The strongest AI strategy is not to put AI everywhere. It is knowing where AI belongs, where it does not and what should remain human.

​

​

Map the work before choosing the technology​

 

Use this framework when a team is asking where AI could be added, but the work has not yet been mapped task by task.

​

A job rarely changes in one neat block. AI usually changes individual tasks, decisions, handovers, evidence and accountability.

​

That is why the starting question matters.

​

​

Do not start with: Where can we put AI?

​Start with: Where does AI actually belong in this work?

​​​

 

The four work decisions 

​

Once the work has been mapped, each task or decision can be considered in one of four ways.

 

1. ​Automate

Use AI for repetitive, rules-based and checkable work where the output can be verified.

​

The important point is verification. Automation should not remove the organisation’s ability to check whether the work has been completed correctly.

​

2. Assist

Use AI to reduce cognitive load through tasks such as summarising, searching, preparing, organising or comparing information.

​

The technology helps someone complete the work, but it does not take ownership of the decision or its outcome.

​

3. Augment

Use AI to support skilled judgement by structuring evidence, identifying patterns or presenting relevant information.

​

A person still considers the evidence, makes the decision and owns the outcome.

​

4. Keep human

Keep the work human when it involves trust, ethics, relationships, exceptions, reputation, wellbeing or final accountability.

​

Human judgement is not automatically an inefficiency. In many situations, it is the control the business needs.

​

 

When to be careful

​

• Do not add AI to a process simply because the technology is available.

• Do not replace a working process if AI would add complexity without improving the outcome.

• Do not treat human judgement as an inefficiency when it provides necessary oversight, context or accountability.

• Do not assume an entire role should be automated because AI can complete one or two of its tasks.

 

The leadership check

​

Use these questions with your leadership team before deciding where AI belongs.

If an answer is vague, leave it blank and treat that as useful evidence.

​

Task or decision

What specific part of the work are we looking at?

​

Current problem

What is slow, risky, costly, inconsistent or painful today?

​

AI role

Should this task be automated, assisted, augmented or kept human?

​

Human accountability

Who checks the work, makes the decision, explains it or owns the outcome?

​

Do not change if  

What would make us leave this process alone for now?

 

​Ownership 

Who owns the outcome, adoption, cost and risk?

​

If the work cannot be mapped clearly

​

The team should be able to explain the task, the current problem, the proposed role of AI and the human accountability that remains.

​

If the work cannot be mapped clearly, the project is not ready for approval.

​

Take a Human Pause first.

​

​