My team spends hours on the same repetitive admin, where would AI actually help?

What do those three actually mean?

Repetitive means the same shape every time. Not identical content, the same shape. A hundred emails asking different things about the same five topics is repetitive. A hundred genuinely different problems is not.

Rules-based means somebody could write down how the decision gets made, even if nobody has. If your best person cannot explain why they do what they do, that is judgement, and judgement does not automate well. It also means the process is worth documenting for its own sake.

Resource-intensive means it costs real time. This is where most enthusiasm dies honestly. A task that takes four minutes a week is annoying, not expensive. Automating annoyance feels good and pays nothing.

Where does this usually land in practice?

The work that happens around the work, rather than the work itself.

In most operations the pattern is the same. Somebody reads an incoming thing and decides where it goes. Somebody copies information from one system into another. Somebody writes a summary of what just happened. Somebody chases a document that has not arrived. Somebody assembles a pack from six places.

None of that is the skilled part of anyone's job, and all of it is repetitive, rules-based and expensive. That is where I would look first, in almost any business.

The work I would leave alone, at least initially, is anything where a human is applying experience to an ambiguous situation. That is the part people are actually paid for.

What should we do before automating anything?

Measure it, and map it as it really happens.

Measuring is unglamorous and it is the difference between a project you can justify and one you cannot. How many times a week, how long each time, how many people. You need that number before, or you will never prove anything after.

Mapping matters because the documented process and the real one are rarely the same. The real one has workarounds in it, and the workarounds are usually the most informative thing in the building. They exist because something does not work, and automating around them cements a problem instead of fixing it.

This is why I fix the process first and apply the technology second. It is not a slogan, it is that automating a broken process gives you a faster broken process.

Will the team actually use it?

Only if they helped build it, in my experience.

The failure mode is consistent and it is not technical. Something gets built, it gets announced, and people quietly carry on as before because the new thing does not fit how the work really flows. A smaller tool that people use beats a better tool they work around, every time.

The practical version is that you need a few hours from the people who actually do the task, not just from whoever signed the contract. They will tell you things about the process that nobody else knows.

When should you not bother?

When the volume is not there, or when the process is about to change anyway.

If three people do the task twice a week, the payback will not cover the build. And if you are mid-way through a system migration or a restructure, automating the current process is money spent on something that will not exist in six months. Wait.

There is also a version of this where the honest answer is to hire someone, or to stop doing the task at all. Both come up more often than the industry likes to admit.

Where to look next

The 3 Step AI Audit on our services page maps a process properly and tells you what would actually pay back, at £1,250 per day. Before that, a free discovery call at our booking page is forty-five minutes on one process of your choosing, and there is no pitch at the end.

For frameworks, the 5 Pillars of AI covers where AI fits across an operation and our guide to choosing AI tools sets out the evaluation criteria.

Frequently Asked Questions

Your next step

Take the free two-minute Quick Check for an AI readiness score, see what the £995 AI Visibility Audit covers, or book a free 45-minute discovery call. Definitions for the terms used here are in the AI glossary and the Claude glossary.