Boring pays the bills: where to actually start with AI in your business

Start with the most boring process in your business. The dull, repetitive, invisible work - rekeying data between systems, chasing approvals, assembling the same report every month - is where AI pays back fastest, because the steps are knowable, the volume is high, and nobody has ever had to defend it in a planning session. Boring pays the bills.

I say that phrase a lot, and it is not a joke about my job. Almost every pound of measurable value I have delivered came from work nobody would put in a press release. The exciting projects get the airtime. The dull ones get the ROI.

What counts as "boring" work?

Boring work is the stuff that happens between your systems and inside your inboxes. It rarely has a name, an owner, or a budget line. It looks like this:

• Rekeying. The same figures typed from one system into another because the two do not talk.
• Chasing. Someone spends a morning a week asking colleagues for a status, a signature or a file.
• Assembling. The monthly pack, the board report, the client update - built by hand from five sources.
• Triaging. A human reads every inbound enquiry, ticket or invoice just to decide where it should go.
• Checking. Manually comparing two lists to find the differences, because no one built the reconciliation.

None of that is glamorous. All of it is expensive. And crucially, all of it is the sort of work you can describe in a paragraph - which is exactly what makes it automatable. The steps are knowable in advance, so what you need is usually a plain workflow rather than an autonomous agent, and it is far cheaper to build than the thing you were imagining.

Why does boring work pay back fastest?

Four reasons, and they compound.

It is high volume. Small savings multiply. A task that takes four minutes and happens two hundred times a month is thirteen hours a month - more than a fortnight of someone's year, hidden in four-minute increments nobody notices.

The rules already exist. Someone in your business can tell you exactly how the task is done, because they do it every day. You are not designing a new process; you are encoding one that already works. That collapses the discovery time, which is usually where projects bleed money.

Success is unarguable. Hours saved and error rates are numbers your finance lead already recognises. You do not need a new framework to prove the value - you count the hours before and after. That matters more than people expect, because the projects that die are usually the ones that were never tied to a number anyone would defend.

The failure mode is cheap. If an automation misfiles an invoice, a human catches it in the review step and you fix the rule. If your customer-facing AI concierge says something strange, that is a different sort of conversation.

The proof points bear it out. Cutting a finance team's month-end close by 70% took a week of work on a process that had never once appeared on a roadmap. Automating CRM updates for a B2B services firm was worth £50,700 a year, and the "innovation" was that two systems finally shared their data without a person in the middle. Neither is an interesting story. Both paid for themselves several times over.

So why does the boring work never make the roadmap?

Because roadmaps are built out of things you can announce.

Planning sessions are competitive. Every item on the list has a sponsor arguing for it, and manual processes have no sponsor - they have victims. The person losing a morning a week to chasing approvals is not in the room, and if they were, they would struggle to describe the problem as anything more than "it's a bit of a faff". Nobody gets promoted for fixing a faff.

There is also a measurement problem. The cost of manual work is spread thinly across many people's days, so it never appears as a line item. It is absorbed into headcount and then re-described as "just how we do it". A £40,000-a-year process looks like nothing at all when it is sliced into four-minute pieces across eleven people. Meanwhile, a £40,000 project has to justify itself to the board in detail.

And there is the pull of the shiny. Every business is under pressure to have an AI story, and "we automated our reconciliation" does not sound like one. So the roadmap fills with the demo-friendly ideas, the boring work continues eating time in the background, and next year's planning session starts from exactly the same place.

The uncomfortable consequence is that the highest-ROI work in most businesses is the work that is structurally invisible to the process that decides what gets done.

What does starting with the exciting thing actually cost?

Anything up to a year, and the appetite for a second attempt.

The ambitious first project is slower to specify, harder to scope, and far harder to call a success. It tends to need data you have not tidied, buy-in you have not built, and a definition of "working" that shifts every time someone new sees the demo. That is a big part of why so much of this work never ships: in S&P Global Market Intelligence's survey of more than 1,000 companies, the share abandoning most of their AI initiatives jumped to 42%, up from 17% the year before, with the average organisation scrapping 46% of its proof-of-concepts before production (CIO Dive).

The real cost is not the wasted budget. It is that a failed flagship project poisons the well. The next time someone proposes automating anything, the answer is "we tried AI, it didn't work" - and the boring, tractable, £40,000-a-year problem stays exactly where it is.

Boring works the other way round. A small win that pays back in weeks is what unlocks budget and appetite for the next one. That is the entire logic behind running a narrow 60-day pilot rather than a transformation programme.

How do you find the boring win hiding in your business?

You will not find it in the roadmap, because by definition it is not on there. Do this instead - it takes an afternoon.

1. Ask a different question. Not "where could we use AI?" but "what part of your week do you resent?" The first question gets you ideas people have read about. The second gets you the actual problem.
2. Ask the doers, not the owners. Ops coordinators, finance assistants, account managers, support leads. The people whose calendars are full of the work nobody has named.
3. Multiply it out. Minutes per instance, times instances per month, times a loaded hourly cost. Do this crudely and in front of them. The number is usually larger than everyone expects, and that number is your business case.
4. Check it is actually stable. A process that changes every month, or that three people describe three different ways, is a process problem first. Run your candidate through this readiness test before committing a day to it.
5. Pick on impact over effort, then stop looking. The best first project is the dullest one with the biggest multiplication. Ship it, count the saving, and use it to fund the next.

Note what is not on that list: choosing a model, buying a platform, writing a strategy. Those come after you have something working, not before. Strategy written in advance of any evidence is just a very expensive opinion.

When is boring the wrong place to start?

I would be overselling it if I said this is always the right approach. Two honest exceptions:

If the tedious process exists because something upstream is broken - the data is scattered, the system of record is a spreadsheet, nobody agrees the rules - then automating it just makes the mess arrive faster. Fix the source first.

And if you have a genuine strategic bet where AI changes what you sell rather than what you spend, that deserves its own track and its own patience. But run it alongside the boring wins, not instead of them. The boring wins are what buy you the credibility and the budget to keep the interesting one alive.

The bottom line

The best place to start with AI is the least interesting problem you have. It is high-volume, well-understood, owned by someone who is quietly sick of it, and worth more than anyone has bothered to calculate. It's not even a bullet-point in a planning session agenda, which is precisely why it is still costing you money.

Boring pays the bills. Find the dullest expensive thing in your business, automate it properly, and let the saving fund whatever you do next.

If you want a second opinion on which of your processes that is, book a thirty-minute discovery call - I will tell you plainly if the honest answer is "fix the process first". See process automation for how I approach the work, and the case studies for what a boring first candidate has been worth.