The hard part about how to implement AI in small business isn't picking a model. It isn't choosing a vendor. It isn't even budget.
It's that most businesses aren't actually ready — and they don't know it until 90 days in, when the implementation stalls and someone has to explain why.
We've seen this enough times to predict it. Here's the readiness check we run before we accept any AI engagement, and a 30-day plan to get there if you don't pass it yet.
The five things that need to be true before AI works
1. You can describe the workflow you want to fix in one paragraph.
If your answer is "we want to use AI to be more efficient," you are not ready. If your answer is "we want every inbound lead from our website to get a tone-matched reply within 5 minutes and a calendar link within 10," you are.
The first version is a vibe. The second is a project.
2. The data the AI needs to do its job actually exists somewhere.
AI is not magic. It needs inputs. If you want an AI agent to qualify leads against your "ideal customer profile," that profile needs to exist as something other than a feeling in the founder's head.
If you want it to draft replies in your voice, you need 20-50 actual past replies it can learn from.
If you want it to route customer support tickets, you need historical tickets and resolutions in a place it can read them.
Missing inputs is the #1 reason implementations fail.
3. A human owns the workflow today.
Counterintuitive but important: AI can't replace a process nobody currently runs. If "leads coming in from the website" is technically somebody's job but in practice nobody does it consistently, AI doesn't fix that — it just automates the gap.
Before AI: at least one human runs the workflow end-to-end, with a documented process, even if it's slow and inconsistent.
4. There's an owner of the AI system once it ships.
Who watches it? Who notices when it gives a bad answer? Who decides when to update its training data?
This person doesn't need to be technical. They need to be accountable. "We'll figure it out as we go" is not an answer. AI systems decay without an owner.
5. You agree on what success looks like, in a number.
"Better customer service" is not a success metric. "First-touch response time under 10 minutes, on at least 90% of tickets, without adding headcount" is.
If you can't name the number, you can't tell if it's working — and you certainly can't tell if it's worth what you're paying for it.
The 30-day readiness plan
If you've read those five and realized you're closer to "we have a vibe" than "we have a project," here's how to fix it in 30 days.
Week 1 — pick one workflow.
Not three. One. The single workflow where AI would make the biggest visible difference if it worked. Most businesses pick from these: lead response, customer support triage, appointment booking, sales follow-up, content repurposing, or internal reporting.
Pick the one with the most pain and the most volume. Write it down in one paragraph using the format from #1.
Week 2 — collect the inputs.
For the workflow you picked, gather the data the AI will need. Past examples, decision rules, qualifying criteria, tone samples, anything a new hire would need to do the job. Put it all in one folder.
This week is unglamorous. Do it anyway. It's the #1 predictor of whether the implementation will work.
Week 3 — document the current process.
Write down exactly how this workflow runs today, step by step. Include what triggers it, what happens at each step, what tools are involved, and who's responsible. If nobody runs it consistently today, document the version you wish was running.
This becomes the spec the AI builds against.
Week 4 — set the success metric and pick an owner.
What number proves this is working? Who owns making sure it stays working? Get explicit commitment on both before you spend a dollar on implementation.
That's 30 days. No code, no vendors, no tools — just the work that makes AI actually stick.
What happens after 30 days
If you've done the four weeks above, you'll have something rare: a documented, owned, measurable workflow ready to be automated. From there, the actual implementation is the easy part.
You'll also know — honestly — whether AI even makes sense for your business yet. Sometimes the answer at the end of 30 days is "we need to fix the human process first." That's a useful answer.
The businesses that skip this prep and go straight to implementation are the same ones writing post-mortems six months later about why their AI investment didn't work.
Ready to see what AI can actually do for your business? Book a Free Audit