Where do you start with AI in a business? With a single task, chosen because it is repetitive, high-volume, and low-risk if the first output is imperfect. Not with a tool, not with a strategy, not with a steering committee. The method has three steps: locate where information breaks down in your organisation, quantify what that breakdown actually costs, then deploy one automation and measure it for a month. Our audits have revealed between €109,000 and €527,000 of hidden costs per year depending on the company, with a return on investment observed at around two months.
The real problem is not AI, it is information
When an executive calls us, the question almost always arrives as: "which AI tool should we use?". That is the wrong entry point, and it explains a good share of the projects that never land.
In a business, AI creates no value on its own. It creates value when it repairs a break in the flow of information: data that exists somewhere but nobody can find, a document rebuilt by hand because it lives nowhere, information that travels through one person's memory. That is where the time goes, and the money with it.
We formalise this diagnostic under the name Information Trinity. Every friction in an organisation reduces to one of three questions:
- The information — does it exist? Is it accurate, current, complete?
- The medium — where does it live? A business system, a spreadsheet, an inbox, someone's head?
- The flow — how does it move from person to person, step to step? Automatically, or through a manual chase?
A friction where the information does not exist cannot be fixed with AI: it is fixed by creating the data. A friction where the information exists but does not flow, on the other hand, is almost always automatable. That simple sort prevents investing in the wrong place — and it is the first thing a serious audit does.
The three steps, in order
Step 1 — Locate, by asking the people who do the work
The mapping does not happen in a management meeting. It happens by interviewing the people who execute the processes, one by one, about what wastes their time. Depending on company size, that means two to fifteen interviews.
What we look for: workarounds. The parallel spreadsheet someone maintains because the official system does not do the job. The double entry everyone considers normal. The document rebuilt every month. Those workarounds are the exact map of what is broken — and nobody mentions them spontaneously, because they have become routine.
Step 2 — Quantify, so you can arbitrate
An unquantified friction cannot be compared. The calculation is simple and must be done in front of the client: time lost × loaded hourly rate × annual recurrence, for every documented friction. The total is always a surprise.
At a 100-employee civil engineering group, that calculation produced €109,000 per year. At a 26-person notary practice, €527,000 per year — notaries spending their days sorting, re-reading and re-typing. Neither executive would have named those figures beforehand; both confirmed they matched what they felt operationally.
Step 3 — Deploy one thing, and measure it
This is where most companies go wrong, launching three or four parallel initiatives because the audit surfaced fifteen. A mid-sized company has no dedicated innovation team: every initiative consumes management time and team attention.
One initiative, carried through to real daily use, produces more value than five pilots abandoned within three months. And it produces something more valuable still: internal proof that it works. That proof is what unlocks the next ones.
Which first task to pick
A good first initiative has four characteristics, and one exists in nearly every company.
It is repetitive. If the task changes shape every time, automation costs more than it returns.
It has volume. A painful but monthly task pays nothing back. Look for the daily and the weekly.
It tolerates imperfection. A first version produces a draft a human validates. Starting with a process where errors are immediately costly means never reaching production.
It has an owner. An automation without an internal owner dies within six weeks. You need someone who uses it daily and has an interest in it working.
In practice, the first initiatives that recur most: producing recurring documents (quotes, reports, memos, submissions), handling inbound requests, reconciling data between two systems, and assembling files from scattered pieces.
What it costs, and when it pays back
- The AI audit — the quantified mapping described above — starts at €2,000 excl. VAT on a tight scope, rising with the number of interviews and the size of the organisation.
- Monthly support sits between €499 and €799 excl. VAT per month: a partner who comes back every month, deploys, trains and measures. No lock-in.
- Custom development starts at €3,000 excl. VAT, priced after scoping, for a tool you own.
The comparison is never "is this expensive", but "what does inaction cost". When an audit has quantified €109,000 of annual losses, monthly support at €499 pays for itself on the first friction addressed. Return on investment observed with our supported clients is around two months.
The four most common mistakes
Buying the tool before having the diagnosis. The most common way to pay for a subscription nobody opens.
Handing the topic to whoever is most available. An AI project must be owned by someone who knows the process, not by whoever has spare time.
Aiming for the spectacular. The gains come from boring tasks. Impressive projects make good presentations and poor returns.
Not measuring. Without a figure before and after, you will never know whether it works — and you will be unable to stop or accelerate knowingly.
And if AI is not the answer
It happens, and it is a perfectly valid audit outcome. When the information does not exist, when volumes are too low, or when the process needs simplifying first, the right recommendation is not to do AI — or not yet. A partner who never tells you no is not advising you, they are selling to you.
What you can do this week
Take a sheet of paper and write down the three tasks your teams complain about most. For each, ask the Trinity's three questions: does the information exist, where does it live, and how does it flow? The one where information already exists but travels by hand is your first initiative. In an hour, you have just done the short version of the audit.