On this page
Where do I start with AI in my small business?How do I pick the first task for AI?Should I use software I already own, a chat assistant, or a custom build?What should I not do first?How will I know it worked?When is waiting the right answer?When does it make sense to get help?Questions people askMany owners are not short on AI ideas. They are short on a first step that does not waste money. The tools change every month, the advice contradicts itself, and nobody wants to buy the wrong thing.
So here is a plan for one week. It costs nothing but some of your time, it works whether or not you ever hire anyone, and at the end you will know what to do next, including doing nothing.
Where do I start with AI in my small business?
Start with your own week, not with a product. Give each day one small job. See AI agent use cases by industry for examples. Working alone? AI automation for solo business owners covers what fits one person.
- Day 1: find where the hours go. Write down everything you and your staff do that repeats. The five-minute time audit walks you through it.
- Day 2: pick one job. Choose the repeating task that costs the most time and has clear steps. Write those steps in plain sentences, including what you do when something goes wrong.
- Day 3: check what you already own. Look in your current software for AI features before you buy anything new. Jobber, for example, offers an AI Receptionist, and Housecall Pro offers CSR AI, which is a paid add-on.
- Day 4: test a chat assistant on the job. Use made-up or non-sensitive examples. See whether it can draft what you need, and note where it gets things wrong.
- Day 5: score your readiness and decide. Run the ten-point readiness checklist, then choose: keep using a tool, change the process first, or get help with a build.
How do I pick the first task for AI?
A good first task has most of these traits:
- It happens at least weekly, in nearly the same way each time.
- The right answer is clear, so you can tell quickly if the AI got it wrong.
- A mistake is cheap, or a person checks the result before it goes out.
- Somebody on your team already dislikes doing it.
Examples that often fit: drafting replies to common customer questions, writing up job notes, following up on quotes with no answer, sorting a shared inbox.
Want this explained for your business?
Ask Glenn first. He reads every note himself and will tell you plainly whether AI fits your business.
Should I use software I already own, a chat assistant, or a custom build?
Try them in this order. Each step up costs more and takes longer, so move up only when the one below does not fit.
| Option | Good when | Watch for |
|---|---|---|
| AI inside software you already use | Your software offers what you need, such as answering calls and booking visits | Check your plan and what it costs. Some features are paid add-ons |
| A chat assistant you use yourself | The job is drafting, summarizing or answering questions, and you can paste in the details | Do not paste client information until you have checked the provider's data settings |
| A custom build | The work spans several of your programs and repeats often | It costs real money and needs upkeep, so be sure the job is worth it |
What should I not do first?
A few mistakes come up often enough to name:
- Do not buy an annual plan before you have tested the tool on your own job.
- Do not put client or customer information into a consumer account. Providers treat consumer and business plans differently. Anthropic, for example, says it does not train on inputs or outputs from its commercial products by default and handles consumer plans under separate rules. Read your provider's terms.
- Do not automate a process your team does three different ways. Agree on one way first.
- Do not start with the flashiest use. Anthropic's own advice on building agents is to find the simplest solution possible and add complexity only when needed.
- Do not feel behind. The US Census Bureau's survey found about 19.8 percent of US businesses were using AI as of May 3, 2026, and fewer than 20 percent of firms with four or fewer employees.
How will I know it worked?
Decide the measure before you start. Write down how many minutes the job takes you now and how often you do it. After two to four weeks of using the AI, count again. Include the time you spend checking the AI's work, because that time is real.
If the saving is small, stop and try a different job. A test that ends in "not worth it" is still a good result. It cost you a week instead of a purchase.
When is waiting the right answer?
Wait if your process changes every month, if the job is rare, or if you have not yet decided how you want it done. Automating a moving target locks in the confusion. Spend the time writing the process down instead. That helps your staff today whether or not AI ever enters the picture.
When does it make sense to get help?
When your week-one test shows the job is worth doing and it needs to connect to your other programs, or you do not have the evenings to build and maintain it. Primus AI Consulting builds three kinds of AI: an assistant from $3,000 to $6,000, a workflow agent from $6,000 to $15,000 that drafts and waits for your approval, and a private on-site agent from $15,000 to $30,000. The services and prices page has the details.
The first step is a 60-minute working session with me, the Blueprint Session, and it is free. You get a written plan within five business days that is yours to keep whether or not you build. If your week-one test shows a tool you already own is enough, I will say so.
Questions people ask
How much should I spend to start with AI?
Nothing, for the first week. Use software you already own and a chat assistant with non-sensitive examples. Spend money only after you know which job is worth it. Custom builds with Primus AI Consulting start at $3,000.
Which AI tool should I try first?
Start with whatever is already in your software, then a general chat assistant. The tool matters less than the job you pick. Test two tools on the same task and compare the drafts.
Is AI worth it for a business with three employees?
Possibly, if one repeating job eats real hours. The test is whether the weekly time saved, after you check the AI's work, is worth the cost. For very small teams, a chat assistant or a feature in your existing software is often the right size.
How long until I see results?
A first test can show you whether it helps within a week. A bigger project takes longer. Measure the job's time before and after two to four weeks, and be honest about the checking time.
Will my staff think AI is replacing them?
Tell them which tasks you want help with and ask what they would hand off first. People are usually glad to give up the dull repeating parts. Be clear that a person still reviews anything that reaches a customer.
What if I try this and decide AI is not for me yet?
That is a fine result. You spent a week and no money, and you now have your process written down. Come back when the volume or the pain justifies it.
Sources
Understanding AI is step one.
When you want to see what it would look like in your business, a Blueprint Session turns it into a written plan and a fixed price. The session itself is free.
