Get your business AI ready

Whatever you run meetings in — Teams, Zoom, Meet — switch on recording and transcription today and leave it on. Store the transcripts somewhere shared, not in the organiser's personal files.

Why: Meetings are where most of a business's real knowledge shows up — decisions, reasons, context — and in most companies it evaporates the moment everyone leaves the call. Every transcript is a page of your business an AI can read later. This is the single highest-value switch because it captures knowledge automatically, from today, with no extra work from anyone.

Pick one shared drive as the home for company knowledge. Agree a naming pattern people can actually follow — date, topic, what it is — and stop filing things in inboxes, desktops and personal folders. Then write the dictionary: one simple document that says what each folder holds, what your systems are, and what your internal terms, acronyms and codenames actually mean.

Why: AI finds information the same way a new colleague does: by looking where things should be and reading what they're called. A file called "final_v7_JT_edit(2)" in someone's personal drive is invisible to both. And the dictionary is the piece almost everyone skips — your business is full of words that mean nothing outside it, and without a page that defines them, even perfectly filed documents read as noise. It's the onboarding document your AI never gets, written once.

Review the permissions on your shared drives and systems. Knowledge an AI should use needs to be readable by the people (and later, the tools) who'll use it — and anything sensitive needs to be properly restricted, not just obscure.

Why: This one cuts both ways. If everything useful is locked in personal accounts, an AI connected to your systems sees nothing. But the reverse is worse: AI is very good at finding things, so a payroll file that was only ever protected by nobody knowing where it was will get found. Fix access before you connect anything.

The things people currently learn by asking the person who knows — how you price, how you handle complaints, what the process actually is — get written down, given headings, and kept to one current version in the shared home.

Why: This is the gap that stalls most AI projects: the knowledge exists, but only in people's heads, so there's nothing for the AI to work from. One well-structured document per topic beats a hundred scattered emails. Clear headings matter more than you'd think — AI navigates a document the way a skim-reader does.

Where you have the choice, keep working documents as plain text, Word docs or simple pages rather than scanned PDFs, screenshots of tables, or designed slide decks. Text an AI can select and copy is text it can read.

Why: AI reads plain text cheaply and reliably. Scans, images and heavily designed files have to be decoded first — which costs more, fails more, and loses structure. The same information in a simple format is worth more, every time it's read, forever.