Your first 90 days of AI in a small business
Spend the first 30 days learning what you already have and writing down the tasks that repeat every day. Use days 31 to 60 to fix one of those tasks, with a tool you already pay for if it can do the job, and days 61 to 90 to fix a second one or build the piece nothing on the market does.
No vendor tour, no 40-tab research project. Ninety days is enough to fix one or two real things, if you spend the first month looking instead of buying. Hope this helps.
Founder, Handled. Greenville, SC. Updated October 2026, 6 min read.
Days 1 to 30: learn what you have
No new tools this month. Just paying attention.
Write down a normal week. What happens when you open the laptop in the morning? What do you type, copy or chase every single day? This is the first thing I ask any business owner, and the repeated stuff jumps out fast once it's on paper.
List every tool you pay for. What it costs, who uses it, and which parts of it you actually open. Then skim each one's feature list. Odds are decent that something you already pay for does the thing you were about to go buy.
Find the spreadsheet. Almost every business has one file that quietly runs everything: the schedule, the order list, the customer tracker. Whatever that spreadsheet is doing is a good hint at what your tools are missing.
Pick the one task that bugs you most. Not the most impressive one. The one that repeats the most and makes everybody groan. That's your days 31 to 60.
Days 31 to 60: fix one thing
One task, done properly, with the people who do it every day.
Write down the before. How long the task takes and how often it happens. You'll want this number later, and you can't get it back once things change.
Check what you already own first. Your CRM, your booking tool or your email platform might already have the setting. Turning on a feature you pay for is the cheapest fix there is.
Shop only when your current tools can't do it. If you need two apps to talk to each other, something like Zapier is the usual first stop. It has a free plan, and paid plans start at $19.99 a month billed yearly. Small and boring is the goal here.
Let your team break it. Have the people who do the task every day test the fix before it goes live. They'll find the weird cases in an afternoon.
Days 61 to 90: fix the next one, or build what's missing
By now you know whether the tools you have can do the job.
If your tools can do it, repeat days 31 to 60 on the next task. It goes faster the second time.
If nothing fits, that's when a custom build makes sense. Keep the tools that are doing their job and build only the piece that's missing, connected to what you kept. Ripping everything out is rarely the answer.
Write down the after. Same task, same measure as your before. Share it with the team, because they did the work and they should see that it mattered.
Where 90-day plans go sideways
1. Swapping the plan halfway through. Around day 45 a new tool will look like a shortcut. Finish the fix you started, then decide.
2. Adding to the fix before it ships. Somebody always wants texts and alerts and a dashboard on top. Ship the first version. Add the rest next month.
3. Skipping the before number. If you don't write down how long it took, you'll never know if the fix was worth it.
Questions people ask me
Got one that isn't here? Ask me on a free call →
What should the first 90 days of AI actually look like?
Days 1 to 30, learn what you already have and write down the tasks that repeat every day. Days 31 to 60, fix one of those tasks, with a tool you already pay for if it can do the job. Days 61 to 90, fix a second one, or build the piece nothing on the market does.
Do we need an AI person on staff to do this?
No. You need one person who knows how the day actually runs and has a couple of hours a week to own it. That's usually you or whoever keeps the place running. Bring in outside help when you hit something your tools can't do, not before.
What if the first fix takes longer than 30 days?
Then it takes longer. Finish the first one before you start the second. One fix that your team actually uses beats two that are half done.
How do we know if it worked?
Before you change anything, write down how long the task takes and how often it happens. A month after the fix, write it down again. That's your answer, in your numbers, not somebody's industry average.
Should the whole team be part of this?
Yes. The people doing the task every day know where it breaks. Ask them what they'd fix first, and let them test the fix before it goes live.
When you don't need me
Most of this plan is stuff you can do yourself.
The first 30 days. Writing down your week and listing your tools doesn't need anybody but you and a notepad.
Your fix is a setting you already pay for. Turn it on, learn it, done.
Your current setup works and your team uses it. Don't go looking for a project just because AI is in the news.
What changed
- October 2026: rewritten top to bottom around learning what you have first. Removed the made-up savings math (hours saved, dollar ROI), the cost and time estimates we couldn't back up, and the advice to bring in a fractional executive. Added the before-and-after way to measure.
- May 2026: first published.
Sources
Every price and outside fact on this page comes from the sources below, checked October 3, 2026. Prices change often, so check before you buy.
More from the resource library
Start with what you have. Build what's missing.
If you want a second set of eyes on your first 30 days, tell me what a normal day looks like. I'll tell you honestly which task I'd fix first, even if the answer is a setting you already have.
Book a free call15 minutes. No sales pitch.
Get the free missed-money checklist.
Seven places a service business quietly leaks revenue, with the actual text message for each one. None of them are about working harder, and none of them involve firing anybody. Steal the templates and change the words so they sound like you.
Sent. Check your inbox in a minute, and your spam folder if it's being weird.
Email's being stubborn on our end. Grab the PDF directly.
Related reading.
What Is a Chief AI Officer? The Honest Definition
What a Chief AI Officer is, what they do, how the role differs from a CTO, what it pays, and whether a small business needs one.
AI Roadmap for Small Companies: The Operator Playbook
An AI roadmap for a small company: learn what you already have, pick a tool if one fits, build what's missing. What each step looks like, what it costs, and the mistakes that stall it.
The AI Readiness Audit: A Framework for Small Companies
A free AI readiness self-check for a small business: five questions about your daily routine, your tools, where your info lives, logins and your team, and what to do next based on your answers.
AI Strategy for B2B Services Companies: Where to Start
AI strategy for a small B2B services company: start with intake, proposals, onboarding and status updates, keep the client work human, and avoid the usual rollout mistakes.
The Fractional AI Operator Role: What It Actually Includes
What a fractional AI operator is, what the role usually includes and leaves out, how it differs from a fractional AI executive, what to ask about cost, and when you need a build instead.