Everyone bought the AI coding tools, but the productivity boom didn’t automatically follow. And Claudio González, CTO at Germany-headquartered software engineering and digital product consultancy intive claims the companies winning aren’t the ones with the best tools – they’re the ones who rebuilt the team around them.
The tools arrived first. Across the industry, engineering teams now reach for AI the way they once reached for Stack Overflow. In 2025, four in five developers said they were using AI in their workflow.
But here’s the twist in the same Stack Overflow Developer Survey: as adoption climbed, trust in what the AI produced actually fell.
González has a theory as to why. He’s watched plenty of clients buy tools and wait for the magic to show up. “Adding a code assistant on top of the old one doesn’t get you there,” he told 150sec.
A faster tool, the same bottleneck?
Google’s 2025 DORA report – one of the most respected studies of software delivery – found that AI behaves less like a magic lever and more like an amplifier: it makes strong teams stronger and struggling teams more chaotic, depending entirely on the process around it. Speed in one place just exposes the sluggishness elsewhere.
González frames the fix as a workflow problem, not a tooling one. On one large platform engagement, he says, intive’s teams hit a 40% increase in total development output, 35% faster feature cycles, and 50% faster modernization – but only after redesigning the process end to end. “These gains come from redesigning the workflow,” he said.
Independent research lands in the same spot. A November 2025 McKinsey analysis found that pairing generative AI with genuine process change pays off far more than bolting an assistant onto yesterday’s pipeline.
From “manual builders” to “architects of intent”
The old instinct was to throw bodies at a build; a big pod of developers grinding through tickets. The new leverage, the intive CTO argued, sits in a small group of people strong on architecture and judgment, with AI agents handling the boilerplate, the test generation, the legacy decoding.
González has a name for the shift: from “manual builders” to “architects of intent.”
“The constraint is no longer how many hands you have,” he explained, “it’s how good your people are at deciding what to build and verifying it’s correct.”
That’s not a startup-only story or a solely enterprise-focused story. It’s the same math for a retailer’s recommendation engine, a logistics platform, a health-tech product, or a fintech starting out. Gartner projects that AI assistants will surge among enterprise engineers within a few years, making the real question less whether to adopt and more who you actually need in the room.
The real blocker is rarely the code
The hardest part, González stressed, isn’t usually writing software. Rather, it’s understanding the software you already have.
Plenty of companies are sitting on systems built decades ago; undocumented, quietly running, maintained by people who’ve since retired. In early 2025, in fact, the UK’s National Audit Office reported that the government was still running at least 228 ageing “legacy” IT systems and, tellingly, didn’t even know how vulnerable most of them were, with no funded plan to fix around half.
The instinct is to blame the old code. González thinks that’s the wrong target. “It’s not the legacy itself. It’s the lost knowledge around it.”
Translate the old language into a new one, and you simply inherit the same debt with cleaner syntax; speed without understanding quietly piles up debt. So intive points AI at the problem differently: as a kind of “systems historian,” decoding buried logic and recovering business rules before anyone rewrites a line.
Strategy first: AI just makes it survivable
For all the optimism, González kept circling back to a limit. “None of these roadblocks are purely technical,” he said. Unclear goals, no early wins, shaky stakeholder confidence – every one of them is an organizational gap, and no model closes that for you.
The executive is far from alone in that reading. A 2025 MIT study of enterprise AI found roughly 95% of corporate generative-AI pilots were delivering no measurable return, and researchers pinned the failures on how companies integrate the technology – not on the models themselves.
Meanwhile nonprofit research organization RAND reached a similar verdict, finding that more than 80% of AI projects fail. About twice the rate of IT projects that don’t involve AI, most often because the problem was misunderstood or leadership was misaligned, not because the tech fell short.
“Strategy still has to come first. AI just makes execution survivable,” González concluded. The tools, in his view, are the minimum now. The teams – smaller, sharper, pointed at the right problem – are the actual edge.
Featured image: TECNIC Bioprocess Solutions via Unsplash+
Disclosure: This article mentions clients of an Espacio portfolio company.
