Economists are split on what AI means for software jobs

AI is reshaping software development skills and roles

AI will change software jobs even if it does not eliminate them. That distinction matters. A company could have roughly the same number of technology employees two years from now while requiring those employees to do substantially different work.

Software developers increasingly need to work with AI-generated code, evaluate its quality, identify errors, and decide where automation makes business sense. Coding ability remains important, but it is becoming part of a broader skill set. The developers who create the most value will increasingly understand how to combine engineering judgment with AI capabilities.

This changes the talent equation for executives. Hiring plans based on yesterday’s job descriptions can quickly become outdated. Companies need to identify which capabilities will become more important, which tasks AI can absorb, and where human expertise remains critical. Recruitment, training, career development, and performance measures may all need to change together.

The nuance for C-suite leaders is that headcount alone is becoming a weak measure of AI’s workforce impact. The more useful question is what work employees perform and how much value they can produce with AI. Leaders should track changes in skills, productivity, quality, and role design rather than assuming AI adoption must translate directly into job cuts.

The opportunity is significant. Companies that retrain capable employees can preserve institutional knowledge while building the skills needed for AI-enabled development. But this requires deliberate investment. AI transformation changes how technical work is organized, measured, and managed.

Automated coding tools are evolving toward greater autonomy

AI coding tools have progressed quickly. Early products mainly suggested individual lines or small blocks of code. More advanced systems can now generate larger amounts of software and handle increasingly complex engineering tasks with less human input.

This development changes the economics of software production. Developers can delegate more implementation work to AI and spend more time defining requirements, reviewing output, testing systems, solving difficult problems, and making architectural decisions. The result is a different development process in which people increasingly supervise and direct automated systems.

Companies are already adjusting their approach to talent. Employers are increasing demand for programmers with AI skills. This suggests that organizations currently see value in combining experienced developers with AI rather than treating automation as a simple replacement for engineering teams.

Executives should also distinguish technical capability from production readiness. An AI system that can generate complex code still needs controls around security, reliability, intellectual property, regulatory requirements, and software quality. Generated code can contain vulnerabilities, incorrect assumptions, or errors that are difficult to detect. Greater autonomy therefore increases the importance of effective human review and automated testing.

Governance becomes particularly important as these systems gain permission to modify code, use internal information, or interact with production environments. Companies need clear rules governing what AI agents can access, which actions require human approval, and who remains accountable when generated software fails.

The strategic direction is clear even if the final outcome is not. Coding is becoming more automated, and the boundary between developer tools and systems capable of completing development tasks is moving rapidly. C-suite leaders do not need to predict exactly where that boundary will settle. They do need an operating model that can adapt as the technology improves.

Demand for AI-capable developers is growing much faster than traditional developer roles

The software labor market is changing quickly. Companies still need developers, but they increasingly want people who can combine traditional engineering skills with AI. That shift is already visible in hiring data.

According to a June report from Randstad Digital, AI-augmented software development roles grew nearly 600% over five years. Traditional developer roles grew 28% during the same period. The difference is substantial. It suggests that employers are not simply reducing demand for software talent. They are changing the type of software talent they value.

For executives, this has implications beyond recruitment. AI skills will increasingly influence workforce planning, compensation, internal mobility, and training budgets. Companies that rely only on external hiring could face stronger competition for experienced AI talent. Developing existing engineers can provide another route, particularly when those employees already understand the company’s systems, customers, security requirements, and business processes.

AI capabilities are becoming more valuable within software development. CEOs, CIOs, and CTOs should examine which capabilities they will need over the next several years and determine whether to hire, retrain, or combine both approaches.

This also changes what leaders should expect from technical teams. Knowing how to use an AI coding tool is not enough. Valuable employees need to understand when AI output is useful, when it is unreliable, and how to verify the resulting software. Strong engineering fundamentals therefore remain important as AI skills become more common.

AI coding tools save time, but developers spend much of it reviewing AI output

Generating code faster is only part of the productivity equation. AI can reduce the amount of time developers spend writing code, but the resulting software still has to be checked. That verification work can consume much of the time automation initially saves.

A survey from software platform provider Harness found that more than four in five developers said most of the time saved on coding is now spent reviewing output from AI coding tools. This is an important result for executives evaluating the financial return from AI-assisted development.

The finding does not mean AI coding tools provide no productivity benefit. Review is already an essential part of professional software development, and AI may allow teams to produce and evaluate more code within the same period. Productivity can also appear elsewhere: faster experimentation, shorter development cycles, fewer repetitive tasks, or greater capacity to address existing work.

But leaders should avoid treating code generation speed as a complete measure of success. If developers generate software faster but spend substantially more time checking incorrect, insecure, or unsuitable output, the net improvement can be smaller than expected. Measures such as delivery time, defect rates, security issues, rework, system reliability, and developer capacity can provide a more complete view.

There is also a workforce implication. As AI takes responsibility for more initial code generation, developers need stronger review and verification skills. They must be capable of detecting subtle problems, understanding generated code, and deciding whether it meets business and technical requirements. Human judgment therefore remains important even as the amount of directly written code declines.

For C-suite leaders, the objective should be measurable business improvement rather than maximum AI usage. Companies should establish a productivity baseline before broad deployment and compare results after adoption. If AI reduces delivery times while maintaining or improving quality and security, it is creating operational value. If review and rework absorb most of the gains, leaders need to improve the tools, processes, training, or use cases before expanding investment.

AI adoption is increasing the cost of software development

AI can accelerate software development, but the financial equation extends well beyond the price of an AI coding tool. AI adoption can increase spending on development tools, data, computing infrastructure, and the systems required to operate AI at scale.

These costs can rise quickly as usage expands. More developers using AI means more requests to AI models, greater computing demand, additional data processing, and potentially higher cloud and infrastructure expenses. Enterprises may also incur costs for security controls, system integration, monitoring, governance, and reviewing AI-generated work. A relatively inexpensive experiment can therefore have a very different cost profile when deployed across a large organization.

This makes AI economics a C-suite issue rather than simply an IT procurement decision. CIOs and CFOs need to understand the total cost of adoption and compare it with measurable improvements in software delivery. Faster code generation has limited financial value if additional infrastructure, human review, or remediation consumes the resulting gains.

Executives should consequently evaluate AI development investments using business outcomes rather than adoption rates. Useful measures include development cost per project, time to production, developer capacity, defect and rework rates, infrastructure spending, and the cost of operating each AI-enabled workflow. These measures can reveal which applications generate sustainable returns and which mainly increase technology spending.

There is an important nuance. Rising AI expenditure is not automatically evidence of poor returns. Higher costs may be justified when AI helps a company release products faster, improve software quality, increase engineering capacity, or create new revenue. The key is to connect spending with outcomes and determine whether those benefits persist as deployment expands.

CIOs need to make upskilling and reskilling a core part of AI transformation

AI transformation changes people as much as processes. As development tools become more capable, IT employees need new skills to operate, supervise, and evaluate them. Training should therefore become a priority for CIOs managing the transition.

Upskilling involves improving capabilities employees need in their current roles, while reskilling prepares them to perform substantially different work. Both matter in an AI-enabled technology organization. Developers may need stronger skills in AI-assisted coding, output validation, security, testing, and system design. Managers may need to learn how to measure productivity when human and AI contributions are increasingly combined.

Ullrich, whose position and company are not specified in the provided excerpt, directly emphasizes this workforce challenge: “Having employees who feel like they have the ability to upskill and reskill is going to be really important, because there’s likely to be a lot of stress amongst employees as this change is happening.”

For executives, training should be connected to actual workforce requirements rather than offered as a broad collection of AI courses. Leaders can identify roles that AI is changing, define the skills those roles will require, measure current capability gaps, and direct training toward those gaps. Employees should also have opportunities to apply new skills to real workflows so the organization can determine whether training improves performance.

Workforce planning deserves similar attention. Some tasks will become increasingly automated, while others will become more important. Organizations may need fewer hours devoted to routine code creation and more expertise in reviewing AI output, software architecture, cybersecurity, data governance, and complex problem-solving. That transition can create opportunities for internal mobility if employees receive relevant preparation.

The nuance for C-suite leaders is that training alone will not solve every workforce challenge. AI adoption can alter organizational structures, career paths, hiring requirements, and the number of people required for particular activities. Leaders need to communicate what is changing while remaining clear about areas where the outcome is still uncertain.

Done effectively, reskilling can help companies retain valuable institutional knowledge while developing new technical capabilities. The objective is not simply to teach employees how to use the latest AI product. It is to build a workforce capable of adapting as the technology, business requirements, and software-development process continue to change.

Key executive takeaways

• Skills are changing faster than headcount: AI may transform software roles without eliminating them. Leaders should update hiring, performance metrics, and workforce plans around AI-enabled skills and adaptability.
• Coding automation is becoming more capable: AI tools are moving from simple code suggestions toward handling complex engineering tasks. Executives should strengthen human oversight, testing, security, and governance as autonomy increases.
• AI-skilled developers are gaining value: Randstad Digital reports that AI-augmented software development roles grew nearly 600% over five years, versus 28% for traditional developer roles. Companies should combine targeted hiring with internal skills development.
• Productivity gains require effective review: More than four in five developers surveyed by Harness spend most of their AI-enabled coding time savings reviewing generated output. Measure AI performance through delivery speed, quality, rework, and security rather than code generation alone.
• AI economics need closer scrutiny: Development tools can add data, infrastructure, governance, and review costs as adoption scales. CIOs and CFOs should connect total AI spending to measurable productivity, quality, and revenue outcomes.
• Reskilling should become a workforce priority: As AI changes technical responsibilities, employees need practical opportunities to develop relevant capabilities. CIOs should align training with emerging roles, internal mobility, and measurable business needs.