What to build next
Spend where the hours are.
Most automation and AI money is spent on the loudest problem, not the biggest one. The loudest is whatever someone complained about this week. The biggest is usually quiet: three minutes a pallet, forty pallets a day, every working day.
01 — The Roadmap
8 ideas, one budget.
Every idea the people of the example company have suggested so far, placed by the hours a year it gives back and the developer-days it takes. Move the budget and see what fits.
An example company. The figures are illustrative.
860hours a year back
Four ideas fit in 20 developer-days. Together they give back 860 hours a year. That is about half a person’s year.
The rule Best hours per developer-day first. An idea that does not fit the days left is passed over for a smaller one that does.
In order of hours per developer-day
02 — How the ranking is made
Three plain rules, and no enthusiasm.
Hours a year come from the steps an idea touches.
Nobody types in a hopeful number. An idea is tied to the steps on the map it would change, and their times, counts and days give its hours. Scan pallets at the dock touches two steps of Booking in, about 610 hours a year between them, and would hand about 430 of those to tools.
Pain counts what hours miss.
Some work is short and still hurts: one missing certificate holds a truck at the border, one person alone knows the credit limits. The people who do the work say how much it hurts. Pain makes the circle bigger; it never adds hours that are not there.
Cost and payback are shown to those who decide money.
Effort is counted in developer-days, which everyone can see and argue with. What a day costs, and how soon an idea pays for itself, is shown to the people who hold the budget, and to nobody else.
A guess looks like a guess.
When nobody has said how long an idea would take, its effort is a guess, and the chart draws it dashed and hollow. It stays that way until someone who would build it gives one number. Then it is drawn solid, with their name on the figure. Nobody has to pretend to be sure.
10 d?
A guess
Ten days, if we reuse the dock scanners.
The developer who would build it
10 d
Said, by name
03 — Ask the Roadmap
Ask it the question you would ask a person.
In plain words, from Claude, any day of the year. The answer is worked out from the map, with the same rule as the budget above.
What if we only had EUR 10k this year?
The Roadmap
At EUR 550 a developer-day, EUR 10,000 buys 18 days. Three ideas fit: Scan pallets at the dock, Carrier quotes in one message and Certificates checklist.
| Idea | Effort | Cost | Hours a year back |
|---|---|---|---|
| Scan pallets at the dock | 6 d | EUR 3,300 | 430 h |
| Carrier quotes in one message | 9 d | EUR 4,950 | 300 h |
| Certificates checklist | 3 d | EUR 1,650 | 90 h |
| Together | 18 d | EUR 9,900 | 820 h |
At EUR 22 an hour of people’s time, that is about EUR 18,000 of work a year, so it pays back in about 7 months. None of the three efforts is a guess. Next in line would be Invoice straight from the CMR: 12 more developer-days for 260 h a year.
Money is shown only to the people who decide it. Everyone else sees hours and developer-days.
The answer changes nothing by itself. People decide.
04 — Proof
Scan pallets at the dock, measured again.
After an idea ships, the same steps are measured again. This is still the example company: the figures show how proof is drawn, not a customer’s result.
Today 19 March
Booking in
- Receive the truck
- Check the pallet
- Type pallets into Excel
Shipped and measured 18 September
Booking in
- Receive the truck
- Check the pallet
- Type pallets into Excel
The same three steps and the same 610 hours a year. About 430 of them have moved to tools: the green share grew from 12% to 82%. The clerk still receives every truck.
The work stays. Who does it changes.
Your own list, a month from one call.
One call becomes the first draft of your map. A few weeks of your people correcting it, and the ideas are ranked by the hours a year they give back.