The AI Readiness Audit: Find the Workflow Before the Model

Organizations often begin AI planning with a vendor comparison. The better starting point is a workflow that people already perform, a bottleneck everyone can see, and an outcome the business values. Model choice comes after the work is understood.

Map the current job end to end

Document triggers, inputs, systems, decisions, handoffs, exceptions, and the final definition of done. Include the unofficial spreadsheets and messages that keep the process alive; they usually reveal the real integration requirements.

• Volume and time spent per case
• Quality errors, rework, and waiting time
• Availability and ownership of source data
• Decisions that require judgment or formal approval
• Systems the solution must read from or write to
• A baseline metric that can prove improvement

Choose a bounded first win

Good pilots have frequent examples, observable outputs, accessible data, and a human recovery path. Avoid starting with a rare, highly regulated, organization-wide process simply because it sounds strategically important.

Plan the operating model

Decide who reviews failures, updates knowledge, approves permissions, monitors cost, and owns the result after launch. AI readiness is as much about operational responsibility as technical capability.