Is AI Eroding Capability? Why Capability Erosion Belongs on the Board Register

The gap between apparent capability and actual capability inside organisations is widening.

While AI tools generate polished, finished-looking work at unprecedented speed, real accountability remains inherently human. The operational risks of unverified AI reliance are already surfacing in courts and corporate balance sheets:

• Legal Accountability: In Moffatt v. Air Canada, a tribunal established that organisations own what their AI systems say, rejecting the airline's attempt to disown erroneous chatbot advice (Moffatt v. Air Canada, 2024).

• Operational Vulnerability: A finance employee at engineering firm Arup transferred $25.6 million USD after joining a deepfake video call where every face and voice was an AI representation. Seeing a leader say something no longer counts as verification (CNN Business, 2024).

• Sycophancy and Echo Chambers: AI models are structurally biased toward agreement and flattering responses. In executive hands, an unconfigured AI assistant automates the echo chamber - offering no competing agenda and no willingness to voice hard truths (Sharma et al., 2023).

An unconfigured AI assistant in a chief executive's hands is no longer just a productivity experiment; it is an unmanaged governance gap.

The 5 Organisational Tripwires: Spotting the Slide

Boards govern in the language of risk, and leadership capability erosion must be treated with the same urgency as cyber risk.

Without deliberate design, organisations slide from intentional AI integration into an unmanaged default. Watch for these five tripwires that indicate capability drift has begun:

1. L&D Budget Cuts via AI Rationale: Cost savings are visible this quarter, but the capability deficit surfaces years later.

2. Facilitation Retreats to Ceremony: Human leadership development shifts from continuous practice to an occasional annual set-piece.

3. Climbing Ratification Rates: AI recommendations are consistently ratified without human scrutiny or interrogation.

4. Absence of Reversals: Zero senior decisions over the past quarter modified or reversed an AI recommendation. Unanimity with the tool is a symptom of ceded judgement.

5. Atrophied Manual Cycles: The organisation can no longer run a critical decision cycle without its AI tools, closing the recovery path.

Action: Assign a Named Owner for Capability Risk

Capabilities don't erode overnight; they lapse when responsibility is diffuse.

Place capability risk directly on the board risk register and assign it a named owner.

Give this owner clear leading indicators tied to the five tripwires, establish explicit decision rights, and enforce out-of-band confirmation rules for high-stakes actions. Until capability erosion has a named owner, it remains everyone’s concern and no one’s job.

Govern AI at the Board Level

Treating human judgement as an asset worth insuring protects your organisation against market volatility and systemic AI dependence.

👉 Download the full ‘The capability you should never hand to a machine’ Whitepaper to access the complete board risk framework, decision-rights templates, and protocols for funding protected human judgement reserves.