AI or automation? Three questions to decide what your task actually needs

Use automation when a task follows the same rules every time, and AI when the work means reading or writing language that changes from case to case. Before either, check two things: whether the task needs doing at all, and whether the process around it works. Those three questions sort almost any repeat task into one of four fixes.

Most businesses that ask us about AI don't need much of it. They need the repeat work off their team's plate, and most of that work turns out to need automation, a better process, or to stop. Below is the same three-question check we run on every task in our free AI audit, so you can run it on your own business.

What's the difference between AI and automation?

Automation follows fixed rules. When an order arrives, it copies the details into your system, drafts the invoice and tells the warehouse, the same way every time. It doesn't think, and that's the point: it's predictable, cheap to run and easy to check. Most of what we build is business process automation.

AI handles work that means reading or writing language: an email worded differently every time, a contract with the key terms buried on page six, a request that doesn't fit a form. It's good at drafting and sorting. It isn't perfect, so anything that matters should be checked by a person before it reaches a customer or touches money. That's how we build AI assistants.

A simple way to tell them apart: if you could write the rules on an index card, it's automation. If you'd need to teach a new hire with examples, it might be AI.

Question 1: Does it need doing at all?

Start here, because the cheapest fix is no work at all. Every business has tasks that outlived their reason: the weekly report nobody opens, the spreadsheet that copies another spreadsheet, the approval step added after a problem years ago.

Automating these is a trap. Automated, the weekly report nobody opens goes unread right on schedule, and now it's harder to notice and harder to stop.

How to check: ask who uses the output and what they'd do without it. If nobody can say, stop doing it for a month and see who asks.

If the answer is no: stop doing it. The best fix is no work at all.

Question 2: Does the process work today?

If the task needs doing, look at the process around it. Automation copies a process exactly, including its problems. A broken process run faster just produces mistakes faster.

Common signs the process needs fixing first:

• Ask three people for the steps and you get three answers.

• Two spreadsheets disagree because nobody decided which one is right.

• The task only works when one particular person does it.

Month-end reconciliation is the classic example. When three spreadsheets disagree, the fix isn't software. It's deciding which one is the source of truth. Once that's settled, much of the matching becomes easy to automate.

If the answer is no: fix the process first. Sort out how it's done before anything gets built.

Question 3: Are the rules the same every time?

Now the task is worth doing and the process works. The last question decides between automation and AI.

If the steps are the same every time, automate it. Chasing overdue invoices is a good example: same trigger, same reminder, same schedule. Copying orders into your ERP is another: structured data, fixed rules, and a simple connector handles it.

If every case is different, use AI with a person checking. Answering “where's my order?” emails is a good example. Every email is worded differently, but the answer is already in your order data. AI drafts the reply from that data, and a person approves it.

The four outcomes at a glance

Why AI isn't always the answer

There's a lot of pressure right now to “do something with AI”. It's worth resisting where it doesn't fit, for three reasons.

• It usually costs more to run. For rule-based work, automation is cheaper and more predictable.

• It needs checking. AI drafts well, but it can be wrong, so someone has to review what matters. That review is worth it for messy work, and wasted on work a fixed rule could do.

• It can hide a broken process. A clever tool on top of a messy process makes the mess harder to see.

In our experience, most repeat work in a growing business needs automation, a better process, or to stop. AI earns its place in the rest: the reading and writing that used to need a person's full attention. That honest split is the whole idea behind our AI consulting.

How to run this check on your own business

1. List the repeat tasks your team does every week. Ask the people who do them, not just the managers.

2. Run each task through the three questions, in order. Stop at the first “no”.

3. Rank what's left by how often it happens and how many hours or errors it costs.

4. Start with one task that happens often and follows clear rules. An early win builds trust for the next one.

Frequently asked questions