An adoption strategy that survives the first 90 days

Ninety days is enough time for a small business to build a new habit, and short enough that it’s easy to admit when something didn’t work. It isn’t enough time to transform the whole company, so don’t try. Think of the quarter as three decisions instead: what to try, whether it worked, and whether anyone other than the person who championed it will keep using it.

Take stock first

Before you buy anything, write down the tools you already pay for, who uses them, and which workflows take up most of the week. You’re trying to see the difference between the software on your credit card statement and the software people use day to day.

Figure out who’s responsible for AI in your business right now. If the truthful answer is “the owner, after hours,” write that down. A plan that assumes a coordinator you don’t have will fall apart around week three, when it turns out the coordinator is also the person answering the phones.

Assess before you buy is the full worksheet for this step. The 30-day plan is a shorter version: one week to look around, one week to pick a workflow and write down how it currently runs, two weeks to run a shadow pilot, and then a decision.

Set a pilot window

A pilot window needs a start date, an end date, and a clear idea of what “done” means. For most small businesses, anywhere from one to six weeks is realistic. Daily work can be judged in a week or two. Weekly work needs longer, but it still needs a date on the calendar.

Success should be something you can observe, like “the intake draft matches the fields we already collect, and the person at the front desk is willing to use it.” Reasons to stop should be just as concrete. Customer data turns up somewhere your rules say it shouldn’t be. The reviewer ends up rewriting every draft. The owner is the only one who remembers the pilot is happening.

Put the pilot on a calendar the whole team can see. If it only exists in a side conversation, the regular week will crowd it out fast.

Prove it works before you roll it out

Proving it works means running a shadow pilot, which is different from telling everyone to start using the tool. The old process keeps doing the work, the new one produces a draft, and you compare the two. The shadow-mode guide walks through the details.

Rolling out comes after you’ve decided to keep it. Start with a few seats for the people who were in the pilot, write down who reviews what, and spend a week watching whether the new habit holds when you’re not around to remind people. After that, bring in more of the team.

Buying seats for the whole company on day one is a common mistake. Having the software available doesn’t mean anyone will change how they work.

Handle pushback

Pushback is usually specific. “It got the gate code wrong.” “I don’t want a customer hearing that voice.” “This adds a step to a job that was already running late.” Treat comments like these as useful feedback on the tool.

The position that tends to work on a small team is that AI helps with the work, and the person whose name is on the result still makes the call. Say that out loud, then back it up by keeping the review step in place. If you say it and then switch on automatic sending, your team will believe what you did over what you said.

You don’t need a big speech about the future of work. Run the pilot. People pick up tools that take care of a specific annoyance for them, and they mostly tune out all-hands meetings about change.

Measure what happened

Use the same measures as the build guides: time spent on the task, specific mistakes you can point to, and whether the person reviewing the drafts wants to keep getting them next month. Read what to measure before you put any number in a note to staff.

If your ninety-day review opens with a percentage nobody tracked, it won’t survive the first question from the person who did the work. Start with real examples from the pilot instead.

Decide whether to expand or stop

At the end of the pilot, and again around day 90 if you kept it, pick one of three options:

• Expand to the next role or the next workflow, one at a time, with the same review process.
• Hold at the current scope, because it works and doesn’t need to grow.
• Stop, because nobody’s owning it anymore or the risk you were worried about showed up.

Holding steady is perfectly fine, since not every workflow you keep needs to become something bigger. Stopping is fine too. Write down why you stopped, so the next vendor demo doesn’t restart the same debate from the beginning.

When to get outside help

Bring in help when you’ve run out of time or the work has become too complex to handle yourselves. Wanting someone to share your excitement isn’t a good enough reason. Signs you need help include having nobody who can own the output, a workflow that spans locations or systems you can’t keep an eye on, or a pilot that’s been revised so many times it’s lost its shape because nobody had the hours to finish designing it.

Managed AI is the option where a team runs the agents, tools, and memory alongside you. A Fractional CTO engagement is a short-term way in that leads toward Managed AI and LaunchpadOS. It isn’t meant to be the end goal, and it doesn’t replace having someone who runs the work day to day. Pricing and the structure of the engagement are on arcasya.ai. If you’d like someone to look at your ninety days with you, book a discovery call and bring your decision notes, including the pilots you stopped.