Skip to main content
French Press Labs
Journal

Cost and timelines · 6 min read

AI pays off on some tasks and wastes time on others. Here is a simple way to tell which is which before you spend anything.

French Press Labs · Updated

The short answer

AI is worth it for a small business when it takes over a task that already costs you real hours or lost leads every week, and the setup and upkeep are smaller than that cost. It is not worth it yet when the task is rare, the process is unclear, or nobody will look after it.

In this article ›

Why is "is AI worth it" the wrong question?

Because AI is not one purchase. It is a way of doing specific tasks. Asking whether AI is worth it is like asking whether vehicles are worth it. A work van for a plumber, yes. A second van that sits in the yard, no.

The useful question is narrower: is AI worth it for this task, in my business, right now? That question has an answer you can work out on paper.

How do you count what a task costs you now?

Pick one task and watch it for two weeks. Do not estimate from memory. Memory flatters busy people. Write down three things.

Hours. How long does the task take each time, and how often does it happen? Include the small interruptions. A front desk person at a dental office who stops to confirm appointments by phone loses more than the call time. They lose the thread of whatever they were doing.

Missed leads. How many calls, messages or form fills went unanswered or were answered too late? An auto repair shop that returns voicemails at the end of the day is often calling people who already booked elsewhere.

Errors and rework. How often does the task go wrong? A double-booked chair, a quote with the wrong address, an invoice that was never sent.

You now have a real picture: this task takes this many hours a week, loses roughly this many inquiries, and causes this much rework. You know what an hour of that person's time is worth to you, and what a typical job is worth. Multiply it out privately. That is what the task costs today.

Our manual work check walks through this kind of count if you want a structure for it.

What do you compare that against?

Three things sit on the other side of the scale.

  • Setup effort. Someone has to choose the tool, write the rules, connect it to your calendar or job software, and test it. This is real work even when the tool calls itself easy.
  • Ongoing cost. Subscriptions, usage fees, and any outside help. We break these down in what AI costs a small business.
  • Upkeep. Somebody has to read the logs, fix the answers that were wrong, and update the rules when your hours, services or staff change.

Be honest about upkeep. It is the part most people leave out. A tool that saves five hours a week and needs one hour of attention saves four. That can still be a very good trade. It is just not the trade on the sales page.

What does the comparison look like for common tasks?

TaskTends to be worth it whenTends not to be worth it when
Answering calls and texts after hoursYou miss inquiries most evenings and weekendsCalls are rare or always complex
Following up with quiet leadsQuotes go out and nobody chases themYou send only a few quotes a month
Appointment remindersNo-shows are a regular problemClients almost always show up
Drafting replies to reviewsReviews pile up unansweredYou get few and enjoy writing them
Summarizing sales or job dataYou want the same numbers weeklyYour records are not in any system yet
Writing estimatesJobs are similar and priced by ruleEvery job is custom and needs a site visit

The left column has a pattern: the task is frequent, it follows a recognizable shape, and its cost today is visible. The right column is the reverse.

When is AI not worth it yet?

Some honest cases where the answer is no, or not now:

  • The task is rare. Automating something that happens twice a month costs more in setup than it returns.
  • The process is unclear. If two people on your team would handle the same call differently, an AI tool has no rule to follow. Agree on the process first.
  • The information is not written down. Prices in the owner's head, service areas nobody has listed, exceptions that live in memory. The tool will guess, and guess wrong.
  • Mistakes are very costly. Anything involving safety, medical guidance, legal commitments or large sums needs a person. AI can prepare a draft there, but should not act alone.
  • Nobody will own it. A system with no owner drifts. Within months it is quoting last year's hours.
  • Your customers want a person. Some businesses are built on a familiar voice. Protect that.

Saying "not yet" is a fine result. It usually points to something worth fixing anyway, like writing down your prices or agreeing on how calls should be handled.

How do you run a small test before committing?

Treat the first project as a trial with a clear end.

  1. Choose one task from your count, ideally the one with the highest cost and the simplest pattern.
  2. Set a baseline. Use the numbers from your two weeks of watching.
  3. Start narrow. After-hours texts only, or follow-up on one type of quote only.
  4. Keep a person checking. Read every conversation or output for the first few weeks.
  5. Count again after a month using the same three measures: hours, missed leads, rework.
  6. Decide. Expand, adjust, or stop. All three are acceptable outcomes.

A cleaning company might test automatic follow-up on quotes that got no reply. If more quotes turn into bookings and the messages read well, expand. If customers find the messages pushy, change the wording or stop.

What do the numbers miss?

Some benefits are hard to count but real. An owner who stops answering texts during dinner. A front desk that can look up and greet the person walking in. Staff who stop dreading a task everyone hated.

Some costs are hard to count too. A customer who got a wrong answer and never mentioned it. A team that feels watched or replaced. Staff who quietly stop trusting the system and go back to the old way while the subscription keeps running.

Put these on the scale as well. They are not decimals, but they are part of the decision.

Where should you start?

Count one task for two weeks. That alone tells you more than any article can. If you want a rough sense of what missed inquiries add up to, try the ROI calculator with your own figures. If the count shows a clear cost and a clear pattern, test one narrow piece.

If you would like a second pair of eyes on which task to test first, you can book a call with us. Bring your two weeks of notes.

Frequently asked questions

How long does it take to know if an AI tool is paying off?

For a task that happens daily, such as answering calls or following up with leads, a few weeks of honest tracking usually tells you. For tasks that happen monthly, you need several cycles. Decide what you will count before you start, so the answer is not a feeling.

Is AI worth it for a business with only a few employees?

It can be, because in a small team one person often carries the phone, the schedule and the follow-up at once. The test is the same at any size: a task that repeats often, follows a pattern, and currently costs hours or lost leads. Without that, it is probably not worth it yet.

What if I try a tool and it does not help?

Then you have learned something cheaply, provided you started small. Choose tools you can cancel, keep your old process available during the trial, and write down why it did not work. The reason is often the process or the data, not the tool.

Should I wait until AI gets better?

Waiting is reasonable for tasks where mistakes are costly and the tools are still unreliable. It is less reasonable for simple, repetitive tasks that already cost you time every day. Judge each task on its own instead of making one decision about AI as a whole.

Related

Published November 13, 2026 · Updated November 13, 2026

Every fix is scoped and priced in writing after a free 20-minute fit call.

CallBook a fit call