When your teams don't share the same version of the truth

One Monday morning, Inès, head of procurement at Granderive, sends the same email to four different people at her supplier, Néovia. One ordinary question: "What's your guaranteed deployment time, and the penalty if you miss it?"

She doesn't know it yet, but that question is about to cost Néovia its biggest contract — and go around LinkedIn.

Four people, four truths

The answers come back over two days.

• Karim, sales, replies from the pitch deck: "Deployment in 5 days, 10% penalty if we overrun."
• Sonia, delivery lead, replies from the production runbook: "Realistically 3 weeks, and there's no contractual penalty."
• Legal sends the contract template: "Lead time 10 business days, penalty capped at 5%."
• The help center shows an article: "Standard deployment in 48 hours."

Four answers. The same supplier. The same week.

Inès doesn't have to invent anything. She copies the four emails one under another, forwards them to her committee, and adds a single line:

"If a company doesn't know what it sells, how can it deliver it to us?"

The renewal — €2M over three years — will never be signed.

The post that set everything off

That same evening, Inès posts an anonymised screenshot of the four answers, side by side, on LinkedIn. Caption: "Our suppliers don't work with the same version of the truth."

11,000 reactions in two days. In the comments, other buyers: "Happens to me every week." "We end up asking three people the same thing on purpose." Néovia's own employees recognise their company. The post becomes a case study in procurement circles.

The worst part isn't the lost contract. It's the public proof that Néovia didn't know its own truth.

The autopsy: nobody had lied

Néovia launches an internal review, sure it will find a culprit. It finds none.

Karim was reading the official sales deck. Sonia, the production runbook. Legal, the approved contract template. Support, a help-center article published by marketing. Four real documents. All approved at some point. None retired. All contradicting each other.

The problem wasn't that someone lied. It's that no one could see the documents contradicted each other.

Néovia didn't have a truth problem. It had a version problem.

Why it happens to almost everyone

A company's knowledge never lives in one place. It scatters — Drive, Notion, SharePoint, email, and experts' heads. It duplicates. And above all, it ages silently: you write the new version and forget to retire the old one. Both coexist, equally "official," until someone lines them up. (Five signs your knowledge base isn't ready.)

As long as that someone is a colleague, you recover. The day it's a customer — or worse, your AI — it's too late.

Because that's where the story turns urgent. Your teams now plug ChatGPT, Claude or an in-house assistant into those same documents. And an AI doesn't arbitrate between two contradictory versions: it answers, confidently, with whichever one it found. Néovia's four truths, served at machine speed, to every customer, with no human re-reading them. That's essentially why your AI hallucinates: it's only as reliable as the knowledge you feed it.

The one thing that would have saved Néovia

Not "write better documents." Néovia had four, all well written.

What was missing: a single version of the truth — a company memory where every team, and every AI, draws the same sourced, traceable answer. And, above all, something that surfaces contradictions before the customer does. The day three documents claim three different lead times, you want to know that day, not on renewal day.

That's exactly what Ragnight's Knowledge Pulse does: it scans your sources, surfaces contradictions, stale clauses and ownerless claims — and gives your AIs knowledge they can answer from without making things up. (How to audit your base before plugging AI into it.)

The real question

Néovia's story is fictional. The moment a customer lines up your answers side by side is not.

Only one question matters: how many versions of the truth coexist, right now, inside your company? You won't know until someone — or something — has lined them up before your customers do.

Book a demo: we measure, on your real data, how many versions of the truth your teams — and your AIs — keep alive today.