Your first 90 days with AI: what actually happens

You've heard the pitch. You know AI exists. You're not sure if it's for you, and you don't want to spend six months and $50,000 finding out it isn't.

Here's what the first 90 days actually look like for a business that's never used AI before.

Month one: learning what's real

Your first month is not about building anything. It's about getting specific.

Start by listing 5 to 10 problems your business solves the same way every week. A dental clinic in Kelowna might spend 2 hours a day answering the same questions about insurance coverage, appointment policies, and post-care instructions. A 12-person accounting firm in Burnaby might spend 3 hours a week manually sorting incoming tax documents by client. A property manager in Surrey might spend 4 hours a week answering tenant questions about lease terms, maintenance requests, and move-out procedures.

These are your candidates.

Next, pick the one that costs you the most time or causes the most friction. Not the one that sounds impressive. The one that actually bothers you or your team. Time spent on repetitive work is time not spent on the work that makes money or matters.

Then, talk to someone who has done this before. Not a vendor. Someone who has actually built an AI tool in a business like yours. This is where most owners get stuck: they talk to salespeople, not to people who have lived through the messy middle.

What you're doing in month one is answering three questions:

• Is this problem specific enough that a tool can solve it. ("We get a lot of questions" is too vague. "We answer the same 12 questions about our return policy 15 times a week" is specific.)
• Would solving it actually save time or money. (If it saves 2 hours a month, it's not worth doing. If it saves 5 hours a week, it might be.)
• Can we describe the solution clearly enough to build it. (If you can't explain what the tool should do in two or three sentences, it's not ready yet.)

Month one usually takes 4 to 6 hours of your time, spread across a few conversations.

Month two: one small test

Now you build the smallest version of the tool that could actually work.

A plumber in Richmond with a dispatcher who spends 30 minutes every morning organizing the day's jobs might build a tool that reads incoming service calls and suggests the best order to visit them. A distributor in Calgary with a warehouse team that manually checks incoming invoices against purchase orders might build a tool that flags mismatches in 30 seconds instead of 10 minutes per invoice.

The key word is small. You're not automating your entire operation. You're solving one narrow problem in a way that a person can still check the result before acting on it.

This is called a pilot project, and it usually takes 2 to 4 weeks. Your team will spend 3 to 5 hours setting it up, testing it, and teaching it to do the right thing. There will be mistakes. The tool will miss cases or misunderstand context. This is normal and expected. You fix it and try again.

The tool doesn't have to be perfect. It has to be useful enough that your team uses it instead of skipping it.

Month two usually costs $3,000 to $8,000 in software and setup. Most of that is labour, not magic.

Month three: the real decision

By week 8 or 9, you'll know whether this works.

If the tool saves the time you expected and your team actually uses it, you have two choices: expand it to solve more of the same problem, or build a second tool for a different problem.

If the tool saves less time than you hoped, or your team doesn't use it, you stop. You've spent 90 days and $10,000 to learn something true about your business. That's not a failure. That's information.

If the tool works but the problem wasn't as big as you thought, you keep it running and move on to the next candidate on your list.

Month three is mostly reflection and decision-making. It takes 2 to 4 hours of your leadership team's time.

What this really costs

Software: $3,000 to $8,000 for month two. Some months you spend nothing.

Your time: 10 to 15 hours total over three months. Most of that is in month one, when you're learning and deciding.

Risk: You spend $5,000 to $15,000 and find out AI doesn't fit your business. This happens. It's not common, but it's real. A business that runs smoothly on phone calls, relationships, and institutional knowledge might not benefit from AI. A business where every job is custom and unpredictable might find that AI tools are more trouble than they're worth.

The honest version: AI fits some businesses and not others. The only way to know which one you are is to try something small.

Where to start

• Write down three to five recurring problems your team solves the same way every week. Pick the one that wastes the most time or causes the most frustration. Spend one hour on this.

• Talk to someone who has built an AI tool in a business similar to yours. Ask them what worked, what didn't, and whether they'd do it again. Ask for a specific example, not a pitch.

• Schedule a conversation with your leadership team this week to decide whether you want to run a small test in month two. If yes, pick the problem and the person who will lead it. If no, that's a valid answer too.