What is Operational Assurance?
Keeping execution aligned with intent.
Operational Assurance is the continuous capability for identifying and managing the gap between what an organisation intends and what actually happens—while work is taking place and outcomes can still be influenced.
Operational Intent is what the organisation wants to happen: intended outcomes, requirements, responsibilities, authority and boundaries.
Operational Reality is what people and AI agents actually do under real operating conditions—shaped by pressure, information, handoffs, tools and judgement.
The Execution Gap is the material difference between the two.
Not every variation is a failure. Assurance focuses on material divergence and leaves room for legitimate judgement and authorised adaptation. The risk lies in gaps that are invisible, unmanaged or allowed to persist.
Operational Intent
What should happen
The Execution Gap
Where material divergence occurs
Operational Reality
What actually happens
The Operational Assurance Model
Five stages, applied continuously
Adapt feeds learning back into Understand.
- 1
Understand
Establish Operational Intent for the selected work.
- 2
Observe
Establish Operational Reality from permitted evidence.
- 3
Compare
Identify whether any divergence is material.
- 4
Assure
Act at Assurance Points with the Minimum Effective Intervention.
- 5
Adapt
Learn from results and improve the work, the assurance or the intent.
Assurance Points
Selected moments, not blanket monitoring.
An Assurance Point is a material moment—an approval, a handoff, an exception, an agent action—where activity can be compared with Operational Intent and, where necessary, influenced. Placing assurance at a few high-leverage moments is more effective, and less intrusive, than watching everything.
A short example
A procurement team intends every package to use the current template and carry the required evidence. In practice, deadlines lead people to reuse old templates and attach evidence late. Rather than adding a new review layer, an Assurance Point at submission prompts for the current template and missing evidence—while the package can still be fixed. The same intent applies whether a person or an AI agent prepares the package, though each may need a different form of assurance.
Begin with one workflow
Where is execution falling short of intent?
Tell us about one important workflow or AI agent. We’ll help identify a useful starting point.