A dashboard shows you a margin rate. It doesn’t tell you whether that rate is good.
That’s the whole difference, and it’s what we built: the VIL — Vertical Intelligence Layer, the module that reads your data through a knowledge base we maintain for your industry — its thresholds, its norms, its seasonality.
Why “vertical”
Because the engine doesn’t reason in the abstract.
A 14% margin can be excellent on a fast-moving product line and alarming on another, in the same warehouse, the same week. A 90-day payment term is unremarkable in some sectors and a red flag in others. These benchmarks aren’t inferred from a spreadsheet — they’re written down, sourced and dated, industry by industry.
That written knowledge is what lets you say whether a number is good, instead of just displaying it.
What the engine looks at
Not what a dashboard shows. Crosses that nobody has time to run by hand, on your procurement as much as your sales:
- which product lines each customer has quietly stopped ordering — a far earlier signal than a drop in revenue;
- who hasn’t bought anything in thirty days despite carrying real weight in your book of business;
- which accounts combine high volume with thin margin — the ones that make turnover without making money;
- which categories are losing momentum month over month;
- how much of your business rides on too few customers.
None of these fits in a single metric. Each requires crossing several dimensions at once — precisely the work a human does badly, not for lack of skill, but because it takes too long to redo every month.
What you get
Five findings, at most. Never a list to sort through.
Each one is typed — alert, opportunity, margin risk, strength, trend — and carries an urgency level. And each one names the account or product line involved, with the amount at stake. Not “margin is slipping,” but where, with whom, and how much.
A short text, readable in a few minutes, where every claim is checkable: the file it came from, the formula that produced it, the business rule that makes it significant — and what was excluded from the calculation.
That last part is the one everyone else skips, and it’s the one that matters. When someone in a meeting asks where a number came from, the answer is already in the text. That’s what separates a defensible figure from one you either accept or reject wholesale.
Written rules that bound the model
The engine relies on artificial intelligence to connect information that doesn’t live in the same place and turn it into a coherent view. That’s what it’s good at.
But it operates inside written rules that bound what the model is allowed to assert: thresholds, exclusions, forbidden phrasings, calculation scope. A model left unconstrained always produces something plausible-sounding — that’s exactly the danger.
That’s the difference between an analysis and a well-presented opinion.
Two things this industry rarely says out loud
An analysis carries its own date
You know when it was produced. And it flags itself when it has aged, or when enough new data has come in since it was generated.
A stale finding that doesn’t say it’s stale is worth less than no finding at all. It’s a simple rule, and almost nobody applies it.
The VIL works without dashboards
If you’re already equipped — Power BI, Tableau, Qlik, anything else — you have nothing to replace and nothing to migrate.
The VIL connects to your data and hands back the reading, without asking you to change tools. Dashboards stay exactly where they are, in the hands of the people who know how to use them. The reading reaches the people who decide.
Your data stays yours
Strict isolation between customers, hosted in France. Every analysis is traced: you know when it was produced, from what, and what triggered it.
Frequently asked questions
How is the VIL different from a general-purpose AI assistant? An assistant comments on whatever you hand it. It doesn’t know what the file leaves out, doesn’t know your industry’s benchmarks, and keeps no record of what it claimed last month. The VIL works on prepared data, checks results against written rules, and ties every claim back to its source.
Do I have to give up my dashboards? No. The VIL works independently. It can complement existing dashboards or be used on its own.
What data do I need to provide? Your exports, exactly as they come out of your management software. No pre-cleaning, no renamed columns — recognizing the format is part of the job.
How often is the analysis produced? Every time meaningful new data arrives, and on demand. Every analysis is dated and flags its own expiry.
See procurement reporting in detail · The fastest way to know what your data can actually tell you is still to look at your real files rather than a demo: write to us.