AI is altering the economics of software development, but who gets a cut?

As software vendors and agencies boost their productivity with AI, the enterprises contracting their services have started to ask what’s in it for them.

According to some studies, developers have increased their output by up to 40% using AI coding assistants. That increased productivity should lead to new benefits for customers of for-hire developers and SaaS providers, some IT leaders contend.

But while the benefits of AI productivity gains for companies employing their own programmers are clear — lower costs and more code shipped — CIOs shouldn’t hold their breath waiting for price drops from outside software engineering firms.

Still, as the advantages for customers begin to take shape, IT leaders should start to demand more from their software vendors, including faster and more flexible development cycles, experts recommend.

“In many cases, AI lowers the cost of producing routine code, documentation, tests, and internal tooling,” says Yoni Michael, CTO and cofounder at Typedef, an AI startup focused on transitioning prototypes into deployments. “That should benefit customers, but the most important customer benefit does not have to be an immediate price cut.”

Other benefits can include faster delivery, better support, better reliability, more frequent product improvement, or more tailored functionality at the same spending level, he suggests.

“From a CIO perspective, the key issue is whether vendors are using AI to create customer-visible value or simply to expand their own margins,” he says. “If a vendor claims major internal productivity gains, enterprise buyers should ask where that shows up.”

Customers of software development firms should gain more leverage in negotiations over time, particularly when AI leads to a significant reduction in the labor required to maintain or improve the product, Michael says.

“But that leverage is likely to show up first in commercial flexibility, bundling, faster services delivery, and outcome-based commitments before it shows up as across-the-board lower subscription prices,” he warns.

Quality before cost cuts

While AI assistants are increasing developer productivity, one downside is the quality of the code, including security flaws, says George Manuelian, chief strategist at software security vendor RapidFort. Because of this, he suggests IT leaders focus first on software quality before price cuts.

“We’re seeing teams ship AI-generated code that pulls in outdated or vulnerable open-source libraries, and code that works functionally, but wouldn’t pass a proper security review,” he says. “While productivity is clearly up, the volume of unverified code entering production is also rising. In many cases, speed is improving faster than the processes needed to validate what’s being shipped.”

Customers should demand quality as a dividend for AI-assisted development, Manuelian suggests.

“Customers should expect better software, not just faster development behind the scenes,” he says. “That means fewer defects, faster turnaround on feature requests, and stronger security built into the product.”

For example, if a vendor can now deliver a feature in two weeks instead of six, the benefit should be faster roadmap delivery or more frequent product improvements for the customer, Manuelian adds.

“The problem is that in many cases, those gains aren’t visible externally,” he says. “Customers are still dealing with the same release quality and support timelines. If the experience isn’t improving, then the benefits of AI aren’t really being shared.”

Customers can push for more value, including faster delivery on features, stronger SLAs, or additional functionality, Manuelian recommends.

Can’t save if you don’t ask

Customers of software development services using AI coding assistants should absolutely ask to share in the benefits when software development times for standard features have dropped by 30% to 40%, agrees Jayanand Sagar, cofounder and COO at AI data aggregation vendor Hyperbola Network.

“Software vendors are having leaner engineering teams today,” he says. “That productivity increase goes straight to margin unless the customer demands something in exchange. Most buyers never ask.”

Many vendors aren’t disclosing their AI use during development, and they are transparent only when customers demand the information up front, Sagar says.

“If a vendor of a software as a service has shipped three major releases in the last 12 months using AI tooling, that pace of delivery is a data point most CIOs are leaving on the table,” he adds.

While price cuts are an obvious ask, other changes may be more valuable over the long run, he says.

“Feature timelines, bug resolution windows, and roadmap access have real dollar value in operations,” Sagar adds. “A value on faster delivery cycles is considerably compounded over a three-year term.”

Informed negotiations

Sagar recommends that software buyers pull their top three vendor contracts and ask for a summary of their delivery performance over the past year before the next renewal discussion.

“If the rate of features being released has sped up and the invoice has not changed, then that is your opening,” he says. “Buyers that win this negotiation come having use data, delivery history and a specific number for what more rapid output is worth.”

Some clients are getting the message. Some customers of software development firm ProSense.Digital have begun to ask for 20% to 30% price cuts on the assumption that Claude Code is doing the heavy lifting, says Asia Solnyshkina, founder and CEO there.

But these demands misunderstand the value that software development firms bring, she says.

“AI is absolutely changing software development economics, but not in the way some clients think,” she adds. “Code generation was never the main source of value. Delivering business value is the harder and more important part.”

While AI can assist with creating business value, human oversight is still essential, Solnyshkina says. “Our clients are not paying for raw code alone; they are paying for judgment, architecture, complex integrations, and accountability,” she adds. “You cannot solve all of that with a prompt.”

In addition, AI isn’t free for development firms, Solnyshkina adds, with both direct and indirect costs, including tooling, tokens, training, process redesign, and quality control. AI shouldn’t enable automatic cost cuts, but better value delivery, including faster cycles and better decision-making, she adds.

“So no, we are not cutting prices simply because Claude Code exists,” she says. “We are using these tools to deliver more value for the same investment.”