Every leadership team we meet has a list of AI ideas. Most of those lists are long, exciting and almost entirely unranked. The result is predictable: the first project chosen is the most interesting one, not the most valuable one — and six months later the company has a clever prototype and no change to the P&L.

After mapping more than 300 business processes across logistics, healthcare, legal and retail clients, we’ve found that the best first projects look remarkably similar. They are boring, frequent and expensive in aggregate. Here are the five we recommend looking at first, and how to decide which one is yours.

Why boring wins

The processes that pay back fastest share three traits. They happen hundreds of times a week, so small savings compound. They follow a recognisable pattern, so a system can learn it. And they sit between tools — email to CRM, form to spreadsheet, ticket to calendar — which is where people quietly lose hours copying and checking.

Notice what’s missing from that list: strategic importance. Your pricing strategy matters enormously, but it’s set a few times a year by people who should keep thinking hard about it. Your quote intake happens 400 times a week and nobody enjoys it. Start there.

The five processes

Across industries, these five show up on almost every opportunity scorecard we build — and they usually rank in the top quartile for return.

1. Inbound request intake. Quote requests, enquiries and referrals arrive by email in a dozen formats. An intake agent can read them, extract the details and draft a first response for a person to approve.

2. Scheduling and reminders. Confirming, rescheduling and filling cancellations is high-volume, rule-heavy work. Automating it cuts no-shows as well as admin time.

3. Document checks. Insurance verification, clause review and invoice matching all compare a document against a known standard — ideal for a system that flags exceptions.

4. Status updates. “Where is my order?” and “Has this been approved?” are questions your systems can already answer. A connected assistant answers them instantly.

5. Weekly reporting. If someone spends Monday morning exporting and formatting, that time can almost always be returned — with a better report.

How to choose your first one

To rank them for your business, score each on three questions: how many hours does it consume per month, how often does the pattern hold, and what does a mistake cost? A process with high volume, a stable pattern and a cheap, catchable mistake is a near-perfect first candidate.

The right first project is the one your team is most relieved to stop doing — not the one that looks best in a board deck.

Talk to the people doing the work before you decide. Leadership tends to underestimate how long the dull tasks take, and the people closest to the process will tell you exactly where the exceptions hide.

What a good first build looks like

A good first build is narrow. It does one job, it keeps a person in the loop for anything that matters, and it reports a single number every week. For Northwind Freight that number was time to quote, which fell from 37 minutes to 14 within the first quarter.

Rule of thumb: If you can’t name the one metric a system will move, it isn’t ready to build. Agree the number — and how you’ll measure it — before the first sprint starts.

Resist the urge to expand scope during the build. Every “while we’re at it” adds edge cases, and edge cases are what turn a three-week project into a three-month one. Ship the narrow version, measure it, then widen it.

Where to go next

Once your first system is live and measured, the second is far easier to justify — you have a real number, a team that trusts the approach and integrations you can reuse. Most of our clients add their second build within eight weeks of the first.

If you’d like a second opinion on your list, our two-week diagnostic scores every idea by return, risk and effort. Or simply start with the five above, and pick the one your team complains about most.