The problem
The specialist-stacking era is running out of road
For most of the past decade, the standard answer to an engineering problem was another specialist. It worked. Then AI made individual engineers dramatically more capable and cross-domain work the norm. The coordination overhead that always existed became the bottleneck.
AI changed two things at once. It compressed what one engineer can do. A single developer with the right tools is now shipping what small teams used to. And it raised the cost of coordination, because AI systems cut across domains by nature. A model depends on data pipelines. Data pipelines depend on infrastructure. Infrastructure decisions affect security posture. None of those conversations happen cleanly across a wall of specialists who only speak their own language.
The manager used to be the person who held the full picture. Increasingly, that is not enough. The pace is too fast, the decisions too technical. What teams actually need are engineers who can operate across that picture themselves.
Faster skill change in AI-exposed roles vs. other jobs
PwC · 2025
Of employers cite skill gaps as their top barrier to transformation
WEF · 2025
Of engineers must upskill through 2027 to keep pace with generative AI
Gartner · 2025
PwC's 2025 Global AI Jobs Barometer found skills in AI-exposed roles are changing 66% faster than in other jobs, up from 25% just a year earlier. The World Economic Forum found 63% of employers consider skill gaps their biggest barrier to transformation, with 85% planning to prioritize upskilling in response. Gartner projects 80% of the engineering workforce will need to upskill through 2027 to keep pace with generative AI. The I-shaped engineer, deep in one lane and nowhere else, is facing structural headwinds.
Agentic AI is redefining platforms1. It is transforming how work gets done, how value is created and how humans and machines collaborate.
Strategy alignment unlocks value. Companies that align their AI, platforms and business strategies achieve on average 2.2x revenue growth and a 37% EBITDA lift.
Winning in an AI-first world requires a new platform strategy. We outline five priorities to help companies seize the opportunity.
How do you expect your platform strategy will need to evolve in reponse to agentic AI
The model
The T is not about knowing more. It is about knowing enough.
The T-shaped model gets misread in engineering circles, usually in one of two directions. Either it becomes an argument for generalism, which alienates engineers who spent years mastering their craft, or it becomes a soft skills addendum, vague talk about "communication" and "collaboration" that doesn't change how anyone actually works.
Neither reading is right.
The horizontal bar is not expertise. It is literacy. Enough to collaborate without slowing down the person you are working with.
A T-shaped backend engineer does not need to design UX flows. They need to understand what a designer is protecting so they don't accidentally break it in a release. A T-shaped data engineer doesn't need to own product roadmap decisions. They need enough product context to know which data problem is worth solving first, and which one can wait.
The vertical is still where the engineer earns their seat. That does not change. Deep expertise in systems architecture, machine learning infrastructure, security, or any core discipline is what makes the breadth worth anything. Without a strong vertical, breadth is just dilettantism.
What enterprise teams are now hiring for, explicitly, is the combination. A 2025 engineering hiring analysis found the most valued candidates can bridge AI innovation with practical business outcomes and translate technical complexity for non-technical stakeholders. SSi People's 2025 market report identified T-shaped skills as a defining hiring trend, noting that skills-based recruitment lets organizations fill capability gaps that credential-first hiring consistently misses.
Agentic AI is redefining platforms1. It is transforming how work gets done, how value is created and how humans and machines collaborate.
Strategy alignment unlocks value. Companies that align their AI, platforms and business strategies achieve on average 2.2x revenue growth and a 37% EBITDA lift.
Winning in an AI-first world requires a new platform strategy. We outline five priorities to help companies seize the opportunity.
How do you expect your platform strategy will need to evolve in reponse to agentic AI
The barriers
Why most engineers stay narrow even when the incentives say otherwise
The demand signal for T-shaped engineers is not subtle. And yet most engineers remain narrowly specialized. The reasons are worth understanding clearly, because they point to where intervention actually needs to happen.
01
Depth feels like identity. Breadth feels like starting over.
Engineers who have spent years building mastery in one domain associate that depth with who they are. Moving into adjacent areas feels like regression. Junior again, uncertain again. That friction is real. But the reframe matters: broadening is not abandoning the vertical. It is building leverage on top of it.
02
What gets measured is still narrow output, not cross-functional impact.
Job descriptions are written as tool checklists. Performance reviews count function-specific deliverables. Engineers are rational. They optimize for what the system rewards. Until organizations change what they measure and recognize, the behavior will not shift, regardless of what they say they want.
03
No one tells them which adjacent domain actually matters.
Engineers who want to broaden tend to absorb whatever happens to cross their desk. A backend engineer picks up deployment habits from a DevOps colleague, but never learns the data modeling fundamentals that would have changed how they build APIs. Useful breadth is not random. It requires knowing which gaps are costing you something.
The engineers who do develop a genuine T-shape are deliberate about it. One adjacent domain per year. A support ticket pulled from a different team. A sprint review attended outside their usual squad. None of it is dramatic. It compounds.
How do you expect your platform strategy will need to evolve in reponse to agentic AI
The organizational implication
You cannot hire your way to a T-shaped team
The instinct is to treat this as a sourcing problem. Find the engineers who already have both the depth and the breadth, hire them, done. That instinct is wrong, and it is expensive to chase.
T-shaped engineers are made, not found. Most of them developed that way through specific kinds of organizational exposure: projects that required working outside their lane, managers who pushed them toward adjacent problems, cultures where the expectation was always slightly broader than the job description. Those conditions do not come with the hire. Organizations have to build them.
When they don't, the cost shows up in a particular way. Handoffs accumulate. Each one is a small delay: a ticket waiting on another team, a decision held until someone from a different function can weigh in. Individually, none of it looks catastrophic. Across a project, it adds up to months.
Mastercard's internal learning platform found that more than a third of employees who took on cross-functional projects made an internal career move within a year. That is not coincidence. Exposure to adjacent domains changes how engineers see their own work, and what they are capable of doing next.
The teams that move fastest on AI are not the ones with the deepest specialists. They are the ones where a few engineers can hold the full picture while still doing the technical work.
How do you expect your platform strategy will need to evolve in reponse to agentic AI
At NMBLR, this is the pattern we see consistently. The variable that most predicts whether an AI engagement moves from pilot to production is not tooling, and it is not budget. It is whether someone on the team can work across the problem end-to-end. Technically, not just organizationally.
Powering Reinvention with Wind & Intelligence
Building a digital core to orchestrate the next generation of energy.
Wind energy is no longer just a resource; it is the cornerstone of the multi-generational decarbonization era. To meet rising global expectations, we are moving beyond simple manufacturing to build a fit-for-purpose foundation for the wind industry. By embedding Agentic AI into our windmill technology, we are transforming static infrastructure into living systems of intelligence—capable of reacting dynamically to atmospheric shifts and orchestrating energy output in real-time. This alignment of AI, platform, and business strategy is what unlocks 360° value, reducing reliance on fossil fuels while driving 2.2x higher operational growth.
The path forward
What it looks like when organizations get this right
The T is not permanent. The domains that matter for breadth keep shifting as AI capabilities expand. An engineer who built cross-functional literacy in data modeling five years ago is now being asked to understand AI governance, model evaluation, and deployment risk. The shape evolves with the work.
What does not change is the logic. Depth first. That earns the engineer their credibility and gives their broader perspective something to stand on. Then intentional breadth, aimed at the specific gaps that are actually slowing the team down, not the ones that happen to be convenient to learn.
The organizations that get this right tend to stop treating cross-functional work as something that happens around the edges. They build it into how projects are staffed, how performance is recognized, and what career progression actually looks like. When an engineer knows that working across domains advances them rather than distracts from their track, they do it. It is not complicated. It just requires the organization to mean what it says about wanting versatile engineers.
A backend engineer at an AI company who does not understand data pipeline contracts is quietly costing their team time on every sprint. That is a solvable problem. The cost is small per sprint and invisible in any individual review cycle, which is exactly why most organizations never get around to solving it.
NMBLR AI Foundry partners with enterprise teams to design, build, and deploy AI-ready capabilities across the full stack, from infrastructure to talent architecture. This piece is part of our Foundry Insights series.
Agentic AI is redefining platforms1. It is transforming how work gets done, how value is created and how humans and machines collaborate.
Strategy alignment unlocks value. Companies that align their AI, platforms and business strategies achieve on average 2.2x revenue growth and a 37% EBITDA lift.
Winning in an AI-first world requires a new platform strategy. We outline five priorities to help companies seize the opportunity.
Agentic AI: The new orchestration layer
Agentic AI is becoming the interface across platforms, spanning systems, reacting dynamically and orchestrating work in real time. It can rebalance supply chains, generate personalized offers and even close financial processes.
Example: Lenovo used Adobe Experience Platform and Microsoft Copilot to orchestrate AI across marketing, customer service and internal workflows. The effort delivered $11 million in efficiency savings and a 12.5% boost in click-through rates—speeding execution and enabling new forms of engagement at scale.
When teams use disconnected AI tools, proprietary knowledge escapes into public models while institutional insights get trapped in silos. 68% of organizations cite data fragmentation as their top data management concern — and only 12% report having data of sufficient quality to support effective AI.
Employees fill workflow gaps with personal tools, and most organizations have no idea it's happening. Shadow AI incidents now account for 20% of all data breaches — and carry an average cost premium of $4.63M compared to standard incidents. The average enterprise unknowingly hosts 1,200 unofficial applications.
The ideas are good. The pilots are promising. They just never make it to production. The average organization scrapped 46% of its AI proof-of-concepts in 2025. AI projects fail at twice the rate of non-AI technology projects. That number isn't improving on its own.
