When NOT to Use AI

The useful question is not whether AI can do a whole job. Most jobs contain several kinds of work. Those parts do not all deserve the same approach.

A customer meeting, for example, includes gathering background, organizing notes, writing a follow-up, making promises, and managing the relationship. AI can help with some of that work. A person should remain responsible for the rest.

Split the Job

Start by separating labour from judgment. Labour includes finding information, organizing it, checking records, drafting, and repeating clear steps. Judgment includes understanding a relationship, accepting risk, making a promise, and deciding what good looks like.

Then place each part in one of three groups. AI can run it, prepare it for review, or leave it with the person who owns the decision.

1. Run Automatically

AI can run a task when the steps are clear, the work repeats, and a mistake can be caught and corrected without serious harm. Examples include sorting routine requests, checking whether required details are missing, and preparing a standard internal update.

Running automatically does not mean running invisibly. The task still needs a record of what happened and a clear way to raise a problem. If the steps never change, a simple rule may be safer and easier than AI.

2. Prepare for Review

This is where AI often removes the most work without taking the decision away from a person. AI can gather the background, compare records, organize choices, and prepare a draft. The responsible person checks the work and makes the call.

Meeting preparation, customer follow-up, research, planning, and first drafts often fit here. The work may matter a great deal. That is why AI prepares it and a person remains accountable for the result.

3. Keep Human

Keep the work human when it can build or break trust. Do the same when the business makes a meaningful promise or the final choice depends on taste, values, or risk. This includes important customer conversations, pricing decisions, hiring, legal commitments, and the direction of the business.

AI can still prepare a brief or show the available options. It should not pretend to own the relationship or the decision.

When to Skip AI

Skip AI when the setup and upkeep will take more effort than the job. Skip it when you do not have the needed information or a simple checklist already solves the problem. Not every task needs a new system.

This is an important form of discipline. A small number of useful workflows is better than a collection of half-built experiments that someone must remember to maintain.

One Job, Three Parts

A client meeting can use all three groups. AI can gather the latest notes and check the customer record. It can prepare a summary and follow-up draft for review. The person in the meeting decides what to promise, handles the relationship, and approves the message.

The goal is not to remove people from the work. It is to remove the parts that drain their attention so they can spend more time on the work that needs them.