Before You Spend a Dollar on AI

1. The money that never hits your books

It's 7:40 on a Tuesday night. A homeowner's furnace quits. She calls your shop, gets voicemail, and calls the next company on the list. They pick up. By morning the job is booked with someone else.

That job never shows up anywhere. Not in your sales report, not in your lost-bid list, not on your P&L. A customer who never reached you leaves no record. That's why these losses are so easy to live with for years.

Most well-run businesses have a few of these quiet leaks:

• Calls and messages that go unanswered after hours, at lunch, or when the office is slammed.
• Estimates that were sent and never followed up, so they went cold.
• Work that was done but never fully billed, such as extra parts, added time or change orders nobody wrote down.
• Invoices that sit past 60 days because chasing them is nobody's job.
• Good customers who drift away because no one reminded them it was time for service.

None of these means you run your business badly. They happen because there are only so many hours in the day, and every one of them depends on someone remembering to do something.

2. What the big competitors have, and what you can now get

Small businesses employ about 46% of America's private-sector workers. [1] Many of them compete against regional and national operators, including roll-ups backed by private equity. It's worth being honest about what those competitors bring:

• Data across many locations and thousands of jobs, so they can see patterns in demand, pricing and performance that a single shop can't.
• Purchasing power and better information on what parts, equipment and supplies should cost.
• Access to capital for trucks, technology, acquisitions and marketing.
• Market intelligence: people whose job is to watch competitors, prices and local demand.
• Staff and systems for the routine work: someone to answer every call, follow up every estimate, chase every invoice and watch the numbers.

Some of these you won't match by yourself. You're not going to out-buy a national distributor's biggest customer. But others are no longer limited to large companies. Today's software and AI tools can provide the follow-through, the use of your own operating data, a clearer picture of your customers, and regular checks on your local market, at a cost a business your size can justify.

Meanwhile, you keep the advantages they struggle to copy: you know your customers, crews and community personally, and you can decide and act the same day. The goal isn't to become a bigger company. It's to add the specific capabilities that pay off at your size, without giving up the judgment and relationships that make your business worth choosing.

Most businesses haven't made that move yet. A 2026 U.S. Census Bureau analysis found roughly 17% to 20% of firms reported using AI in late 2025 and early 2026, with use much higher among larger firms. [2] That gap is real. But using AI and making money from it are not the same thing, and that's where most owners should be skeptical.

3. Find your leaks in 30 days, before you buy anything

You don't need new software to find out where you're losing money. Most of the evidence is already in your phone system, your scheduling or field-service software, and your accounting system. Spend a month counting:

Then put a rough dollar figure on each one. Use what a typical job actually earns you after parts and labor, not the ticket price. Now you know which leak is worth looking at first, and you have a "before" number to judge any fix against.

Be clear about what a 30-day count can and can't tell you. It shows how many calls were missed. It can't show how many of those callers would have booked a profitable job. Keep three kinds of numbers separate:

Only verified results should be used to decide whether to keep paying for something, and a fair comparison should test and reduce plausible alternative explanations. Comparing with the same weeks last year is a start, but weather, prices, staffing and customer mix may have changed too. Stronger tests switch the change on and off in alternating weeks, or compare similar crews or locations with and without it.

4. What the research shows, and what it doesn't

There's real evidence that AI can make people and businesses more productive. There's also a lot of hype. Here's what the stronger studies actually found:

• Customer support. A study of 5,172 support agents at a large software company, published in The Quarterly Journal of Economics, found an AI assistant raised issues resolved per hour by about 15% on average. Newer, less-experienced agents gained the most. [3] In other words, AI can help a whole team work more like your best people.
• Office work. A field experiment across 66 companies and 7,137 workers found that employees who used an AI tool built into their email and documents spent about two fewer hours a week on email. But the researchers saw no change in what work people actually did. [4] Saving time didn't change the business on its own.
• Home services. McKinsey's 2026 look at U.S. home services points to scheduling, customer retention, pricing, purchasing and back-office efficiency as the main places to find value. [5] Those are estimates that assume things go well, not guaranteed savings.

5. Run the numbers honestly

Here's how to judge whether fixing a leak could pay. Take a hypothetical home-services company with $3.0 million in sales and $300,000 a year in operating earnings, or EBITDA (earnings before interest, taxes, depreciation and amortization). It fixes two leaks, missed calls and cold estimates, and tightens up billing and collections. Every figure below is an estimated opportunity, not a verified result. We show a low, middle and high case, because no honest forecast is a single number.

Two rules keep these numbers honest:

• Collecting faster improves cash, not earnings. Getting paid in 30 days instead of 75 puts money in the bank sooner, but you were going to earn that money anyway. It only adds to earnings when an invoice would otherwise have gone unpaid. That's why faster collections sit on their own line, outside the earnings total.
• Each job is counted once. A job booked because a call was answered goes in the first line. An estimate that closed because of a follow-up goes in the second. Billing improvements only count work that was already done, never the new jobs above. If one job could fit two lines, it goes in one.

In the base case, EBITDA rises by about $21,120 a year. To earn that much more by selling more, at a 10% margin you'd need roughly $211,200 in new sales. In the high case, it's like adding about $723,200 in sales without adding a truck.

The cost line is the first year's all-in cost, including one-time setup, so later years should cost less. Even so, the low case loses $7,160. That's not a reason to skip the idea. It's the reason to start with one leak, measure it, and stop quickly if the numbers don't show up.

6. You stay in charge

Your business runs on judgment that doesn't fit in a software field: which customers need to hear a real voice, which tech to send on the tricky job, when a small concession builds loyalty and when it just invites more discount requests. Good systems protect that judgment. They don't override it.

The right split is simple:

• Routine, low-risk work you've approved gets handled and checked automatically: confirming appointments, sending reminders, following up an estimate.
• Anything unusual or high-stakes comes to you, with just the facts you need to decide quickly.
• Upset or emotional customers go to a person, even when software could technically answer.

Your data is a business asset, too. Federal guidance sets out the basics: the FTC recommends that businesses keep only the sensitive information they need and limit who can reach it, and the NIST framework for managing AI risk calls for clear human oversight, defined roles and ongoing checks. [6] [7]

NewellOI adds commitments of its own on top of that guidance. You should always know what information a system can see, what it did and why, and how to switch off its access. And your data and workflows are yours: you should be able to take them with you if you change vendors.

There's a longer-term payoff here as well. Every routine you document and every decision you capture is knowledge that no longer lives only in your head. That makes the business less dependent on you day to day, and more valuable if you ever decide to sell it.

7. Start with one leak

The fastest way to a real result isn't a big rollout. It's one problem, fixed and measured. NewellOI works in three steps:

Define

Pick one leak from your 30-day count, with a clear "before" number. Decide what can be handled automatically, what needs a manager, and what always comes to you.

Deploy

Use the tools you already have wherever possible, and add only what's missing. Keep a manual backup, clear permissions and the ability to switch it off at any time.

Operate

Measure against the "before" number and a fair comparison, so the result is verified rather than assumed. Track the results that matter: jobs won, errors, customer experience, profit and cash. Fix what's not working. Expand only when the numbers justify it.

Where to start depends on your trade. An HVAC company might begin with after-hours calls and maintenance-plan renewals. An auto shop might start with estimates waiting for approval. A pest-control company might focus on renewals. A roofer, with fewer repeat customers, might care most about responding to leads, tracking referrals and warranty follow-up.

One leak is where it starts, not where it ends

Fixing a first leak earns its keep on its own. But it also starts something bigger. Every call answered, estimate followed up and invoice sent adds to a shared, owner-controlled record of your customers, your decisions and your results. Over time, that record becomes an operating capability that keeps improving: knowing which customers are due, at risk or worth extra attention; a short regular briefing that brings you only the things that need your judgment; a clearer view of local competitors, prices and demand; routines that are measured instead of remembered; and the sense to step in, or to bring you in, at the right moment. That's the capability large competitors pay whole departments for, built one proven piece at a time.

8. Eight questions to ask any AI vendor, including us

1. Which specific problem in my business are we fixing?
2. What is it costing me today, and how do we know?
3. What part of it can today's technology actually handle?
4. What result will tell us it worked, in dollars and not activity?
5. Who approves exceptions, and what always comes to me?
6. How will mistakes be caught and fixed?
7. What is the full yearly cost, including support and my team's time?
8. If I leave, do my data and workflows come with me?

If a vendor can't answer all eight clearly, keep your wallet closed.

Sources

Sources accessed November 2026. Figures and findings keep the scope and limits set by their publishers. About these examples: the company in Section 5 is hypothetical, and its numbers are assumptions you should replace with your own. Third-party research is not a NewellOI customer result, and nothing here is a forecast or guarantee of performance.

[1] U.S. Small Business Administration, Office of Advocacy (Feb. 2026). Frequently Asked Questions About Small Business 2026. SBA Advocacy FAQ (advocacy.sba.gov)

[2] U.S. Census Bureau (May 2026). Large Firms With at Least 20 Employees Biggest AI Users. Census Bureau story (census.gov)

[3] Brynjolfsson, E., D. Li & L. R. Raymond (2025). Generative AI at Work. Quarterly Journal of Economics 140(2), 889-942. Journal article (doi.org)

[4] Dillon, E. et al. (2025). Shifting Work Patterns with Generative AI. NBER Working Paper 33795. NBER working paper (nber.org)

[5] McKinsey & Company (Feb. 25, 2026). Value Plays in US Home Services: Where Opportunity Meets Reliability. McKinsey article (mckinsey.com)

[6] National Institute of Standards and Technology (2023). AI Risk Management Framework 1.0: Core. NIST AI RMF Core (nist.gov)

[7] U.S. Federal Trade Commission. Protecting Personal Information: A Guide for Business. FTC business guide (ftc.gov)