How ready are you for agents that act?
Most organisations can already run a model. Far fewer can let its output change something without a person retyping it. Twelve questions, about four minutes, and the score is on screen the moment you finish — no form in the way.
Data foundation
An agent cannot act on a number it cannot get, or trust one it cannot trace.
Decision clarity
Automation needs a decision with a known owner, a known input set and a known success measure.
Governance and trust
The question is never whether the model is clever. It is who is accountable when it is wrong.
Execution path
A recommendation that a human must retype into another system is not automation.
What the score actually tells you.
A low score is not a verdict. It usually means one dimension is holding the others back, and that is a smaller problem than it looks.
Reporting
The organisation can describe what happened. Decisions are made from exports and experience. Start by writing down one recurring decision and the inputs it really uses — not the ones it is supposed to use.
Analysing
Prediction exists somewhere, usually in one team. The gap is the handover: the forecast is produced, then a person carries it to the system where the action happens.
Recommending
Recommendations reach the people who act on them, with reasoning attached. The remaining work is usually governance: agreeing which actions may proceed without a signature.
Acting
Agents can be given real scope: bounded decisions they execute, logged and reversible, with people supervising the exceptions rather than the routine.
Get the written breakdown.
The score tells you roughly where you stand. The breakdown tells you which single change moves it most, based on what we see in organisations that answered the way you did.
- Your four dimension scores, with the two questions dragging each one down.
- The decision we would automate first in an organisation shaped like yours.
- What would have to be true before that decision could run without a person.