The four-level maturity model.

This file is one self-contained piece of the AI IR Overlay™ framework. Cross-references to other pieces point to other packages in the same set, which you can obtain at jacobideji.com.

The AI IR Maturity Roadmap

Four levels. One question at each: can you do this, on demand, under pressure?

This is the framework view of maturity, the model used to assess and benchmark organizations. The operating view, with cadences, drills, and pitfalls, is in Playbook 20: AI IR Maturity Roadmap (Operating View).

The Four Levels

                       ┌──────────────────────────────┐
                       │   Level 4  RESILIENT         │
                       │   Continuous improvement     │
                       │   "We get better every Q"    │
                       └──────────────▲───────────────┘
                                      │
                       ┌──────────────────────────────┐
                       │   Level 3  PROVABLE          │
                       │   Evidence under pressure    │
                       │   "Here is what happened"    │
                       └──────────────▲───────────────┘
                                      │
                       ┌──────────────────────────────┐
                       │   Level 2  CONTAINABLE       │
                       │   Stop harm without stopping │
                       │   "We're already contained"  │
                       └──────────────▲───────────────┘
                                      │
                       ┌──────────────────────────────┐
                       │   Level 1  AWARE             │
                       │   Inventory of agents/tools  │
                       │   "We know what we run"      │
                       └──────────────────────────────┘

Level Definitions

Level	One-sentence definition	Test
1. Aware	The organization has basic visibility into its AI assets.	Can you produce a current inventory of every agent → identity → tool → write target in under 5 minutes?
2. Containable	Harm can be contained without a complete shutdown.	Can Tier-1 SOC activate Modes M1 to M4 within 10 minutes, in production, without escalation?
3. Provable	The organization can demonstrate scope under time pressure.	Can the team export the Minimum Evidence Set within 60 minutes for any agent?
4. Resilient	Continuous improvement with measured recovery.	Are the Six Metrics (Playbook 13) trending in the right direction over rolling 90 days?

Mapping the Levels to Framework Controls

Level	Required controls (see `framework/01-minimum-viable-overlay.md`)
1. Aware	MVO-1 Inventory (current)
2. Containable	MVO-1 + MVO-2 Safe Modes (M1 to M4 implemented and tested)
3. Provable	Level 2 + MVO-3 Minimum Evidence Set (A–F exportable in 60 min)
4. Resilient	Level 3 + MVO-4 Controlled Re-Enable + quarterly tabletops + measured metrics

The Honest Self-Assessment

A common failure mode is over-claiming. The Overlay’s stance:

If a capability has never been tested in the last 90 days, you do not have it.

Apply this rule to every level:

Level 2 means M1 to M4 tested in production within 90 days, not “documented in the runbook.”
Level 3 means the Minimum Evidence Set was actually exported within 60 minutes within 90 days.
Level 4 means tabletops happened and metrics moved.

Most organizations who claim Level 3 are honestly at Level 1.

Why This Matters for the Board

Board question	Level required to answer “yes”
Do we know which AI we run?	Level 1
Can we stop harm fast?	Level 2
Can we tell the regulator what happened?	Level 3
Are we measurably getting better?	Level 4

The Board-Ready Scorecard (Playbook 24) maps directly to these.

Operating the Roadmap

The operating cadence, drill design, common pitfalls, and the 30-minute single-agent reality check are detailed in Playbook 20: AI IR Maturity Roadmap (Operating View). Start there for execution. Playbook 13 (Six Metrics) carries the trend-line measurements that Level 4 depends on, and Playbook 14 (Testing for Agent Failure Modes) carries the drill discipline that produces the Level 2 and Level 3 evidence.

Source: AI IR Overlay newsletter, Issue #20, “AI IR Maturity Roadmap,” by Jacob Ideji. https://www.linkedin.com/in/jacobideji/