AI readiness checklist: 8 signs your company's knowledge is ready for an AI assistant
An AI assistant is only as good as the documents it reads. Before a rollout, the question isn't which model to choose or how many licences to buy. It's whether the knowledge behind those licences is accurate enough to rely on. This checklist covers the eight signals that separate organisations ready to scale AI from organisations about to find out the hard way.
Readiness, in other words, is a documentation question dressed up as a technology decision. Most rollout plans budget for licences, training and change management, and skip the one variable that decides whether any of that pays off: the state of the content the assistant will actually draw from.
What "AI readiness" actually means
AI readiness is the state of an organisation's documentation: how current it is, how consistent it is across teams, and how clearly it's owned. It has very little to do with computing power, model choice or seat count.
An AI assistant doesn't invent answers from nothing. It retrieves and summarises what's already sitting in shared drives, wikis and policy folders. If two documents disagree about a retention period, the assistant can't tell which one is right. It picks one and states it as fact. That's the core mechanic: AI doesn't create bad information, it scales it. A conflicting policy that used to confuse the one employee who opened the wrong file now reaches everyone who asks.
Even the best system tested on 500,000 company documents gets 1 in 3 answers wrong (EnterpriseRAG-Bench, 2026). So the rollout question changes. "Are we AI-ready?" isn't about infrastructure. It's this: is our documentation accurate, current and owned well enough that an assistant repeating it at scale is a good thing?
The core question to ask before rollout
Before you expand an AI assistant to more teams, ask one question: if someone asked your assistant something today, would the answer be reliably correct?
Not "would it sound sure of itself". Assistants sound sure whether or not the content underneath is accurate. The question is whether the source documents would hold up to scrutiny. Teams that have already given an assistant to a pilot group often find the honest answer is "it depends which document it pulls from". That's the readiness gap this checklist is built to surface.
Early signs you may already be seeing
Readiness gaps often show up in daily work long before anyone calls them a documentation problem. Watch for these, especially if an assistant is already live for a first team or department:
- Helpdesk or support tickets that trace back to someone acting on an AI answer that turned out to be outdated or contradicted by another document.
- Answer quality that varies by team for no obvious reason. Usually one department's documentation is in better shape than another's, rather than anything about the assistant.
- People quietly double-checking AI answers against the source document before acting on them, which cancels out much of the time the rollout was meant to save.
- Access that's been informally ring-fenced to the folders or teams everyone already relies on, because nobody is sure what would happen if it were opened up further.
- Leadership asking why the rollout hasn't delivered the expected return, with no clear answer, because the real constraint (the documents) was never measured.
None of these mean the AI is failing. They mean the assistant is faithfully reflecting documentation that was never checked for consistency before people started relying on it. The checklist below is designed to catch exactly this, ideally before a pilot expands rather than after.
The 8-point readiness checklist
Each point is a specific condition you can check in your existing documentation. None of them need new tools to start: a working knowledge of your shared drives is enough for a first read.
1. Do you have conflicting versions of the same policy?
Search your shared drives for your most-referenced policies, such as expense limits, remote work rules, data retention periods and approval thresholds, and check that every copy says the same thing. A retention policy that says 24 months in one document and 36 months in another is the textbook case. Neither is marked as outdated, but only one can be true. An assistant with no way to know which is authoritative will serve either.
If a spot check turns up even a handful of these, treat it as a signal, not an exception. Conflicts cluster wherever a policy was updated in one place and copied, but not updated, somewhere else.
2. How old is your most-referenced content?
Age alone isn't the problem. A well-maintained document from three years ago can be perfectly accurate. The risk is content that hasn't been touched in a long time and still gets opened regularly, because nobody has checked whether it's still right.
A practical check: pull the most-viewed documents in each department and look at the last-modified dates. Anything untouched for well over a year that still gets regular traffic deserves a review before an assistant starts quoting it as current.
3. Does every area of knowledge have a named owner?
For each major topic (HR policy, IT security, supplier onboarding, pricing) ask: "if this document turns out to be wrong, who fixes it?" If the honest answer is "not sure", or "whoever wrote it, and they've left", that's an ownership gap.
Ownership gaps matter more once AI is involved, because the informal review that used to happen disappears: a colleague asking around, someone noticing that something looked off. An assistant doesn't flag its own uncertainty. It needs a working human loop behind it.
4. Are placeholder or unfinished documents still live?
"TBD", "TODO" and "to be confirmed" are common in drafts and usually harmless, until a document containing them is treated as finished and an assistant reads it. A placeholder that a person would recognise as unfinished reads, to an AI, like any other content.
A quick search for these terms across your busiest shared drives usually turns up a handful of documents that were never actually finished.
5. Do different departments answer the same question differently?
This is the cross-team version of the conflict check. Ask the same operational question in two departments ("what's our current approval threshold for a purchase?" works well) and see whether the documented answers match. Departments that work fairly independently, which is most of them, build up their own local versions of shared policies without noticing that another team's version has drifted.
This matters for AI rollouts because an assistant doesn't respect the department lines that people informally do. It retrieves whichever version is closest to the question, whichever team wrote it.
6. Do you know how many people a wrong answer would reach?
Not every inconsistency carries the same weight. An outdated note in a folder nobody opens is low priority. A conflicting figure in a policy hundreds of employees read every week is a different kind of risk. Readiness includes knowing, roughly, which of your inconsistencies sit in high-traffic documents and which sit in quiet corners, because that decides how fast a wrong answer spreads and to how many people.
7. Has anyone tried to clean this up by hand before, and did it stick?
Many organisations have already run a clean-up at some point: someone audited a shared drive, flagged outdated files and asked owners to review them. If that happened and the same problems have crept back, that tells you something. The issue isn't a one-time mess, it's an ongoing maintenance gap. One-off clean-ups tend to decay within months once normal work resumes, because nothing is watching for new conflicts as they appear.
8. Would you be comfortable if this content were repeated, word for word, to every employee who asked?
This is the readiness question in its simplest form. Read a sample of your most-referenced documents and ask: if an assistant repeated this exact content to someone as the final word, would it hold up? If the honest answer for a meaningful share of your documentation is "probably not", that's the gap to close before you scale further.
Why readiness varies by department
Readiness is rarely even across an organisation. IT documentation is often the best maintained, because it tends to have the clearest ownership and the most frequent reviews. Legal and HR policies are high-stakes but slow-moving, so conflicts can sit unnoticed for a long time. Sales content (pricing, packaging, competitive positioning) tends to change most often, and so is the most likely to drift out of step across regions or teams.
The practical result: a company-wide rollout can work well in some departments and cause real problems in others, purely because the documentation underneath is uneven. Assessing readiness area by area, rather than as one yes-or-no verdict for the whole company, shows you where to expand first and where to slow down.
How to run a light readiness audit
A first audit doesn't need new infrastructure. A useful first pass looks like this:
- Pick your three or four highest-traffic knowledge areas. Usually IT policy, HR policy, and whatever touches customers directly (support macros, pricing, onboarding).
- Pull the most-referenced documents in each area and run them through the eight checks above.
- Note where you find conflicts, outdated content or ownership gaps, and roughly how many people each document reaches.
- Give each finding to a named person with one specific decision to make. "Confirm which retention period is correct" is actionable. "Review our documentation" is not.
- Check again in a few weeks. A manual audit gives you a snapshot. The real test is whether the same areas still look right the next time you look.
A few notes make the manual version go more smoothly:
- Start with your highest-traffic content, not your largest folders. A shared drive with ten thousand files is not a useful starting point. The twenty documents your team opens every week are. Volume isn't the risk, reach is.
- Treat the audit as ongoing, not a one-off project. Documentation keeps changing after the audit ends. A single snapshot tells you about the past, not about next month.
- Give findings to people, not teams. "HR should look into this" tends to go nowhere. "Priya, can you confirm which retention period is correct by Friday?" tends to get resolved.
- Track how long findings stay open. A conflict that has been open for six months is a different risk from one flagged yesterday, even if they look similar on paper.
This works for a handful of departments checked by hand. It gets much harder to keep up as the organisation grows, or as more of the company relies on AI answers every day.
What changes once you scale past a pilot
Rollouts often start small and look fine. A pilot team gets access, feedback is positive, and the obvious next step is to open it up. That's usually when readiness gaps start to show. A small pilot group tends to know, informally, which documents to rely on and which to double-check. That knowledge doesn't travel with the assistant once it reaches people who weren't in the pilot.
The organisations that scale AI smoothly tend to treat the readiness check as an ongoing discipline, not a one-time gate before launch. They revisit it as documents change, teams reorganise and new areas come into scope.
Where Anchoris fits
Anchoris runs the checks behind this list on your own documents in Google Drive and SharePoint. It finds contradictions between documents, outdated and expired content, people named as owners or contacts who have left, unfinished placeholders, broken links and duplicates. It ranks each issue by how many people can open the document, suggests who should fix it, and fixes it at the source with approved workflows (move, rename or move to trash), with every run logged. You get a quality score from 0 to 10 for your documents, and first results within an hour.
You choose which drives Anchoris can reach, and nothing changes without your approval. It works alongside the assistant you already use, whether that's Copilot, Gemini, Claude or Glean: fix the source once, and every assistant that reads it answers correctly.
See where your documents stand before you scale.
Book a demoFrequently asked questions
How do I know if my company is ready to roll out an AI assistant?
Readiness comes down to your documentation, not your technology. If your most-referenced policies are consistent across teams, reasonably current and clearly owned, you're in good shape. If you find conflicting versions of the same policy, outdated content still in use, or documents nobody is responsible for, close those gaps before you expand access.
What is an AI readiness audit?
An AI readiness audit is a review of an organisation's documentation to find the issues that would make an AI assistant give wrong or inconsistent answers: contradictions between documents, outdated content, missing owners and unfinished documents. It can be done by hand as a one-off project, or with a tool that does the checking for you.
What causes an AI assistant to give a wrong answer?
Most wrong answers trace back to the documents, not the AI model. If two documents disagree, if a document hasn't been updated in a long time, or if information is missing, an assistant retrieving from that content repeats whatever it finds, including the wrong version.
How often should we re-check our documentation once an AI assistant is live?
Documentation changes all the time, so a one-off audit only reflects a single moment. Many organisations find that a review every few months catches the major issues, but the gap between checks is exactly where new conflicts build up unnoticed.
Can't the AI assistant we already pay for, like Copilot or Gemini, catch this itself?
General-purpose assistants are built to find and summarise documents, not to check whether those documents agree with each other. They present what they find with equal certainty, accurate or not. Catching contradictions, outdated content and ownership gaps needs something that looks across and inside documents for exactly that, which is a different job from answering questions.
Does AI readiness mean the same thing for every department?
No. IT documentation is often better maintained than HR or sales content, simply because of how often each is reviewed. Assessing readiness area by area, rather than as one company-wide verdict, shows you where to expand AI access first.
Is a documentation clean-up a one-time project or an ongoing task?
Ongoing. Documentation keeps changing after any clean-up ends, and new conflicts, outdated content and ownership gaps build up again from a cleaner starting point. Organisations that treat readiness as a regular discipline keep the gains from a first clean-up far longer than those that treat it as a single project.
What's the difference between an AI readiness audit and document management?
Document management tools help people store, organise and find files. An AI readiness audit looks specifically for the conditions that make an assistant give a wrong answer: contradictions between documents, outdated content and knowledge nobody maintains. Good storage and organisation don't guarantee that two documents don't quietly disagree.
How to find duplicate files in Google Drive
Read nextMore from the blog
Your AI pilot worked. Why did it break at scale?
The model didn't get worse. The rollout reached documents the pilot never touched.
The anatomy of a documentation audit: what to look for
Most audits end as a 400-row spreadsheet. The four things a useful audit looks for, and how to keep its findings from decaying.
Why does AI give wrong answers? Often it is the documents
The model rarely invents your policy. It repeats the wrong version of it. Four document failures behind most wrong answers, and how to find them.