Luboslava Uram is COO and CTO at Solvd Group, a subsidiary of Allianz Group. Transforming the claim management experience
Every week seems to bring another headline about AI transforming software development. On Forbes, I recently read the article “AI Coding Power Users Are Churning Out 46X More Code Than The Rest,” while TechCrunch published an article with the headline “Coders are refusing to work without AI — and that could come back to bite them.”
These headlines create a simple boardroom narrative: AI coding tools are making developers more productive; we can ship code faster, bugs can be seen earlier, and routine work automates itself.
So the question from the CEO follows quickly: “If AI speeds up our engineers, can we run with fewer of them and spend less on technology?”
The Highway Analogy That Changed How I Think About AI Productivity
Thinking about how AI was changing the work of my own engineering teams, I came across the idea of induced demand. The analogy changed how I looked at developer productivity. When a city widens a congested highway, the reasonable expectation is less traffic. What usually happens is the opposite: The extra capacity invites more trips, and within a few years, the road is just as packed and now carrying more cars. Economists call this induced demand.
I increasingly see software development behaving in the same way. Make delivery faster and cheaper per unit, and the organization doesn’t collect the difference as savings; it consumes more delivery capacity via induced demand. Faster coding does not empty the backlog. It brings forward work that had been postponed and makes previously uneconomic ideas look achievable.
Atlassian describes developers using AI-enabled capacity for more concurrent projects and broader work. McKinsey similarly argues that higher engineering productivity allows companies to build more products, modernize more systems and automate more workflows. What I take from this is that AI often expands the amount of work an organization can justify. It does not automatically shrink the existing cost base.
What Changes Inside A Development Organization
In conversations with engineering leaders and partners, I hear the same pattern repeatedly.
The first effect is visible quickly. Developers move through routine and familiar tasks faster. AI can assist with testing, generate documentation, explain unfamiliar code and reduce the time spent searching for solutions. These gains are real. What becomes visible later is that the work does not disappear. It moves.
As more code is produced, teams face more decisions about architecture, integration, security and long-term maintainability. Faster releases also create additional operational responsibility and dependencies that may need to be managed for years.
At the same time, the business notices the increased speed. Delivery expectations rise, previously deferred ideas return to the road map, and the faster pace quickly becomes the new baseline. This was one of the most important realizations for me: AI can make development teams more productive and more overloaded at the same time.
In one discussion, I saw a team demonstrate how AI could shorten part of the development cycle. The immediate reaction was not to reduce the road map. It was to ask which additional features could now be brought forward. That moment captured the issue for me. The productivity gain was real, but it had already been allocated before anyone had discussed whether it should become growth, quality improvement or cost reduction.
In my experience, converting these gains into financial impact requires deliberate changes to the operating model, incentives and portfolio governance. Without those choices, impressive productivity figures can coexist with an unchanged technology cost base.
Three Decisions You Still Have To Make
I now frame the discussion around three decisions.:
• First, decide where the capacity should go. Use the new speed to accelerate the road map, improve quality or modernize legacy systems faster. These choices create value, but they do not create savings and should not be presented internally as if they do.
• Second, decide whether faster delivery becomes the new baseline. Once faster delivery becomes simply “how we work,” the gain is absorbed into expectations instead of remaining visible as a measurable outcome.
• Third, decide what must change structurally. Cost savings are possible, but only alongside deliberate changes to team structures, skill mixes and portfolio scope. This requires careful planning and honest communication, not a spreadsheet exercise in subtracting engineers.
Framing AI as a headcount-reduction tool tends to produce the worst outcomes. AI can accelerate implementation, but it cannot take responsibility for architectural quality, security or long-term maintainability. It accelerates decisions while amplifying the consequences of the architectural and product choices behind them. In most environments, that makes experienced engineers more valuable. Cutting capacity before redesigning the work can allow complexity and risk to rise faster than costs fall.
Before You Put The Savings Into The Plan
Before AI savings enter a financial plan, I would ask three questions. Where is the released capacity expected to go? Which work will the organization consciously stop doing? And what will change in the operating model so that productivity can affect the cost base?
If there are no clear answers to these questions, the productivity may be real while the savings remain hypothetical.
AI changes the economics of software development, but not in a straight line. Speed may improve almost immediately. Value rises when the released capacity is directed toward the right work. Costs change only when leaders make deliberate choices about scope, priorities and organizational capacity.
Without those choices, the gain is absorbed into growth, broader scope and higher expectations.
The economic impact of AI coding tools will not be determined by how much code they generate. It will be determined by what leaders choose to do with the capacity they release.
The question I now ask is not simply, “How much faster are our developers?” It is, “What have we consciously decided to do with the time AI has given back?”
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
