Every company produces information.
Organizational intelligence begins when that information can change a decision while there is still time to change the outcome.
The familiar four stages are useful only when they describe a real capability, not a maturity badge. A company does not become predictive because it owns a model, or prescriptive because a dashboard displays a recommendation.
- 01 Descriptive What happened?
- 02 Diagnostic Why did it happen?
- 03 Predictive What is likely next?
- 04 Prescriptive What should change?
1. Descriptive: the business can describe the past
Descriptive intelligence collects and organizes facts. It can show that complaints increased, margin fell, delivery slipped, stock moved, or sales slowed.
This is necessary. Without reliable description there is no stable reality for the next stage to examine. But description alone is a mirror pointed at the past. The owner can see the consequence after the business has already absorbed it.
Visibility is not yet steering.
2. Diagnostic: the business can recover context
Diagnostic intelligence connects a visible result to the conditions that produced it. “Complaints increased” becomes a pattern tied to a customer segment, a promise, a handoff, missing context, and an unclear decision owner.
This stage needs more than data infrastructure. It needs shared definitions, process meaning, evidence quality, rules, and ownership. The company must be able to distinguish a plausible explanation from one supported by reality.
Without this layer, prediction is dangerous. A business that cannot explain why something happened should be careful when asking a model what will happen next.
3. Predictive: the business can see early enough
Predictive intelligence identifies a pattern before it becomes the only fact everyone is forced to react to: a customer drifting, a supply risk emerging, a complaint cluster forming, or margin pressure building.
This is where AI and machine learning can create genuine value, provided the first two stages are strong enough. Prediction built on weak description produces noise. Prediction built on weak diagnosis produces confident wrong answers.
The value of prediction is not foresight as theater. It is time to act.
4. Prescriptive: the business can recommend action
Prescriptive intelligence does not mean the system replaces judgment. It means the business can produce a recommendation close enough to the decision to change what happens next.
A responsible recommendation is specific, timed, owned, evidence-backed, and honest about its limits. It names the action, the decision owner, the uncertainty that remains, and the result that must return as learning.
At this stage the company begins to steer, not because the machine decides alone, but because intelligence is finally connected to decision, action, and consequence.
The danger of skipping a stage
Many transformation efforts try to become prescriptive before they are diagnostic. They automate before they understand, build models before definitions are stable, and ask for recommendations before decision ownership is clear.
The result looks advanced but leaves the owner no closer to truth. More output is produced while the same decisions still arrive late.
A missing stage is therefore not a reason to add another feature. It is a precise design constraint: strengthen the place where reality, context, or ownership is failing.
Every domain starts again
The four stages are not a ladder the whole company climbs once. Pricing may be diagnostic while complaints remain descriptive. Production may predict one constraint while customer decisions still depend on instinct.
Each decision domain has to earn its way forward. The evidence that supports a sales action cannot simply be borrowed for a supply decision. Definitions, timing, ownership, and learning have to be true in that part of the business.
The owner’s question
The useful question is not which stage sounds advanced. It is which stage the company can honestly sustain when a real decision has to be made.
- Can the business describe reality accurately?
- Can it recover the context behind what happened?
- Can it see early enough for action to matter?
- Can it recommend a bounded action and learn from the result?
Move forward by strengthening the stage where truth is breaking.