Nikola Kapraljević, CEO of Infinum, delivered the opening keynote at .debug 2026 – Croatia’s largest technology and developer conference – with two questions that cut to the core of the industry: why do clients work with agencies at all, and what happens now that AI is here to stay?
“Two reasons why agencies exist,” he said. “Talent and flexibility.”
Talent because enterprise companies building complex technology need access to specialised skills they can’t always hire for. Agencies bring hiring infrastructure, training programmes, shared knowledge, and constant upskilling built up over years of delivery.
Flexibility because, within large organisations, headcount is political, layoffs are expensive, and hiring freezes are real. Agencies absorb that volatility. Whether a client needs a full-time embedded team, fractional support, or a focused one-off build, the model scales with their business’ needs.
Infinum has been doing exactly that for 21 years.
We operate across 8 offices with a global team of nearly 400 people, with four acquisitions expanding our impact worldwide – adding brand and creative, cybersecurity, and a stronger foothold in global markets.
Those fundamentals haven’t changed, and 21 years of growth suggest the model still works. Or? What happens to agencies when AI enters the picture?
The engineering shift
The common misconception is that AI can now write code, so agencies that build software are in trouble. The reality is more nuanced.
Yes, AI is hitting engineering hard, and Infinum’s own data confirms it. Comparing November 2025 – January 2026 to February – April 2026, pull requests are up 45%, and commits are up 77%. The build is getting faster.
But software delivery has never been just about writing code.
Consulting and strategy, design, project management, QA, security – everything that happens before and around the code is becoming more complex, more consequential, and harder to shortcut. As AI accelerates the build, that surrounding work isn’t shrinking. It’s growing in importance.
– NIKOLA KAPRALJEVIC, CEO
And as with any major shift, the gains come with a cost.
The pressure on great engineers
The gap between top engineers and average ones is widening. Senior developers are under serious pressure: PRs are getting larger and harder to review and management expectations have risen accordingly.
Infinum’s response has been practical.
Phantom, our Slack-native in-house AI agent, lets product teams such as design or QA quickly raise issues directly in Slack and receive a pull request and a build to test. The bottleneck used to be engineering availability, but Phantom shipped 135 jobs in May alone.
The result: Infinum teams are now regularly landing at the lower end of project estimates, or below them. On new agentic projects, where the team has freedom to configure things from the start – internal documentation, coding standards, proper feedback loops, automated test suites, code coverage gates, security scanning – code quality holds up.
For large legacy builds, Nikola’s honest message is different: you’ll need to slow down first to speed up.
The new age of the software engineer
We have also noticed something else.
Onboarding to new projects, even with unfamiliar tech stacks, is noticeably faster and easier than it used to be. AI effectively lowers the barrier to working across technologies – an engineer doesn’t need to be a specialist in a stack to contribute meaningfully.
The knock-on effect is a blurring of boundaries. Frontend and backend, specialist and generalist – the lines that once defined how teams were structured and how work was divided are becoming less fixed.
Full-stack is the direction things are moving, not as an expectation imposed on engineers, but as a natural consequence of how AI is reshaping what’s possible within a single role.
For Infinum, this has translated into a concrete strategic shift.
External hiring has slowed down. The focus now is on further upskilling the team — developing engineers who can operate across a wider surface area, rather than adding headcount with narrower specialisations.
The cost of keeping up
Token costs are the new line item nobody budgeted for. Since January, Infinum has seen a 5x spike. Token spend will need to be managed like any other resource, or it won’t be sustainable. Nixa was direct: it puts real pressure on margins.
As for client budgets: smaller one-off projects will increasingly be handled in-house with AI tools, and projects in maintenance mode may scale down.
Ongoing programs that want to stay competitive won’t – because their competitors are also accelerating, and clients can’t afford to stand still.
Finally, the industry is heading toward consolidation. Smaller agencies have been struggling with stagnant or declining revenue and shrinking room to invest. New business development is increasingly expensive even at scale. Some of that pressure will accelerate acquisitions.
So finally – why do agencies still matter?
AI doesn’t make the problems that agencies solve disappear. If anything, the stakes get higher. Clients’ competitors aren’t slowing down. They’re building faster, releasing more, and pushing harder for users’ attention. Staying ahead requires more value delivered, not less.
The craft of building things that actually work, thoughtfully, responsibly, with real judgment behind every decision, only becomes more valuable as building gets cheaper.
That’s exactly why Infinum keeps reinvesting – in cybersecurity through AMR CyberSecurity – Part of Infinum, in brand and creative capability through Your Majesty – Part of Infinum, and now in AI-native ways of working across every engagement.
Infinum has navigated 21 years of technology shifts by changing before it had to. This wave is no different. If you’re navigating the same questions, our team would be glad to compare notes.