Business performance · 6 min read
You do not need a strategy deck to start with AI. You need one repeated task, a two-week trial and a person who checks the work. Here is the plain version.
French Press Labs · Updated
The short answer
To use AI in your small business, start with one repeated task that eats time every week, like answering the same customer questions or writing estimate follow-ups. Try a general assistant on it for two weeks, keep a person checking the output, and measure the time saved before you add anything else.
In this article ›
Why is one small task the right place to start?
Most owners who give up on AI tried to change everything at once. They signed up for several tools, connected none of them properly, and went back to the old way within a month.
One task is different. You can see whether it worked. You learn how the tool behaves, where it gets things wrong, and how much checking it needs. That knowledge carries over to the next task, and the next one goes faster.
A small start also limits the damage. If an assistant writes a clumsy reply to one customer email and you catch it, nothing is lost. If you hand it the whole inbox on day one, you find out about the mistakes from your customers.
Which tasks make a good first pick?
Look for work that is frequent, boring and follows a pattern. Good candidates in a typical small business:
- Replying to common questions. Hours, service area, what to bring, how long a job takes.
- Following up on estimates. The polite nudge three days after a quote went out.
- Writing first drafts. Job descriptions, a notice about holiday hours, a reply to a review.
- Summarizing. A long email thread, a supplier contract, notes from a site visit.
- Tidying data. Turning a messy list of contacts or parts into a clean table.
Poor first picks are tasks where a mistake is expensive or hard to see: final pricing, legal wording, payroll, anything involving medical advice. Those can come later, with more controls, or not at all.
If you are not sure where your time goes, write down every task you repeat for one week. The manual work finder walks through the same exercise in a few minutes.
How do you run a two-week trial?
Keep it simple enough that you will finish it.
- Write down how the task works today. Who does it, how often, how long it takes, what a good result looks like.
- Pick one tool. A general assistant such as ChatGPT, Claude or Gemini is enough for most first tasks. Do not shop for specialist software yet.
- Give it real context. Paste in your hours, your service list, two or three replies you were proud of. The more it knows about your business, the less generic the output.
- Use it on the real task every day. A landscaper might draft every estimate follow-up with it. A salon might draft every reply to a booking question.
- Review everything before it goes out. Fix what is wrong and note what you fixed.
- At the end, compare. Was it faster? Was the quality the same or better? Would you keep doing it this way?
Save the instructions that worked. A short page of "how we ask the assistant to write follow-ups" is worth more than any tool.
What should you measure?
You do not need a spreadsheet full of metrics. Three plain questions cover it:
| Question | How to check it |
|---|---|
| Did it save time? | Time the task before and during the trial |
| Was the work good enough? | Count how often you had to rewrite, not just edit |
| Did anything go wrong? | Keep a short list of mistakes you caught |
If you rewrote most of the output, the task is a poor fit or the instructions need work. If you only made small edits, you have a keeper. The list of mistakes matters most, because it tells you what kind of checking the task will always need.
Why keep a human check?
AI tools make mistakes with total confidence. An assistant can invent a warranty you do not offer, quote hours you changed last spring, or promise a same-day visit nobody can deliver. It does not know it is wrong.
A human check does not mean rereading every word forever. It means matching the level of review to the risk:
- Low risk, like an internal summary: skim it.
- Medium risk, like a customer email: read it before sending.
- High risk, like a price, a contract term or a safety instruction: a person writes or approves it every time.
Privacy belongs here too. Before anyone pastes customer details into a tool, decide what is off limits. Payment details, health information and passwords should stay out. If more than one person will use the tool, write the rules down. A one-page AI policy for a small team is enough.
What should you avoid in the first month?
- Buying before trying. Many specialist tools do what a general assistant already does. Prove the task first.
- Automating a broken process. If follow-ups are late because nobody owns them, AI will not fix the ownership problem.
- Letting it talk to customers unreviewed. That comes later, with rules and logs, once you trust the pattern.
- Expecting it to know your business. It only knows what you tell it or connect it to.
- Hiding it from the team. They will find the mistakes fastest if they know to look.
When is it time to go past a chat assistant?
A chat assistant waits for you to type. That is fine for drafting and summarizing. It is not enough when the work has to happen while you are on a ladder or asleep.
Signs you have outgrown copy and paste:
- The task depends on speed, like replying to a new lead within minutes.
- The tool needs to read your calendar, job software or inbox to be useful.
- You are doing the same paste, check and send routine dozens of times a week.
At that point the next step is an agent or an automation connected to your systems, with approval rules and a log of what it did. The guide to which tasks to automate first explains how to judge whether a task is safe to hand over.
Where should you start this week?
Pick the one task that annoyed you most last week. Time it. Try a general assistant on it tomorrow, with real examples pasted in, and review every result. In two weeks you will know more about AI in your business than any article can tell you.
If you would rather see the whole picture first, the short operations check shows where time and leads are leaking. And if you want a second opinion on what to hand over, how we help explains the way we approach it.
Frequently asked questions
- Do I need technical skills to start using AI in my business?
No. A general assistant works by typing plain instructions, the way you would brief a new hire. The skill that matters is being clear about what you want, giving it your real details, and checking what comes back before it goes to a customer.
- How long before I know whether AI is helping?
Two weeks on one task is usually enough to tell. Write down how long the task takes today, run the trial, and compare. If you cannot see a difference in time or quality, change the task or the instructions before you add more tools.
- Is it safe to put customer information into an AI tool?
Treat it like any outside vendor. Read the privacy settings, avoid pasting payment details, health information or passwords, and remove names where you can. If you work in a regulated field, check your obligations before sharing any customer records.
- Should I tell my team I am testing AI?
Yes. The people doing the task know where it goes wrong, and they will spot mistakes fastest. Tell them the goal is to remove the boring part of the job, ask them to review the output, and agree on what must never be shared with the tool.
Related
Published November 11, 2026 · Updated November 11, 2026

