AI Saves You Money. It Also Costs You Customers.

AI can help you take on more work without adding to payroll. It can also make it easier for your customers to do the work themselves.

That is the tradeoff in a recent PYMNTS Intelligence report on small and medium-sized businesses. Frequent AI users estimated that using AI saved them the equivalent of roughly 16% of annual revenue. They also estimated losing roughly 7% to customers who used AI instead of hiring them.

Those are meaningful estimates. They are not audited savings or guaranteed results. But they point to a practical question for any small business owner: where does AI make your business more useful, and where might it make your service easier to skip?

AI can add capacity without adding payroll

For a small business, an extra hour matters. So does the ability to take on a task without adding another recurring expense.

AI tools can help with parts of everyday work, from drafting marketing copy to organizing information or preparing a first pass at a document. Used with care, they can help you move routine tasks along and free up time for work that needs your knowledge of the business, your customers, or the situation at hand.

That is one way to grow without hiring. Not by making every task automatic, but by using tools to extend the capacity you already have.

The PYMNTS report suggests that some frequent AI users see a sizable financial benefit. Their estimate of savings, roughly 16% of annual revenue, covers work they otherwise might have paid someone else to do. It is a survey finding, not a forecast for your business. Your results will depend on which tasks you use AI for, what the tools cost, and how much checking or rework they require.

A useful starting point is to list the tasks that take time but do not require your particular judgment. Then test whether AI can help with one of them. Keep a human responsible for reviewing anything that goes to a customer or affects a business decision.

The same tool can make your service easier to replace

AI does not only change how you work. It can change how your customers work.

A customer who once hired you to draft, research, design, or organize something may now try to do part of it with an AI tool. The PYMNTS report found that frequent AI-using businesses estimated losing roughly 7% of annual revenue to customers using AI instead of hiring them.

That is not a claim that every AI user loses that share, or that every customer-facing service is at risk. It is a reported estimate from businesses surveyed. Still, it is worth asking which parts of your offer a customer could now attempt without you, and what they would miss by doing so.

The answer is not always "use AI too." Sometimes a service is valuable because you know what questions to ask, what information to trust, or what a good result needs to do. Sometimes a customer wants a finished outcome, not another tool to learn. And sometimes the customer needs a person to take responsibility for the work.

The distinction is between selling a task and delivering an outcome. If the offer is only "we produce a first draft," a customer may see AI as a substitute. If the offer includes direction, review, context, and accountability, the value is harder to reduce to a prompt.

Human work is finding its own demand

The report also points to a countercurrent. About one in four businesses surveyed gained customers who specifically wanted a person to do the work. And 14% picked up paid work fixing AI-generated output.

Those findings do not mean every customer is turning away from AI. They do show that human involvement can be part of the value customers seek, and that AI output can create work that needs a capable person to check and improve it.

This is not a case for rejecting AI. It is a case for being clear about where your judgment matters. An AI-generated draft may be quick, but someone still needs to decide whether it is accurate, suitable for the audience, and ready to use. A generated design may look finished but miss the brief. A summary may leave out the detail that changes the decision.

For a small business, those gaps can become either a liability or a service. The difference is whether you have the skills and process to catch them.

A practical way to decide what to automate

You do not need a large AI strategy to begin. Start with a bounded test.

1. Choose one repeatable task. Pick work with a clear input and a result you can judge. Do not begin with a high-stakes decision.
2. Set a human review point. Decide who checks the output and what they check for. If no one can verify it, do not treat it as ready.
3. Count the whole cost. Include the tool, setup time, review, corrections, and any customer communication the process creates.
4. Watch for customer effects. Note whether the tool helps you serve customers better, or simply makes it easier for them to bypass your service.
5. Keep what works. If the test saves time without lowering the quality or trust customers rely on, make it part of the process. If it creates more review than value, stop or change the task.

This keeps AI adoption connected to actual work rather than to the pressure to adopt a tool because it is new.

Keep the person accountable

The central business question is not whether AI can produce something. It is whether the result is useful, correct, and worth paying for, and who stands behind it.

TresPies takes a human-led approach to research, engineering, and design. That means using judgment to understand the problem, building tools and systems with a clear purpose, and making the handoff and responsibilities explicit. AI can support parts of that work. It does not replace the need to know what the work is for, check what it produces, and answer for what gets delivered.

If you are weighing an AI feature or process, our practical guide to AI integration lays out questions to ask before putting a tool into an existing app. TresPies also lists an AI governance audit among its engineering services.

The PYMNTS findings are a useful prompt, not a promise of what will happen in your business. Frequent users reported both savings and customer losses. Some businesses also found customers who specifically wanted human work, and paid work correcting AI output.

Use AI where it helps you serve people well. Keep people responsible for the parts that require judgment. And make sure customers can see the difference between an automated first pass and work that someone has checked and is prepared to stand behind.

Source: PYMNTS Intelligence, "AI Goes to Work: How SMBs Grow Without Growing Payroll". The figures above are survey estimates reported by frequent AI-using small and medium-sized businesses, not audited business results.