Is your business process ready for AI automation?

Start with one repeated task

Choose a task you can describe from trigger to finished output: turning meeting notes into an action list, for example. Identify who uses the result and what currently causes delay. “Use AI across the business” is too broad for a useful first test.

Check six foundations

• Opportunity: does the task repeat, and is the current delay or effort worth investigating?
• Process: can someone explain the steps, exceptions and approval owner?
• Information: are the necessary records available, accurate enough and permitted for this use?
• Systems: can information be accessed and transferred safely, even if the first test is manual?
• People: is there an owner who can review outputs and manage changes?
• Risk: what happens if an answer is wrong, incomplete or disclosed to the wrong person?

Choose the simplest useful route

Clear rules and structured inputs often suit ordinary automation. AI may be worth testing when the work involves varied text, summarising or drafting. Missing information, unclear ownership or high consequences can make process repair and stronger controls the first step. Paper or Excel alone does not disqualify a business.

Example: meeting notes to a reviewed action list

Fictional example: a small project team has meeting notes but no consistent action register. First, agree fields for action, owner, due date and status. An AI-assisted draft could extract candidate actions; missing owners or dates must be flagged, not invented. A project lead checks the draft against the notes before sharing it. Restricted information needs an approved handling method before any tool is used.

Define a small test before choosing software

• Use permitted, representative samples, including incomplete notes and difficult exceptions.
• Record the current time and rework so there is a baseline.
• Agree acceptance checks: correct actions, no invented facts, visible unknowns and required human approval.
• Compare accuracy, review time and total effort with the current process. Stop or redesign if review effort outweighs the benefit.

What a readiness result means

The free SIGMA check provides a preliminary profile and practical next actions. It does not approve feasibility, security or compliance. A separately scoped review can examine evidence, map the workflow and define requirements. Any pilot or implementation depends on confirmed capability, feasibility and agreed scope.

Method and further reading

This practical guide draws on the risk-oriented approach of the NIST AI Risk Management Framework and its Govern, Map, Measure and Manage playbook. These voluntary resources support context-specific decisions; this guide and the SIGMA check are not NIST certification or a formal conformity assessment.