Monday moves for AI governance: five steps in 90 days

At a glance

What this is: NHI Mgmt Group's analysis, drawing on C1.ai's blog post: Practical AI governance starts with intake, telemetry, containment, reuse, and measurement rather than waiting for a full programme buildout.

Why it matters: It matters because IAM, PAM, and governance teams need controls that work before AI sprawl, shadow AI, and unmanaged access paths outrun the official process.

👉 Read C1.ai's article on Monday moves for AI governance: five steps in 90 days

Context

AI governance in this article means the operating controls that decide what gets approved, logged, wrapped, reused, and measured before the organisation lets AI into production workflows. The article argues that governance fails when teams treat those controls as a later-stage programme instead of the first deployment guardrail.

The core problem is not the absence of ambition. It is the mismatch between how quickly AI projects move and how slowly identity, logging, and approval paths are usually formalised. That gap is exactly where shadow AI grows, because the fastest path to access tends to become the real path.

For IAM and identity governance teams, the article is really about sequencing. It says a workable first 90 days can establish enough control to reduce unmanaged access, even before the wider AI programme is mature.

Key questions

Q: How should security teams start AI governance when there is no mature programme yet?

A: Start with a single intake funnel and one visible owner for every AI project. That gives you inventory, accountability, and a place to attach controls before access fragments across business units and shadow channels. The first goal is not perfect policy. It is a governed front door that every AI request must cross.

Q: Why does unmanaged AI become a governance problem so quickly?

A: Because the unmanaged path is often faster than the governed one, and users choose speed when approvals, logging, or access requests are slow. Once that happens, shadow AI becomes the default operating model. The risk is not theoretical. It is a process design failure that rewards bypassing formal control.

Q: What do teams get wrong about audit logging for AI tool use?

A: Teams often log the server error but not the identity event. For MCP, the useful record is which agent called which tool, on whose behalf, with what arguments, and when. Without that detail, incident response and compliance review become reconstruction exercises instead of evidence-led investigation.

Q: When should organisations use identity-aware wrapping instead of rewriting an application?

A: Use wrapping when an application has high blast radius and you need governed access quickly without waiting for a full rebuild. The wrapper creates a control layer around the legacy app, reduces exposure, and buys time to modernise later. It is a containment move, not a final architecture.

Technical breakdown

Why a single AI intake funnel matters

A single intake funnel is a governance choke point that forces every AI project, whether built, bought, or internal, through one visible process. Technically, that means projects are not approved through scattered email threads, ad hoc tickets, or business-unit exceptions that no one can reconcile later. The value is not perfect policy enforcement on day one. The value is a durable record of what entered the AI estate, who owns it, and whether it is already under review. That creates inventory, accountability, and traceability before controls fragment.

Practical implication: route every AI request through one intake path before any build or purchase work begins.

Why audit logging must come before policy

The article reverses a common mistake: organisations often write policy first and discover telemetry gaps later. In practice, policy without logging is unenforceable because no one can answer what agents did, when they did it, or which systems they touched. A central audit log creates the evidence layer for future rules, reviews, and incident analysis. For identity teams, this is the difference between governing behaviour and guessing at it after the fact. The technical issue is observability before control automation.

Practical implication: establish central audit logging for in-production AI before you codify governance rules.

How identity-aware proxy wrapping contains blast radius

Identity-aware wrapping sits in front of legacy applications and mediates access without forcing an immediate rewrite. Technically, it interposes a control layer that can preserve existing application behaviour while changing how access is authenticated, logged, and de-risked. The article links this to the three highest-blast-radius apps first, which is sensible because blast radius is a governance problem as much as a technical one. The approach also deprecates older service-account paths, which reduces the chance that legacy access routes remain the easiest route into sensitive systems.

Practical implication: place identity-aware controls in front of high-risk applications before you attempt deeper app modernization.

NHI Mgmt Group analysis

Shadow AI is fundamentally a governance-speed problem, not just a discovery problem. The article makes clear that unmanaged AI wins when the governed path is slower than the unsafe one. That means the real failure is not only missing inventory, but a process design that makes official access harder than unofficial access. Practitioners should treat path-speed as a control attribute, not just a convenience metric.

Single-funnel intake is the right first control because it creates an enforceable identity boundary for AI programmes. Without one front door, ownership fragments across business units, platform teams, and improvisational purchases. That fragmentation is exactly how AI estates become unreviewable. The implication is that governance needs a visible entry point before it needs sophistication.

Fastest-path-delta is a useful named concept for AI governance maturity. The article’s weekly measurement model shows that the safest route must not lag the unmanaged route by more than 25% or shadow AI returns. That is not just a performance metric. It is a governance signal that access design, approval design, and telemetry design are out of alignment. Teams should measure control speed with the same seriousness they apply to control coverage.

Identity-aware wrapping shows that containment can outrun reconstruction. The article correctly prioritises mediated access for the highest-blast-radius applications instead of waiting for full app rewrites. This validates a pragmatic governance pattern for large estates: reduce exposure first, then modernise. For identity programmes, the message is that control placement matters more than architectural purity when risk is already in production.

Reusable agent assets point to the next phase of governance maturity. Once teams can intake, log, and contain, the organisation can start standardising the parts that every team keeps reinventing. That shifts AI governance from one-off approvals to governed reuse, which is where scale becomes manageable. The practitioner conclusion is simple: standardisation is a control strategy, not just an efficiency play.

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.
• Read next: Agentic AI Identity Maturity Model

What this signals

Fastest-path-delta: governance breaks when the unmanaged path is materially faster than the safe one. That gap turns policy into theatre, because users route around controls that slow them down. Identity programmes should treat control latency as a first-class risk signal, not an implementation detail.

The next maturity step is to make AI governance measurable in operational terms. If teams cannot prove intake coverage, logging coverage, and contained access coverage, they do not yet have a governable AI estate.

For practitioners

• Define one AI intake funnel Create a single front door for every AI project, including build, buy, and internal use cases, and require enough fields to establish ownership and initial risk triage.
• Stand up central audit logging first Make telemetry available before policy finalisation so you can answer what agents did this quarter from one query and prove which systems were touched.
• Wrap the highest-blast-radius applications Place identity-aware proxy controls in front of the three applications where a failure would cause the most damage, then retire old service-account paths on a dated schedule.
• Publish reusable AI assets Create one real agent, one MCP tool for a frequently used enterprise system, and one governance template that other teams can adopt without rebuilding the pattern.
• Measure the fastest path weekly Track time-to-credential, time-to-first-agent, and time-to-tool-access, then treat any 25% or greater gap between safe and unsafe paths as a governance incident.

Key takeaways

• AI governance fails early when intake, logging, and access mediation are added after projects are already moving.
• The article’s method prioritises speed and visibility, showing that governance has to be deployable in weeks, not years.
• For identity teams, the immediate task is to make the governed path faster, clearer, and more reusable than the unmanaged one.

Key terms

• Single AI Intake Funnel: A single AI intake funnel is one governed entry point for every AI project, whether built internally, purchased, or introduced by a business unit. It creates an auditable record of ownership and scope before access fragments across informal channels.
• Fastest-Path Delta: The gap between the quickest governed access route and the quickest unmanaged route. In identity programmes, it is a practical measure of whether people and systems will choose the controlled path or route around it when speed matters.
• Identity-aware Proxy: An identity-aware proxy combines routing with authentication and authorization logic. It checks tokens or certificates, applies policy at the edge, and forwards verified identity context to the backend so applications do not have to re-implement security decisions inconsistently.
• Reusable Agent Asset: A reusable agent asset is a standardised AI component, tool wrapper, or governance template that can be consumed by multiple teams instead of being recreated repeatedly. It supports controlled reuse, clearer ownership, and more consistent governance across the estate.

What's in the full article

C1.ai's full article covers the operational detail this post intentionally leaves for the source:

• The exact five-step, 90-day sequencing the author recommends for AI governance rollout.
• The role assignments behind each move, including which existing functions should own intake, logging, and measurement.
• The target metrics for funnel coverage, log coverage, and fastest-path performance.
• The quarter-by-quarter maturity path from first deployment to reusable assets and reporting.

👉 C1.ai's full post covers the 90-day sequencing, ownership model, and operational targets behind the methodology.

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.