governance assumes that the approving team can see the business context, the request, and the downstream risk at the same time. That assumption breaks when access is created in product teams, engineering workflows, and contractor relationships faster than a central queue can process it. The result is not just delay. It is degraded decision quality, because the queue becomes the policy and speed replaces scrutiny.

Practical implication: move approval authority closer to the business event while keeping policy, evidence, and enforcement centrally governed.

How ticket-driven access becomes standing privilege

Ticket systems work well for tracking work, but they are a weak control plane for identity lifecycle. When tickets become the mechanism for approvals, people optimise for closure, not correctness. Temporary access becomes permanent because no lifecycle trigger forces removal, and every exception creates a new baseline that is hard to unwind.

Practical implication: tie access changes to joiner, mover, and leaver events rather than treating every change as a manual request.

Why AI agent access amplifies NHI governance risk

AI agents behave like non-human identities with broad, fast-moving reach across systems, vendors, and data paths. Unlike humans, they can be provisioned quickly, used by multiple operators, and left with standing permissions that outlive the original business need. That expands the NHI problem from credential management into accountability, ownership, and runtime scope control.

Practical implication: govern agents as privileged NHIs with explicit ownership, time-bounded access, and auditable system-of-record evidence.

Threat narrative

Attacker objective: The attacker objective is to preserve broad, unreviewed access long enough to increase blast radius and reduce the chance of timely detection or revocation.

1. Entry occurs when access is created through tickets, side channels, or distributed team requests without a consistent lifecycle trigger. Escalation happens as temporary exceptions remain active and accumulate into standing privilege across people and AI agents. Impact follows when audits, reviews, and offboarding cannot reconstruct who approved what, leaving excessive access in place and invisible.
2. The objective is to keep work moving while bypassing the controls that would otherwise constrain privilege growth and evidence quality.
3. The breach evidence pattern most relevant here is long-lived exception drift rather than a single exploit, because the failure is operational accumulation.

Breaches seen in the wild

• Coupang Signing Key Breach — Unrevoked signing key credentials expose 33.7 million records after employee offboarding failure at Coupang.
• Meta AI Instagram Account Takeover — 20,225 Instagram accounts hijacked via compromised Meta AI support chatbot with overprivileged access.

Read our 52 NHI Breaches Analysis report for a comprehensive view of breaches impacting Non-Human Identities including AI Agents.

NHI Mgmt Group analysis

Access management is now an operating model problem, not an approval problem. The article is right that the central failure is structural: access is created where work happens, but governance still expects a central team to keep pace. That model collapses as headcount, apps, contractors, and AI agents expand. The practitioner conclusion is that governance has to move with the business event, not behind it.

Ticket queues are not just inefficient, they become a hidden policy engine. Once the backlog determines what gets approved, organisations stop enforcing least privilege and start optimising for throughput. That is how rubber-stamping becomes normal and why standing privilege grows one exception at a time. The practitioner conclusion is that queue design is an access control issue, not a service desk issue.

Role clarity plus lifecycle automation is the real control pair. RBAC defines what access should look like, while lifecycle triggers define when it should change. Without both, automation only accelerates overprovisioning. The practitioner conclusion is that access governance must be built around role templates, HR signals, and revocation logic, not one-off approvals.

AI agents turn access governance into NHI governance by default. When an agent can act across internal systems and external vendors with standing permissions, the old assumption that access review will catch drift becomes weaker. This is exactly where NHI programmes need to absorb agent access before it becomes invisible sprawl. The practitioner conclusion is to treat agent access as privileged non-human identity, not as a special automation case.

Identity blast radius is the right concept for growth-stage access risk. The article surfaces a useful named concept: the amount of damage one identity can cause when permissions, ownership, and evidence are loosely governed. That blast radius grows with every stale role, unrevoked exception, and unmanaged agent. The practitioner conclusion is that controlling blast radius is now a core identity objective, not a secondary audit outcome.

From our research:

• 80% of organisations report their AI agents have already performed actions beyond their intended scope, including accessing unauthorised systems, inappropriately sharing sensitive data, and revealing access credentials, according to AI Agents: The New Attack Surface report.
• 52% of companies can track and audit the data their AI agents access, while 48% still operate with a blind spot for compliance and breach investigation.
• For the broader control model, see Ultimate Guide to NHIs and lifecycle processes for managing NHIs for the access governance patterns that need to extend into agent oversight.

What this signals

Identity governance teams should expect access controls to be judged by how well they survive growth, not by how well they look on paper. As apps, contractors, and AI agents multiply, the meaningful question is whether access can still be tied to lifecycle events and revoked without manual chasing. The control objective shifts from approval volume to removal certainty, which is where most programmes still struggle.

Identity blast radius: the practical measure of how far one identity can move before governance catches up. In growth-stage environments, that blast radius expands when exceptions age, roles blur, and AI agents inherit human-style permissions without the same accountability boundaries.

With 52% of companies able to track and audit AI agent data access, the other 48% are already one investigation away from a reconstruction problem. That makes audit evidence design a runtime control, not a compliance afterthought, and it aligns closely with the NIST AI Risk Management Framework and OWASP Agentic AI Top 10.

For practitioners

• Map access creation to lifecycle triggers Tie joiner, mover, and leaver events to automated access changes so permissions do not depend on someone remembering a checklist.
• Replace ticket queues with policy-backed decision paths Keep approvals close to the business context, but enforce the rule set centrally and capture evidence automatically for every access grant.
• Classify AI agents as privileged NHIs Assign explicit owners, document intended systems, and require time-bounded permissions for every agent before production use.
• Measure access hygiene with operational signals Track time to revoke after departure, percentage of time-based access, and age of exceptions older than 30 days.
• Make senior access subject to the same controls Require executives and founders to follow the same approval, logging, and revocation process so exceptions do not set the real standard.

Key takeaways

• The core failure is structural: access is created at the edge while governance stays centralised.
• Lifecycle automation and clear roles matter more than adding approval headcount, because they stop exceptions from becoming standing privilege.
• AI agents should be governed as privileged NHIs, with owners, time-bounded access, and audit evidence built into the access model.

Key terms

• Access control operating model: The set of roles, approvals, evidence paths, and change processes that keep an access control system trustworthy in production. It defines who can change policy, how exceptions are handled, and how failures are recovered, which is often more important than the policy syntax itself.
• Standing Privilege: Standing privilege is access that remains active even when no immediate task requires it. For NHI programmes, it is a common failure mode because long-lived credentials and persistent roles create unnecessary exposure. Reducing standing privilege usually means tighter expiry, on-demand access, and clearer review of who or what still needs access.
• Identity Blast Radius: The amount of damage a compromised identity can cause across systems, data, and infrastructure. In NHI environments, it is shaped by permissions, network reach, and administrative capability rather than by the credential alone. Reducing blast radius is a containment strategy that limits lateral movement and data exposure.
• Lifecycle Trigger: A lifecycle trigger is the event or source signal that causes an access change, such as a role update, termination, or transfer. If the trigger is stale, missing, or poorly governed, the identity system can keep access alive long after the business need has ended.

What's in the full article

Cakewalk's full article covers the operational detail this post intentionally leaves for the source:

• The conversation-level examples behind the distributed decision model for access approvals.
• The practical sequence for tying RBAC to joiner, mover, and leaver events.
• The specific metrics Peter Kovacs uses to measure access hygiene and audit readiness.
• The internal approval path pattern for AI initiatives and non-human access requests.

👉 Cakewalk's full article covers the access operating model, lifecycle triggers, and non-human identity controls in more detail.

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 security strategy or NHI governance in your organisation, it is worth exploring.