A framework by Farid Mheir
AI Fog Index
A score, from 0 to 5, for how much uncertainty remains around an AI use case. And what a leadership team should do about it.
I reply personally. No obligation. No mailing list.
The question it answers
For a defined AI application, what is understood, what remains uncertain, what does that mean for the business, and what should an executive consider or challenge?
The answer is an analysis, not just a number. Its most useful part is often an overlooked prerequisite, a warning about a claimed benefit, a better alternative, or a question to put to the team proposing the project. The score is a consistent shorthand for the uncertainty that remains.
Eight criteria
Every use case is examined against the same eight criteria, each scored on its own.
Technical maturity and predictability
How well the technical basis, its capabilities and its limits are understood.
Operational performance and scalability
How well quality, failure, security and behaviour at scale are understood.
Business value and economics
What is known about benefits, lifecycle costs and net value against the alternatives.
Deployment and integration requirements
How well the prerequisites, the work and the recurring obligations are understood.
Legal, regulatory and accountability conditions
How well obligations, rights and responsibilities are understood for the activity and the jurisdiction.
Human and social conditions
How well adoption, human interaction and the distribution of benefits and harms are understood.
Political, market and ecosystem conditions
How well supplier dependencies, policy direction and market continuity are understood.
Environmental and resource conditions
What is known about resource demands and environmental consequences against the alternatives.
The score
Each criterion is scored from 0 to 5. The overall Fog is their average, with equal weights, shown to the nearest half-point.
0 · Negligible uncertainty
1 · Low uncertainty
2 · Moderate uncertainty
3 · High uncertainty
4 · Very high uncertainty
5 · Fundamental uncertainty
Uncertainty is not risk
A high score means substantial unresolved uncertainty. It does not mean that the technology is dangerous, expensive, immature, or undesirable. A low score means little uncertainty remains. It does not mean that deployment is advisable.
So each criterion also carries a judgment: favorable, mixed, unfavorable, or insufficient basis for judgment. Strong evidence that a workflow costs more than its alternatives gives low Fog and an unfavorable judgment. A promising prototype with little deployment evidence gives high Fog and a cautiously favorable one.
What you get
An assessment ends with conditional guidance, written warnings, and questions to ask. The guidance falls into four categories.
Consider deployment
External evidence supports serious consideration for the defined arrangement.
Run a bounded pilot
A limited, reversible experiment can resolve the material questions at acceptable exposure.
Monitor
There is potential, but action is not justified yet, or a dependency should be watched.
Avoid in the stated context
Evidence supports not proceeding under the described conditions.
No score threshold picks the category. A high-Fog proposition can warrant a pilot when exposure is limited and the learning is valuable. A low-Fog negative result can warrant avoiding the application.
Fog and Leading with AI
The Fog Index looks outward: what comparable deployments have required, what has been demonstrated, and under which conditions. Leading with AI looks inward, at your organization’s mindset, skillset, toolset and actionset: are those conditions present here?
Together they are the two frameworks of my own that anchor the AI work I do through FMCS. See the consulting work →
See a sample AI Fog Index report.
One assessed use case: the eight criterion scores, the judgments, the guidance and the warnings. Leave your email and I will send it to you personally.