Where to Start With AI: A 30-Day Plan for Small Businesses

Most small businesses don’t need an AI strategy to get going. They need one task that’s annoying enough to fix and a month to test whether AI actually helps. If you want to start using AI in your business without turning it into a big technology project, this is the plan I walk people through.

It’s the same approach I shared at the Asheville Chamber’s Smart Series in September: pick one task, measure it, test AI on it, and only then decide what’s next. Below, it’s broken into four weeks, with what to do in each one and what to watch for.

Before you start using AI: pick the right first task

The first task decides whether the whole experiment works. A good one has most of these traits:

• It happens every week, ideally every day
• It takes at least 30 minutes a week of someone’s time
• It’s mostly reading and writing: email, meeting notes, documents, short reports
• A wrong draft is easy to catch and cheap to fix
• One person owns it and can judge whether the result is good

Good first tasks we see again and again:

• Sorting the inbox each morning to find which emails actually need a reply today
• Turning meeting notes into a list of tasks with owners and due dates
• Pulling a short morning briefing together from email, calendar and your project tool
• Answering questions from your own documents, like policies, price sheets or product specs
• Keeping an eye out for information the team is waiting on, so nothing slips through

Bad first tasks are anything customer-facing that would go out without a person checking it, anything involving legal or financial decisions, and anything nobody does the same way twice. Save those for later, if at all.

Week 1 (days 1 to 7): measure, then set up

Write the task down. List the steps the way the person actually does them, not the way they’re supposed to. Five to ten lines is plenty.

Time it for a week. This is the step everyone skips, and it’s the one that matters most. Without a baseline, you’ll end the month with a feeling instead of a number. A rough log is fine: date, minutes spent, anything that went wrong.

Choose a tool you already have, if possible. If your team works in Microsoft 365, check whether Copilot Chat is on your plan. Google Workspace Business plans have Gemini built in. If you need something else, use a business plan such as ChatGPT Business or Claude Team, not personal accounts. Our guide to what AI implementation costs compares the current prices.

Set ground rules before you start using AI

Set a few ground rules. Until you have a proper AI use policy, three rules cover most of the risk:

• No client financial, health or personal data goes into the tool unless you’ve confirmed the plan allows it and you’re comfortable with the vendor’s terms
• A person reviews every output before it’s used
• If you’re not sure whether something can go in, ask first

Write the instructions. This is where people expect magic and get mush. Be specific about the job, the format and what to do when something is missing. For meeting notes, an instruction like this works well:

You turn meeting notes into a task list. For each commitment, write the task, the person responsible and the due date. If an owner or a date isn’t stated in the notes, write “not stated” instead of guessing. List anything that sounds like a decision separately.

That last line about not guessing matters. AI will happily invent an owner or a deadline to fill a gap, and those are exactly the mistakes that cause trouble later.

Week 2 (days 8 to 14): run it side by side

Do the task both ways for a few days: the old way, and with AI. Then compare.

Keep the same simple log, plus two columns:

• Corrections. How much did you have to fix in the AI’s draft?
• Misses. Did it leave out anything important?

Most of the improvement in week 2 comes from tightening the instructions. If the AI keeps writing long summaries when you wanted bullets, say so. If it misses commitments buried in the middle of the notes, tell it to read the whole document before answering. Small changes to the wording often fix the biggest problems.

Keep a person in the loop for everything. The goal this week isn’t speed. It’s finding out where the AI is reliable and where it isn’t.

Week 3 (days 15 to 21): make it the default

If week 2 went reasonably well, switch to doing the task with AI first and checking the result, rather than doing it by hand and comparing.

Two things help this stick:

• Save the instructions where the team can reuse them. ChatGPT has projects and custom GPTs, and Claude has projects, so nobody has to retype the setup each time. A shared document works too.
• Add a second person. Someone who wasn’t part of the setup is the best test. If they can use it with a one-page explanation, it’s ready to keep. If they can’t, the process needs simplifying, not the person.

This is also the week to notice the small stuff. Is the output going somewhere useful, or does someone still copy and paste it into another app? That copy-paste step is often the next thing worth automating.

Week 4 (days 22 to 30): measure again and decide

Time the task again for a few days and compare it with your week 1 baseline. Then make one of three calls:

• Expand. It saves real time and the quality holds. Roll it out to everyone who does the task, and pick the next one. Our AI rollout plan covers how to bring the rest of the team in.
• Adjust. It helps, but not enough yet. Give it another two weeks with better instructions or a narrower job.
• Stop. It didn’t help. That’s a fine result. You learned it for the price of a month, not a big project.

Write a short summary for yourself: what changed, how many minutes a week it saves, what went wrong. When you’re deciding where to spend money on AI next, that paragraph is worth more than any vendor demo.

What a typical first month looks like

Here’s a composite example, not a specific client, of how this tends to go when a small office starts using AI on meeting follow-ups:

Week	What happens	Time on the task
1	Baseline. Notes are turned into tasks by hand after each meeting.	About 4 hours
2	AI drafts the task list. Lots of corrections at first, then fewer as the instructions improve.	About 4 hours (doing it both ways)
3	AI goes first, a person checks. A second team member starts using it.	About 2 hours
4	Instructions are saved and shared. The team decides to expand.	About 1.5 hours

Your numbers will be different. The pattern usually isn’t: week 2 is slower, week 3 is where the time savings show up, and week 4 tells you whether it’s worth keeping.

Mistakes to avoid when you start using AI in your business

• Buying seats for everyone on day one. Start with the people who have a clear use for it.
• Starting with the hardest problem. Your messiest process is a bad first test. Build confidence on something smaller.
• Skipping the baseline. Without the week 1 numbers, you can’t tell whether it worked.
• Letting AI send things on its own. Keep a person in the loop until you’ve seen it work for weeks, not days.
• Using personal accounts for company work. Business plans come with admin controls and clearer data terms. They’re worth it.

After the first 30 days

Once one task works, the next steps tend to fall into two groups. Some businesses repeat the same process with a second and third task. Others find that the real bottleneck is moving information between systems, and that’s when connecting AI to tools like QuickBooks Online, NetSuite or an automation platform like Zapier starts to pay off. We cover the current options in our guide to connecting AI to QuickBooks, NetSuite and Shopify.

Either way, you’ll be making that decision with a month of real data instead of a guess.

Getting help in Asheville and across North Carolina

You don’t need outside help to start using AI this way. If you’d like some, there are good options:

• The Asheville Area Chamber of Commerce runs the Smart Series, where local business owners share practical topics like this one.
• The NC Small Business and Technology Development Center offers no-fee, confidential business coaching across the state.
• The Small Business Center Network at North Carolina’s community colleges runs free and low-cost seminars and free counseling.

And if you’d rather have someone walk through it with your team, that’s what we do. Good People Technologies helps with AI implementation for small businesses in Asheville, Western North Carolina and across the state. A free 15-minute operations and technology audit is a good place to figure out which task to start with.

Frequently asked questions

How much time does this plan take?

Beyond the task itself, plan on an hour or two a week: setting up in week 1, comparing results in week 2 and measuring in week 4. Week 3 should take almost no extra time.

Do we need to hire someone to start using AI in our business?

No. One curious person with a clear task can run this plan. Outside help is most useful when AI needs to connect to your business systems, or when several teams want to use it in a consistent way.

Which AI tool should we start using?

Usually the one you already pay for. If your team lives in Microsoft 365, start with Copilot. If you’re on Google Workspace, start with Gemini. The plan matters more than the tool.

What if our first task doesn’t work?

Pick a different task, not a different tool. Most failed first attempts come from a task that was too vague or too big, not from the AI itself.