If AI Makes Your Work Cheaper, What Will You Sell More Of?

Suppose you run an accounting firm, and AI cuts the cost of closing a client’s books in half.

For a while, that looks like an excellent business. You can serve more clients without hiring as many people. You can lower prices and still make more money per account.

Then your competitors acquire similar capabilities.

What happens next?

You might keep some of the savings but your clients are going to capture most of them. That’s capitalism. But smart entrepreneurs are going to discover a different business hiding inside the one you already run.

That third possibility is where the discussion about AI and professional services gets interesting.

Rick Zullo at Equal Ventures recently made a useful argument in “Ephemeral Margins in AI-Native Services.” An early automation advantage can produce attractive margins, but those margins may disappear as competitors catch up and lower prices. He uses insurance-claims administration as an example: making claims cheaper to process doesn’t necessarily create more claims to process. It just lowers your revenue and margins.

That’s a challenge to the assumption that every dollar currently spent on labor becomes an equally large revenue opportunity for AI companies.

But it leaves an important question: how much valuable work isn’t being done because it costs too much?

Return to the accounting firm.

A client probably doesn’t need five versions of its monthly financial statements. It might, however, benefit from understanding profitability by customer, product, or location. It might want to know why a growing account is becoming less profitable, or which service lines are absorbing more staff time than their revenue justifies.

Perhaps that analysis was always possible. It simply wasn’t economical for the firm to provide it at a price the client would pay.

If AI changes that calculation, the firm has an opportunity beyond producing the same statements more cheaply. It can help the client make decisions that previously went unsupported.

The word “opportunity” matters. More analysis is not automatically more value. Someone must use it, and someone must be willing to pay for it.

Imagine a marketing agency that uses AI to produce emails, ads, and landing pages with much less labor.

It can offer more content for the same fee, or the same content for less. Neither proposition is necessarily a durable advantage if other agencies can do the same thing.

Now imagine the agency connects its campaign work to what happens in sales.

It learns which objections stop purchases. It discovers that a successful-looking ad attracts prospects who misunderstand the offer. It identifies the explanation that repeatedly helps a salesperson move a qualified prospect forward.

Those findings change the next campaign.

Marketing brings in better-qualified prospects. Sales conversations reveal what the marketing needs to explain. The two workflows improve each other.

The agency isn’t merely adding sales support to its list of services. It is making its existing marketing work more valuable.

Take that one step further. Connect acquisition information to the client’s actual customer economics.

One campaign produces inexpensive leads. Another produces more expensive leads. Judged only by cost per lead, the first looks better.

But what if its customers need extensive support, buy lower-margin services, or leave quickly? The more expensive campaign might be building the better business.

Access to reliable financial and delivery information could change whom the agency targets, what it promises, and how it evaluates success.

That is a much more consequential service than generating additional ad variations.

This is the part of the software-and-services combination I find most interesting.

A second workflow can create value through the work it performs. It can also create value through what it teaches the first workflow.

Sales conversations improve campaigns. Customer profitability improves targeting. Onboarding problems reveal promises that marketing should stop making.

Building Rustproof has made this question less abstract for me: when AI makes the work cheaper, do you use it to deliver the same service at a lower cost or take responsibility for something more valuable?

Connecting the work requires more than connecting the software.

The agency needs to recognize that a campaign lead, a sales opportunity, and a customer in the accounting system refer to the same business. It needs shared definitions of a qualified prospect, a successful sale, and a profitable customer. It needs permission to use the information and a way to check whether its conclusions are right.

That is where implementation work, domain expertise, and human judgment enter the picture.

Some of that work may become reusable software. Some may remain a valuable service. The test is whether it creates a better result and makes subsequent delivery more effective.

None of this is a new management insight. Good firms have always tried to connect marketing, sales, and customer economics. The AI opportunity is to make that coordination economical for more customers, with less manual work between each step.

There is an easy way to take this argument too far.

An agency could conclude that it should now handle marketing, sales, onboarding, finance, and operations. An AI startup could decide it needs to become the operating system for the entire industry.

But breadth does not create an advantage by itself.

If each new service requires different expertise, extensive customization, and another team to maintain it, the company may be assembling several difficult businesses rather than one improving system.

And more activity is not necessarily useful.

A small client may not have enough traffic to learn from dozens of marketing experiments. Its customers may not want more communication. Its biggest obstacle may be a weak offer or insufficient delivery capacity—neither of which is solved by producing more content.

You also don’t have to own a workflow to benefit from it. A marketing agency may need good accounting information without becoming an accounting firm.

The useful boundary is not “everything we could automate.” It is the set of activities that become meaningfully better when connected.

Even when the customer benefits, the provider still has to build a viable business.

The agency might create more value but struggle to charge for it. It might sell a broader engagement while underestimating the cost of integrations, review, and ongoing exceptions. Competitors might eventually offer a comparable result for less.

So the question is not just whether AI expands what a firm can do. It is whether the firm can deliver something customers value, get paid for it, and retain an advantage as the technology becomes widely available.

I would ask three questions of an existing professional-services firm or a new AI company:

AI can make an existing service cheaper. It can also make a previously uneconomical service possible.

Those are different opportunities. The first invites a competition over cost. The second gives you something new to prove.

When your work gets cheaper, the most important question may not be how much margin you can keep.

It may be what your customers can finally afford to have you do.