Dr. Jonah Tebaa has spent much of his recent advisory work watching the same failure pattern repeat across otherwise well-run companies: an AI system is deployed to handle a slice of real operational work, and within months it is either drifting into tasks nobody assigned it, absorbing blame for outcomes outside its actual remit, or getting micromanaged into near-uselessness after a single visible mistake. His diagnosis is not technical. It is structural.
In his work with founders and operations leaders across Lebanon, the wider MENA region, and internationally, Tebaa argues that organizations already possess the discipline needed to prevent this. They apply it to every human hire, in the form of a job description: a defined scope, a named manager, a review schedule. What they consistently fail to do is apply that same discipline to the AI systems now doing comparable work inside the business. The AI gets access. It rarely gets a role.
Tebaa's proposed fix is deliberately unglamorous. Rather than a governance program or a committee structure, he advocates for what he calls a role charter: a short, writable document built around six specific elements, completed before an AI system's first day of real operational work.
The six-item charter
According to Tebaa, each AI role deployed inside a live workflow should have:
- A precise title and scope — naming the exact decisions or tasks owned, with the same specificity as a human job title, rather than a vague functional label.
- A single named reporting line — one accountable human, not a team or a department, responsible for the system's output.
- A defined escalation trigger — a concrete, auditable condition (a dollar threshold, a request type, a customer tier) rather than an unmeasurable standard like "when uncertain."
- A scheduled review cadence — a calendared audit, not a reactive one, on the logic that some failures never generate a complaint on their own.
- A performance metric aimed at quiet underperformance — something that surfaces scope creep or drift, since visible errors tend to report themselves.
- A renewal clause — a fixed date to expand, narrow, or retire the role, treating AI deployment as a role to be periodically reassessed rather than a permanent installation.
What distinguishes Tebaa's framing from the broader AI-governance conversation is its scale. He is explicit that this is not a six-month compliance initiative. It is a document that takes a few minutes to write, using management logic his audience already applies daily to people. The relief, in his view, is that the fix for AI drift was never a bigger AI system. It was better management of the one already in place. The labour data supports treating this as a management problem rather than a replacement one: the ILO's refined global index of occupational exposure to generative AI concludes that "As most occupations consist of tasks that require human input, transformation of jobs is the most likely impact of GenAI" — and transformed roles are exactly what a role charter exists to define.
For an audience of founders, COOs, and department heads who are operating AI in production rather than experimenting with it in isolation, Tebaa's argument reframes a technical anxiety as a familiar managerial task, one most leaders already know how to do well.
