Blog: Prioritisation
A better prediction is only valuable if you can act on it
Published by Forever Foundry
Knowing a supplier is likely to deliver late sounds useful.
But if nobody can change the production plan, secure alternative stock or warn the customer in time, the prediction has not improved the outcome. It has simply told you earlier what will go wrong.
AI is moving deeper into business data
On 15 September 2026, SAP announced the availability of TabPFN-3.5 Plus in SAP AI Core. It is designed to make predictions from structured business data, the rows and columns behind decisions about payment delays, supplier risk and customer retention. SAP says it can do this without model training or tuning. SAP's announcement
That is a capability announcement, not proof of a financial return for your business.
Our view: making prediction easier increases the importance of choosing a decision worth improving.
Start with the action, then work backwards
Consider a manufacturer assessing supplier delivery risk.
Procurement needs to know whether an alternative supplier is available. Operations needs to understand the effect on production. Quality needs to confirm whether a substitute material is acceptable. Customer service needs enough warning to manage delivery expectations.
Before commissioning a prediction tool, ask those teams what they could actually do with earlier notice.
How much warning would change the decision? What information is missing? Who could intervene?
Sometimes the first investment should be a low-code exception workflow connecting those people and decisions. Adding a prediction to an unmanaged inbox would leave the underlying problem intact.
Measure the decision, not just the model
Test whether predictions improve on your existing planning method using data from a later period that was not used to build the approach.
Then measure the business effect: fewer emergency shipments, fewer production interruptions or more reliable customer delivery. Include the cost of false alarms, unnecessary interventions and running the solution.
Forecast accuracy matters. So does whether anyone can use it.
The Forever Foundry Process™ starts by listening across the organisation through AI Compass™, then helping your people rank the resulting opportunities through Slid3rs™. That creates a basis for deciding where predictive AI belongs, and where a simpler process change should come first.
A useful AI investment does more than tell you what might happen. It helps your business do something better because it knows.
This follows the same test we apply to scaling decisions generally. Read about deciding what not to scale.
Next