Technology Consulting

Five problems, one budget. Which one first?

In this article

Nori.mx7 min read

Five frontsWarehouse sheetAnnex 24EDI deadlineNew machinesAIThree criteria1Cost today?2How many touch it?3Can it be undone?in this orderONE PAGEWhat goes firstToday's costOwner, by nameIn 90 daysWhen to stop

The list usually looks about like this. The warehouse runs on a spreadsheet. An Annex 24 review is coming and nobody will sign that the balance reconciles. A customer sent an email asking for EDI, with a date. The new machines report to a screen nobody looks at. And in the management meeting someone asked what we are doing about artificial intelligence.

Five fronts, one budget, and a team that is already busy. The question is not which of the five matters, because all five do. It is which one you touch first.

The mistake is not choosing wrong, it is choosing without a rule

Almost nobody fails by picking the wrong item off a list. They fail because they never wrote down the rule they were picking with, and then the front that wins is whichever one came up in the most recent meeting.

That has a concrete consequence. When the project is halfway through and starts costing more than planned, there is no way to defend it. Nobody remembers why it was chosen, so it is dropped and another one starts. Three of those in a row and the organization learns that systems projects do not finish.

A written rule does not guarantee you are right. It guarantees that if you were wrong, you will know, and that the next decision starts from something.

Separate what is process from what is system

Before ordering anything, split the list in two. Some fronts a system solves; on others the system only makes the mess move faster.

The test is simple: if the problem goes away when two people agree, it is not a software problem. If two departments report different numbers because each counts the same thing its own way, no system will reconcile them. Deciding which of the two is authoritative is a human call, and it has to be made before anything is bought.

We covered this in the ERP that does not reach the floor: the one piece of data, one owner rule is not technical. It is organizational, and it is free. It is also the first thing everyone skips, because it cannot be quoted.

In practice, out of a list of five fronts there are usually one or two that a meeting and a written agreement will fix. Take them off the investment list and do them this week.

Three criteria that actually order the list

For what is left, three questions. In this order.

What does the error cost today, measured? Not estimated. Measured. Customer chargebacks are priced per shipment in the supplier manual. An Annex 24 discrepancy has a calculable tax cost. A line stop for a missing part has an hourly cost and somebody knows it. If nobody can state the number for a given front, that is information in itself: it means it is not being measured, and what is not measured cannot be defended in the budget meeting.

How many people touch it? A fix that changes the work of two people lands in weeks. One that changes the work of forty, across three shifts, is not a software project: it is a change project, and it takes three to five times longer. Both can be worth doing. What is not worth doing is confusing them while planning.

Can it be undone? If in eight weeks it turns out to have been a bad idea, can you go back without losing data or stopping the operation? A pilot on one line can be undone. Replacing the system that invoices cannot. All else being equal, reversible first: you learn faster and you risk less.

When all three point at the same front, the decision is made. When they do not, the first one wins.

There is an order you do not get to choose

Some fronts are prerequisites for others, and that order does not negotiate, however loudly the urgency argues.

The most common case on the border is this one: a customer asks for EDI and your warehouse does not know what is in each box before it is sealed. The shipping notice describes packaging structure — pallet, box, piece — so a valid one cannot be issued without that information. Buying the EDI first gains you nothing; it only makes the discrepancy travel faster.

The same goes for machines: connecting them before knowing what decision the data will drive produces a nice screen nobody looks at, which is exactly what has to be ready first. And it goes for artificial intelligence: an agent works on your data, so it inherits its quality. If the inventory is wrong, the agent is wrong faster and with more confidence.

The short rule: whatever produces the data comes before whatever consumes it.

The trap of starting with what shows

Between a dashboard and a part-number table, the dashboard wins every time. It can be shown in the meeting, it is understood in ten seconds, and it feels like progress.

But the table that maps your part number to the customer’s is the one that decides whether EDI works, whether inventory reconciles and whether Annex 24 can be defended. It cannot be presented, it impresses nobody, and it usually lives in the head of one person who has been there nine years.

This is not an argument against dashboards. It is an argument about order: the boring thing first, because it is what holds up the visible one. A dashboard built on data you do not trust is worse than no dashboard, because now the bad decisions come with a chart.

Signs that it is not time to buy anything yet

Sometimes the right answer is to wait. Four signs:

  1. Nobody owns the data. If you cannot name the person who answers for the inventory balance, buying a system only changes where the wrong balance lives.
  2. The process changes every week. Automating something still being invented freezes a provisional version and makes it permanent.
  3. The problem is in the contract. Sometimes what hurts is not the operation but what was signed: an agreement that treats a forecast as a firm order is not fixed with software.
  4. You cannot state today’s number. With no baseline there is no way to know whether the project worked. And without that, the next budget is negotiated blind.

“Not yet” is a valid conclusion. It is cheaper to find out now than halfway through a rollout.

A decision that fits on one page

Once you have the order, write it down. One page, five lines:

  • Which front gets attacked first and which one waits.
  • What that error costs today, with the number and where it came from.
  • Who owns it: one person, by name, not a department.
  • What will look different in ninety days, measured with that same number.
  • What would have to happen for us to stop.

The last line is the one nobody writes and the one worth the most. A project with no exit condition never gets cancelled: it just drags.

That page is the whole deliverable. If someone hands you forty slides instead, what you bought was a presentation, not a roadmap.

There is also a front that rarely makes the list because nobody buys it: the old system you already have and nobody documented. If that is one of your five, the decision is not what it costs but whether to rescue it or replace it, and that is answered in an afternoon without quoting anything.

Two fronts that are nearly always on the list and nearly never named as such: the delivery date promised without being calculated and the customer base living inside one person’s head. Both can be diagnosed in an afternoon, and both are fixed sooner by a written rule than by a system.

And if one of the five fronts is “add artificial intelligence,” put a number on it before you put a priority on it: what an AI agent actually costs almost always comes out cheaper than feared and dearer than promised, and the same arithmetic avoids both surprises.

And if you want a first ordering without booking anything, Chira asks exactly these questions and returns the document: what to fix first, what is process and where a system belongs. It takes about ten minutes, and sometimes the answer is that you do not need to buy anything yet, which is also worth knowing.

Related service

Technology Consulting

We do not only build software. We help you decide what to build first.

See the service

Articles on this service

New bank accountFrom: billing@suppliermx.comNew account···· ···· 4821bank-statement.pdf"It has to go out today"Before changing it1Call the number already on file2Someone else approves the change3Small first payment, or wait3,046million USDBEC losses · FBI 2025

Technology Consulting

The email that changes a supplier's account

It comes from the usual address and says the supplier switched banks. How the fraud that costs businesses the most works, and the control that stops it.

5 min read

tuempresa.comRegistrantA former employeeRenewal notices go toPersonal emailAuto-renewalExpired cardRuns on itEmailWebsiteAccount recoveryCertificatesIf it lapses30daysto redeem it

Technology Consulting

Who actually owns your domain?

Your website, your email and the recovery of almost every account depend on your domain. How to find out who owns it, what happens if it expires and how to fix it.

5 min read

Technology auditThe document is yours1Inventory2Risks, ranked3What to do first4Process or software5Without buying anythingRisks, ranked1dated2no date3no date7questions it covershalf are not technical

Technology Consulting

What a technology audit should hand you

A technology audit is an inventory with judgment, not a hunt for culprits. What it reviews, what it must leave you with and how to tell if it is a sales pitch.

6 min read