Don't know where to start with AI in your business? Start here.

Sounds familiar?

• "Everyone says we should be using AI. Nobody can tell me for what."
• "I tried ChatGPT for a week. It was fun. Nothing in the business changed."
• "We bought a tool with AI in the name. Two people use it."
• "I'm not technical and I don't want to be sold a platform I can't evaluate."
• "My team is drowning in manual work, but I don't know which part to fix first."
• "I want someone to just tell me what to do and then do it."

Start with the work, not the tool

Three questions decide whether a workflow is a good first AI project. If you can answer yes to all three, it is a candidate. If not, it is a candidate for later, or for a process fix rather than software.

1. Does it happen often? Daily or weekly, not quarterly. Automation pays back on volume.
2. Is it mostly rules? "If the form says X, do Y" — with a few judgment calls a person can still make. Pure judgment work is a bad first target.
3. Is it stable? If the process changes every month, document and settle it first. Automating a moving target wastes the build.

Notice that none of those questions mention a model, a vendor or a platform. That comes last, once you know what you are automating and can judge tools against the actual job.

The first five things worth automating in most small businesses

Notice what is not on the list: a customer-facing chatbot, a content generator, or anything that talks to your customers unsupervised. Those can come later. They are rarely the first win and they are frequently the first embarrassment.

A 30-day plan you can run yourself

1. Days 1–5: time three recurring processes as they actually happen, not as documented. Write down every hand-off and every place data is retyped.
2. Days 6–10: pick one. Apply the three questions above. Write a one-page description of the inputs, the rules and the output.
3. Days 11–20: build the simplest version. For many first projects that is a no-code automation platform (n8n, Make, Zapier) plus a model call; for some it is a small custom system. Run it alongside the manual process, not instead of it.
4. Days 21–30: compare. Count the hours, count the errors, ask the person who used to do it. Decide whether to switch fully, extend, or stop.

Our guide to the first 30 days with AI in a small business walks through each step in detail, with the questions to ask at each gate.

What to avoid in month one

• Buying a platform before you know the workflow. You will configure it around what it does, not what you need.
• Starting with the most painful process. Painful usually means exception-heavy, which means hard.
• Letting AI talk to customers without a person in the loop. Not yet.
• A six-month roadmap. Ship one thing in 30 days and learn from it.
• Any vendor whose answer to "where do I start" is their product.

How we work if you'd rather not run it yourself

1. Diagnostic call — twenty minutes with an operator, not a salesperson. We ask about the operation, not the technology, and name the one or two workflows worth looking at first.
2. Audit — we observe the work, cost each recurring process in hours and dollars, and hand you a short ranked plan: what to automate, in what order, and what not to touch. You keep the plan whoever builds it.
3. Build — our own engineers build the first system, connected to the tools you already run. First system live in two to four weeks for most businesses.
4. Run — we monitor it, fix what breaks and extend it, or hand it over fully documented. Your choice.

Who this is for — and who should just buy a tool

This is for owner-led businesses with real recurring volume — a clinic, a brokerage, a firm, a multi-location service company — where a few broken workflows quietly cost a large number of hours every month. If you are a solo operator with one repetitive task, an off-the-shelf tool and an afternoon will likely do it, and we will point you at the right one rather than sell you an engagement.

How pricing works

The diagnostic call is free. The audit is a fixed-scope engagement whose output — costed process map, ranked plan — is yours regardless of who builds anything afterwards. Builds are quoted as fixed scope against the hours and dollars they remove, and we decline builds where the maths does not hold up.