How long does it take to get AI up and running in a small business?

You cannot put your business on hold for six months while somebody builds a system around it. That is the real worry behind this question, and it is the right one to have. The honest answer for most small businesses: days to weeks, not months, and nothing stops while the work happens.

What can be live in the first week?

A single automated workflow is a fast build. Following up website leads, chasing unpaid invoices, sending booking confirmations: these connect tools you already use, and they can be running within days. In the first 11 sessions of our Cook-a-Long training, attendees built 41 working tools, most inside a single session. That is people with no technical background, building something that runs, in an afternoon.

What takes a few weeks?

Anything that has to make judgements rather than just move data. An AI agent that reads incoming email, decides what each message is about and drafts a reply needs testing against your real inbox before you trust it, so plan for two to four weeks. A bespoke system that joins several processes together, say quoting, scheduling, follow-up and reporting, is typically four to six weeks from first conversation to running inside your business.

What actually decides the timeline?

Not the technology. Two things stretch every AI build. First, nobody can describe the process being automated: if it lives in three people's heads and the versions disagree, week one is spent writing it down. Second, messy data: customer records split across a spreadsheet, an inbox and someone's phone. Neither is fatal. Both are fixable. But they are where the time goes, and any agency quoting you a timeline without asking about them is guessing.

If someone quotes you six months to automate one process in a ten-person business, the six months is for them, not for you.

Does the business stop while it is built?

No, and walk away from anyone who suggests otherwise. A new system should run alongside the old way of working until it has earned trust, then take over piece by piece. When we rebuilt the booking platform for a national in-home vet service, we replicated all 25 of their live workflows like-for-like, backed by 248 automated tests, before changing anything. Bookings never stopped. Customers never noticed. That is what a competent build looks like: the switchover is boring, on purpose.

The trap to avoid is the open-ended discovery phase. The big consultancy model is months of workshops producing a roadmap and nothing that runs. You want something live in your business within weeks, proving itself on real work, then extended from there.

So the practical sequence: pick the one process that hurts most, get a single workflow live fast, and let the wins fund the ambition. If you are weighing up what that first build should cost, read our guide to what AI automation costs a UK small business.