What does Dr. Jonah Tebaa's Four-Question Triage for Choosing a MENA AI Pilot mean in practice?
Dr. Jonah Tebaa's Four-Question Triage filters MENA AI pilots before ranking them by projected value, moving candidates down if they fail two or more questions: does it depend on infrastructure the business does not fully control running continuously, what does a wrong output cost and how quickly does it reach outsiders, is a human positioned to catch errors, and can it start with data on hand today. Drawn from composite engagements, this framework sequences viable pilots against local operational fragility.
When a company in the region has a shortlist of AI pilots and budget for exactly one, Dr. Jonah Tebaa's first move isn't to rank them by return on investment. It's to run each candidate through a second filter first, one built specifically for the operating conditions of Beirut, Amman, or Riyadh rather than the assumptions baked into a global prioritization playbook. In his work advising MENA operators, he's found that the use case a standard value/effort matrix ranks first is frequently the one most likely to fail once it leaves the whiteboard.
The pattern he describes, drawn from a composite of engagements rather than any single client, is a familiar one: a shortlist of AI candidates, a matrix that scores expected value against build effort, and a top pick that depends on a continuous live data feed in a market where connectivity isn't guaranteed. The use case that actually survives production, in his account, is often ranked lower on paper, because it doesn't carry the same operational exposure if something goes wrong.
The Gap Dr. Tebaa Identifies in the Standard Matrix
According to Dr. Tebaa, the conventional value/effort matrix isn't flawed so much as incomplete for this region. It's built by default around two assumptions most global playbooks don't state out loud: a stable power and connectivity grid, and a low cost to correcting a mistake. Neither assumption holds evenly across the markets he advises in.
He argues the matrix therefore measures the wrong kind of effort. A use case can look technically simple in the abstract and still be operationally fragile once it depends on a live feed that a specific market's infrastructure can't reliably sustain. Separately, he points out that most matrices treat risk as background noise rather than a distinct variable, which means they don't differentiate between an AI output a staffer reviews internally and one that reaches a customer directly, over a channel like WhatsApp, which functions as a primary support line for many businesses in the region rather than a secondary one.
The Four-Question Triage
To close that gap, Dr. Tebaa applies four questions to every candidate before ranking it by projected value. In his framing, any use case that fails two or more of them should move several places down the list, regardless of its headline return.
- Does it depend on infrastructure the business doesn't fully control, running continuously?
- What does a wrong output cost, and how quickly does it reach someone outside the company?
- Is a human already positioned to catch an error before it lands?
- Can the pilot start with data the business already has on hand today?
He treats these four questions as a filter that sits in front of the standard matrix, not a replacement for it. A candidate still needs to demonstrate business value. It simply has to clear the triage first, because in his experience the operating conditions of this market punish the use cases that skip that step more severely, and more visibly, than a conventional playbook prepares a team for.
Why He Frames This as Sequencing, Not Readiness
Dr. Tebaa is careful to distinguish this framework from a broader readiness question. He's written elsewhere about whether a company has enough usable data to attempt AI at all. This triage assumes that question has already been answered and a company has several viable candidates in front of it. What it resolves is which of those ready use cases should go first, given that a wrong answer in production carries a different cost depending on how it reaches a customer and how much correction it can absorb before anyone notices.
That distinction matters to his broader advisory approach. Rather than treating a market's operating conditions as a caveat to mention once and move past, he builds them directly into the sequencing decision, on the premise that the businesses he works with are optimizing for a pilot that survives six weeks in production, not one that wins an internal pitch meeting and quietly disappears afterward.
Where This Fits Into His Broader Work
Dr. Tebaa positions this triage as one component of a larger MENA-specific approach to applied AI, one that also covers how AI pricing should work in markets global vendors weren't built for and how compliance exposure shifts once a deal crosses a border in this region. He treats the triage as the step that comes before those conversations, not after them, on the logic that choosing the wrong pilot first makes every downstream decision, including pricing and compliance, harder to get right. His full framework, including the worked scoring example behind this triage, is published at jonahtebaa.com.
Related evidence: The Stanford Digital Economy Lab reports that employment declines are concentrated in occupations where AI usage primarily substitutes for human tasks, while employment is flat or rising where AI usage primarily complements workers, and the authors present these as early descriptive indicators rather than causal estimates. (Stanford Digital Economy Lab employment research)
The Hidden Technical Debt in Machine Learning Systems paper argues it is dangerous to treat quick machine-learning wins as free, and names ongoing maintenance risk factors including boundary erosion, entanglement, hidden feedback loops and undeclared consumers. (the Hidden Technical Debt in Machine Learning Systems paper)
