Home › The business questions

Readiness

Is our information ready?

Updated November 2026 · About 5 minutes to read · Watch the 3-minute video

The short answer

Probably not all of it, and that's normal. You don't need every document ready, only the ones your first project uses.

For those, you need one current version of each, an owner who keeps it current, and a record of who may see it.

This guide in under three minutes. Captions are on; the CC button turns them off. Watch on YouTube.

Think of a library that never throws anything out

Ask for the travel policy and the librarian finds five: the 2019 one, the 2021 one, two drafts, and the current version.

The AI reads all five, and may quote the wrong one, with confidence.

Where the picture stops working: a good librarian checks the date. The AI only knows which version is current if someone has marked it.

Data quality worries leaders more than security

In a 2026 survey of 300 senior IT and data executives, these were the top concerns about using company data for AI.

33%

Getting data in real time
Live numbers and constant changes.
50%

Security and privacy
Who may see what, and where data goes.
50%

Data spread across systems
The same thing in five places, slightly different in each.
59%

Data quality and consistency
Old, duplicated or contradictory information.

Source: K2view, The 2026 State of Enterprise Data Readiness for GenAI, a survey of 300 senior IT and data executives at US and UK companies with 1,000+ employees (K2view, March 2026). K2view sells data software.

Six common problems, and the fix for each

Old and duplicate versions

Five travel policies, and nobody sure which is current.

Fix: keep one current version; archive the rest where the AI can't read them.

Documents nobody owns

Written by someone who left, never updated since.

Fix: name an owner for each document the project uses, with a review date.

Scanned pages and pictures

Old contracts, faxes, diagrams and tables saved as images.

Fix: budget time to convert and check them. Tables are the hardest.

Answers in people's heads

"Ask Maria, she knows how we handle that."

Fix: write it down first. The AI can only answer from what's written.

Spread across many systems

Shared drives, SharePoint, email, the intranet, a wiki nobody updates.

Fix: start with one or two places, and say which one wins when they disagree.

No record of who may see what

Salary files sitting next to the holiday calendar.

Fix: sort out access before launch. More on SecureLLM.ai.

A one-afternoon test

Collect twenty real questions your people ask. Give a colleague five minutes per question to find the current answer in your files.

16 to 20 found
The documents are in good shape for this project.
Ready
10 to 15 found
Fix the gaps the test showed, then start.
Fix first
Fewer than 10
The answers aren't written down yet, or can't be found.
Not yet

If a person can't find the answer, neither can the AI.

Three ways to get ready

Clean up only what the first project needs

The handful of documents behind one set of questions.

Best for: almost everyone. Weeks, not months, and you learn what clean means for you.

Gives up: the rest of the mess is still there for the next project.

Clean up everything first

A company-wide tidy before any AI.

Best for: companies already planning a move to a new document system.

Gives up: months of work before anyone sees a result, and support fades.

Switch it on and fix as you go

Point it at the shared drive and see what happens.

Best for: a private test by a small team, never staff or customers.

Gives up: trust. People who get a wrong answer early rarely come back.

What goes wrong

Pointing it at the whole shared drive. Old drafts and personal files come back as answers.
Cleaning up once. Without owners, it's messy again within a year.
Assuming the vendor will sort it out. They can convert files. Only you know which version is right.
Forgetting who may see what. The AI can surface a file nobody realized was shared.

Questions to ask your team or vendor

Which documents will the first project read, and where do they live?
How do we know which version of each is current?
Who owns each one, and when is it next reviewed?
How many are scanned or images, and how will they be read?
Who may see each document today, and will the AI respect that?

Other names you may hear

Data readiness, AI-ready data: Information in good enough shape for AI to use: current, findable and owned.
Content governance: The rules for who owns documents, how they're reviewed, and when they're retired.
Single source of truth: One agreed place where the current version lives.
OCR: Optical character recognition: reading the text out of scanned pages and pictures.
Metadata: Labels on each document, such as date, owner and department.
Permissions: Who is allowed to see each document.