Most organisations hold years of knowledge in PDFs, reports, policies and web pages. People can read them, slowly. AI often can’t, or it gets them wrong. That is becoming a business problem.
More and more, AI is how customers, staff and stakeholders find answers. They ask ChatGPT, Copilot or Google’s AI summaries a question, and the answer is built from whatever information the AI can find and read. If your information isn’t in a form it can use, it answers from someone else’s, or it guesses.
Inside the organisation it’s the same story. Staff ask an AI assistant about a policy or a past report, and the answer is only as good as the documents behind it.
Why AI struggles with your information
Most organisational information was written and published for people. A report designed for print, a long PDF with tables and footnotes, a web page that buries the key fact in the fourth paragraph: a person can work through these. AI tools often can’t.
They lose the structure of a table, miss a heading, merge two sections or read an old document as current. The result sounds confident but isn’t accurate. And because the answer looks right, nobody notices until it matters.
What AI-ready information looks like
We think about it in four steps. Information is AI-ready when AI can:
- Find it. It’s published where AI tools can reach it, not locked behind a scan or a login it doesn’t need.
- Extract it. Its content and structure come out intact, including tables, headings and references.
- Understand it. Key facts are stated plainly, with clear headings and structured data that says what each thing is.
- Reuse it accurately. There’s one current version, and the answer can be traced back to its source.
Measure, then fix
That’s the problem we’re working on. We make an organisation’s information easy for AI to find, extract, understand and reuse accurately. Then we measure how AI interprets it and fix what it gets wrong.
For websites, that means our AI discoverability work: we check how AI search describes a business and answers questions about it, compare that with the facts, and change the website so the answer is right.
For documents, it means turning them into content that people and AI can both use.
25 years of law reform, made machine-readable
The Victorian Law Reform Commission is one example. It had 25 years of law reform locked in PDFs. We are making those documents machine-readable and web-searchable, with people reviewing anything the system flags. The same problem exists across government, law, education and many other industries.
The first version started as a capstone project with four RMIT Vietnam students, who placed third out of more than 40 projects at this year’s awards. Read about the team.
Where to start
Three questions will tell you how ready your information is:
- Which documents or pages do people ask about most?
- If you ask an AI tool a question about them today, is the answer accurate?
- What does AI search say about your organisation, and is it right?
If the answers make you uneasy, you’re not alone. Talk to us →