AI readiness · Pillar 2 of 3

Process: AI agents need clear instructions. People need clear control.

Design one process so that it is clear what AI does, what people approve, and what happens when something goes wrong. Adding AI to an old process does not automatically improve it. Organisations need to decide which tasks AI can do, what people must check, who approves important actions, and what happens when the AI gets something wrong.

Five steps to prepare one process for AI

Start with one process, not the whole organisation.

Map the work
Write down how the work is done today: each step, who does it, what information it uses, and which decision it ends in.
Decide who does what
For each step, decide what the AI may do on its own, what it may only suggest, and what stays with a person. As AI does more of a step, the person’s part changes: first they do the work, then they check the AI’s work, and in time they may only approve it. Decide who checks at each stage, and give newer staff practice in checking, with an experienced colleague to discuss their decisions.
Set where a person checks
Name the point where a person checks the AI's work, what they check, and who signs off before anything leaves the team.
Test what happens when the AI is wrong
Run the process with a wrong or incomplete AI answer. See who notices, how quickly, and what it costs.
Measure the result
Agree before you start what will show that the new process works, for example errors caught, time taken or decisions reversed. Check it on a set date.

A fictional example

Buying from suppliers, with an AI tool in the process

Fictional example. Who checks and who signs off will differ in your organisation.
Task	What the AI may do	What a person must check	Who signs off
Compare supplier quotes	Collect the quotes and build a comparison	That every quote covers the same work, and what the comparison left out	Category manager
Recommend a supplier	Rank suppliers on the agreed criteria	The evidence behind the ranking and the conditions in each contract	Budget holder
Flag unusual invoices	Mark invoices that differ from the contract	Each flagged invoice, before any payment is stopped	Finance or compliance owner

When the AI is an agent that acts

Some AI tools now act, not only advise. They send messages, update records or place orders. A process that uses an AI agent needs four more answers.

Which systems and tools the agent may use.
Which actions need a person's approval first, for example any payment above a set amount.
What the agent records, so that a person can see what it did and why.
Who can stop it, and how quickly.

The checks should match the consequence. An agent that drafts a reply needs fewer checks than an agent that pays a supplier.

How AIR APAC helps

Process Review

A fixed-scope review of one process. We map the work, design who does what with AI, test the design against cases where the AI is wrong, and hand over a plan your team or any builder can carry out.

See the Process Review

Process readiness briefing

A working session with your leaders on one process you plan to change with AI. We go through the five steps together. Your team writes a map of who does what, where a person checks, and who signs off.

Training for the people in the process

The people who check AI work need practice. Programmes for teams and business functions include Team Decisions with AI, AI in Operations and Supply Chain and AI Governance for Boards.

Does AIR APAC redesign the process for us?

Yes, one process at a time, in a Process Review, with the people who do the work. If you would rather redesign it yourselves, the briefing teaches your team the method. AIR APAC does not build software or set up AI agents. Exeter Labs, the product company of AIR APAC, can do that under a separate agreement, or you can use any other supplier.

Do we need to use AI agents?

No. The five steps apply to any AI tool, from a chat assistant to an agent that acts on its own.

Delivery partners