In plain English: List what your team does every week, score each task by hours × rule clarity × error cost, and automate the top three. The boring ones win.
Ask your team what they'd love to automate and you'll get a wish list worthy of a tech conference: an AI sales assistant, a customer portal, a dashboard for absolutely everything.
Now ask what they actually do every week. Suddenly the list gets a lot less glamorous: copying order data into the invoicing tool. Chasing approvals. Rebuilding the same weekly report from the same three exports.
That second list is boring. It's also where the money is hiding. Here's how to dig it out in about two hours — no consultants, no software, just a room, a whiteboard and some honesty.
Step 1: Collect the candidates (30 minutes)
Get the people who do the work in a room — or a call — and ask one question:
"What do you do more than once a week that follows the same steps every time?"
Write every answer down. No judging, no "we'll never automate that". You're looking for things like:
- Re-typing data from one system into another
- Creating standard documents (quotes, contracts, onboarding packs) from a template
- Sending reminders and follow-ups
- Checking whether something has happened — a payment, a signature, a delivery
- Compiling a report from several sources
For a company of 20–50 people, twenty to forty items is normal. If you get fewer than ten, somebody isn't telling you the truth.
Step 2: Score each one (45 minutes)
For every candidate, estimate four things. Rough numbers are fine — you're ranking, not budgeting.
- Frequency — how many times per month does it happen?
- Minutes per occurrence — including the context switching around it (that part is always longer than people admit).
- Rule clarity (1–3) — 3 if the steps are always the same, 1 if it needs judgement every time.
- Error cost (1–3) — 3 if a mistake costs real money or reputation (invoices, compliance), 1 if it's merely annoying.
Then do the maths:
Hours per month = frequency × minutes ÷ 60
Priority = hours per month × rule clarity × error cost
The formula deliberately punishes tasks that need judgement. They're not impossible to automate — with today's AI tools, often they're not even hard — but they're riskier and more expensive to do well. Start where the rules are clear and the stakes are high.
// quick one
Want us to do this for you? We only take a few new clients at a time.
See if you’re a fitStep 3: Watch the exciting ideas sink (the fun part)
Here's an illustrative shortlist for a 30-person services company — the kind of list this exercise typically produces:
| Task | Per month | Minutes | Hours / month | Rules | Error cost | Priority |
|---|---|---|---|---|---|---|
| Copy won deals from CRM into invoicing | 60 | 12 | 12 | 3 | 3 | 108 |
| Build weekly utilisation report | 4 | 180 | 12 | 3 | 1 | 36 |
| Send contract reminders | 40 | 5 | 3.3 | 3 | 2 | 20 |
| Onboarding pack for new clients | 8 | 45 | 6 | 2 | 1 | 12 |
| Qualify inbound leads | 50 | 10 | 8.3 | 1 | 1 | 8 |
The least exciting task on the list — syncing won deals into invoicing — wins by a landslide. It's frequent, rule-based, and invoicing mistakes are expensive. Lead qualification, the team's favourite, finishes dead last for now: it needs judgement, and the occasional misjudgement costs little.
That's not a failure of ambition. That's the order that pays.
Step 4: Sanity-check the top three (30 minutes)
Before anyone writes a line of code, ask three questions about each winner:
- Is the process stable? Automating a process that's about to change is money set on fire. Fix the process first, then automate it.
- Where does the data live? If the systems involved have APIs (most modern tools do), integration is straightforward. If the data lives in a PDF attachment or in Brigitte's head, add an extraction step to the estimate.
- What happens when it breaks? Every automation fails eventually — an API changes, a field comes through empty. Decide now who gets notified and how the task gets done by hand in the meantime.
The payoff: roughly €16,000 a year from three boring tasks
In the example above, the top three items together save around 27 hours a month. At a loaded cost of €50 per hour, that's roughly €16,000 a year — before counting the invoicing errors that simply stop happening. A typical integration of this kind costs a fraction of that and pays back within months.
The bigger win is harder to put in a spreadsheet: your team gets its time back for work that actually needs a human. And once three automations have visibly paid off, the ambitious AI project gets approved in one meeting instead of five.
The four ways to ruin it
- Automating a mess. If a process has five exceptions for every rule, simplify it first. Automating chaos just gives you faster chaos.
- No owner. Every automation needs a person who notices when it stops working.
- Silent failures. An automation that fails quietly is worse than none at all — because everyone assumes the work got done.
- Tool first. Buying an automation platform before you know what to automate is how companies end up paying for features instead of outcomes.
Want the shortcut? The free Digital Readiness Scorecard shows you in three minutes where your quickest wins are most likely hiding. Or plug your own numbers into the copy-paste calculator on our homepage — fair warning, it can sting.