Most AI that gets built into a business never returns money or time. Not because the tools are bad — the tools are extraordinary, and you already have access to all of them. It is the order. The wrong thing gets built first.
The trap, in one sentence
Somebody complains on Monday. A customer could not get an answer fast enough. That becomes the project, and you build a chatbot.
Meanwhile a full day every week disappears into typing invoices by hand. Nobody complains about that one. Everyone quietly puts up with it, and has done for three years.
Both are annoying. Only one is bleeding you.
The loud problem gets complained about. The expensive problem gets tolerated in silence.
So the loud one gets your attention, your weekend, and your budget — and the expensive one keeps running in the background, costing exactly what it always has.
A better way to decide
Complaints are close to the worst available signal for deciding what to automate. Score candidates on three things instead.
- 1Hours × frequency. A 20-minute task done 40 times a week is worth far more than a 3-hour task done once a month. Volume is where the money hides.
- 2How little judgment it takes. If a brand-new hire could do it from a written doc, AI can do it. If they could not, AI cannot either — automating a messy process just makes it fast *and* messy.
- 3Whether anyone complains about it — inverted. Silence usually means expensive.
Then build one at a time
The most common failure is not picking the wrong workflow. It is picking fifteen. Fifteen half-built systems nobody trusts, and three months later everyone is back to doing it by hand.
A tax and advisory firm — two partners, six staff — did the boring version of this. Three systems, installed one after another rather than all at once. Repetitive work fell from half of a senior's time to 18%, roughly 11 hours a week back per person. That recovered time turned into about $210K of added advisory revenue in a year, with the same team and no new hires.
The full breakdown is in the tax and advisory firm case study, and the workflows specific to accounting practices are covered in AI for accounting firms.
How to find yours this week
You do not need a consultant for the first pass. For one week, write down every task you or your team repeat more than five times. Do not automate any of it yet.
At the end of the week the expensive problem will be obvious — and it almost certainly is not the thing anyone has been complaining about.