IS YOUR BUSINESS
READY FOR AI?

GARBAGE IN DOES NOT
SOUND LIKE GARBAGE OUT.

That is what has changed, and it is why the old instinct does not protect you any more. A spreadsheet given bad inputs produces an obviously broken number. A model given disconnected, stale or contradictory records produces a fluent, confident, plausible answer that is wrong — and it produces it in the same tone as a correct one.

So readiness is not a question about the technology. It is a question about what the technology would be standing on.

THE TWELVE
CHECKS.

These are what TMI works through before recommending that anything be built. They are deliberately unglamorous.

Most failed projects fail one of the first four, and almost always before a line of code was written.

THE FOUR CONDITIONS
THAT ACTUALLY MATTER.

A clear task. “Use AI” is not one. “Draft the follow-up on quotes older than fourteen days” is.

Records that can be trusted. Not perfect records — trusted ones, where everyone agrees which source wins.

Agreed permissions. Decided before the system exists, not discovered afterwards when something confidential turns up in a summary.

A defined handoff. What the system does when it is not sure, and who picks it up. A system with no handoff will invent one.

AI is one implementation option. It is frequently not the first one, and a good firm will tell you that before invoicing you.

WHAT TO DO IF YOU FAIL
SEVERAL OF THESE.

Nothing dramatic, and certainly not a replacement programme. The usual answer is to connect two systems that already work, agree a source of truth out loud, and write down the judgment that currently lives in one person. That is unglamorous, comparatively cheap, and it is what makes everything afterwards possible.

It is also the point at which a lot of companies discover they did not need AI for the thing they were about to buy it for.

COMMON QUESTIONS

Can we do a small AI pilot to find out?

You can, and the common outcome is a pilot that works in the demo and dies quietly in month three, because the pilot ran on a clean sample and the business runs on the real records. If you do pilot, pilot on the messy data.

Do we have to fix everything first?

No. Fix what the specific task depends on. Readiness is per use case, not a grade the whole company has to earn before anything can start.

Is our data too old or too messy?

Old is usually fine and often valuable; twenty years of quotes is a dataset. Messy is a cost, not a disqualification. Contradictory with no agreed source of truth is the genuine blocker, and it is a management decision rather than a technical one.

How do we know which of the twelve we fail?

Walk them with whoever actually does the work rather than with whoever owns the software. TMI scores all twelve inside the Intelligent Company Audit.