Value Tracking Framework
Start with Value
The majority of AI investments made today are unmeasured. Organisations build, deploy, and expand AI capabilities without a governed mechanism to attribute value, track costs, or demonstrate financial return. This creates a fragile foundation, one that struggles under board scrutiny, resists proper capital allocation, and cannot distinguish a genuinely transformative AI programme from an expensive experiment that happened to coincide with a good business quarter
The FernAI Value Tracking Framework exists to close that gap. It is a structured, stage-by-stage approach to measuring the true financial return of AI, from the earliest ideation conversations through to perpetual, live ROI observability in production
The three principles this framework is built on
- Value must be defined before anything is built
- Not all outcomes are caused by AI
- Funding should follow evidence
Stage 1 : Ideation
Define value hypotheses, model costs and returns, score and prioritise use cases before any build begins, understand dependancies
Stage 2 : Pilot
Run time-boxed experiments with OKRs, weekly cost and return tracking, causality weighting, and funding gates tied to milestone performance
Stage 3 : Implementation
Perpetual ROI observability. Live cost and return attribution. Quarterly gate reviews for scale, sustain, optimise, or sunset decisions.
Causality weighting, the framework’s most important discipline
Causality weighting is the mechanism by which observed outcomes are discounted to reflect only the portion credibly attributable to the AI intervention. It is the difference between a number that can withstand a CFO’s scrutiny and one that cannot
The causality weight is multiplier added to the value between 0 and 1, set by the cross-functional measurement team, representing the probability that the AI intervention, and not other concurrent factors caused the observed outcome
Causality weights are not static. As evidence accumulates, longer time series, cleaner controls and more comparable external data, weights should be updated. Weights can go up or down. All changes are documented with rationale. This is the audit trail if the numbers are ever challenged
Continuous cost & returns tracking
During implementation, all cost streams are tracked monthly at a minimum, with compute and model costs tracked in near-real-time via cloud billing and API usage dashboards. Any cost line exceeding +15% of the monthly forecast triggers an automatic review.
Return streams are tracked against the KRs established at the pilot stage, with causality weights applied monthly. The measurement owner produces a monthly attribution report that feeds into the quarterly board pack
