AI readiness assessment

Start with the right diagnosis

Connect the topic to a decision and a business outcome. Gather facts before choosing a solution.

• Describe the task, frequency, cost, errors, and exceptions.
• Classify data by sensitivity and quality.
• Define where human approval is mandatory.

A practical approach

Work in short stages, with clear ownership and criteria to stop, correct, or continue.

• Choose a reversible, limited-risk use case.
• Compare a simple rule, conventional automation, and AI.
• Test with controlled data and acceptance criteria.
• Monitor quality, drift, security, adoption, and total cost.

What to measure

Choose a small set of indicators directly tied to the original problem. Measure before, during, and after the change.

• Net time saved
• Accuracy
• Exception rate
• Total cost
• Incidents and adoption

Common mistakes

A technically possible solution is not necessarily the best economic or operational decision.

• Exposed confidential data
• Unverified output
• Automating a bad process
• No accountable owner

Checklist to use

• Describe the task, frequency, cost, errors, and exceptions.
• Classify data by sensitivity and quality.
• Define where human approval is mandatory.
• Choose a reversible, limited-risk use case.
• Compare a simple rule, conventional automation, and AI.
• Test with controlled data and acceptance criteria.
• Monitor quality, drift, security, adoption, and total cost.

Useful sources

• OECD — AI adoption by small and medium-sized enterprises
• NIST — AI Risk Management Framework

Last reviewed: 2026-08-13