Why the next breakthrough in AI agents comes down to leadership, not technology

For most of the past decade, artificial intelligence has been framed as potential. Powerful, promising, but out of reach. A distant frontier. 

That frontier is no longer ahead of us. It's here: jagged, messy and moving fast.

For the first time in the generative AI cycle, AI agents, particularly coding agents, are being paid for at scale, with clear return on investment. Unlike earlier AI hype waves, this demand is translating into real revenue, committed budgets and measurable productivity gains inside enterprises.

This shift is visible in the numbers. Leading AI platforms have seen annualised revenues accelerate sharply, with growth increasingly constrained by available compute rather than lack of demand. Recent analysis found that 4% of GitHub public commits are being authored by Claude Code, with that number expected to increase to over 20% by the end of 2026.

AI agents have crossed the economic threshold. At the same time, the systems that support them – compute, memory, energy and data-centre capacity – remain finite. 

This is compounded by the fact that workforce readiness remains a significant barrier to realising the full value of AI. Kyndryl's 2026 People Readiness Report found that only 23% of organisations believe their workforce is fully ready for the technology – a decrease of six percentage points from last year – despite enterprise-wide adoption surging to 57%.

In this environment, the organisations that will succeed will be those that can design, deploy and govern AI agents effectively, with people firmly in the loop. The challenge for many organisations is not just AI adoption itself, but also ensuring their workforce, operating models and governance frameworks are evolving at the same pace.

The real constraints leaders are underestimating

The next phase of agentic AI adoption will not be limited by model capability. It will, however, be shaped by physical and economic constraints. Across the AI value chain, capacity is under pressure. Data-centre build-outs require huge upfront investment. Advanced chips and high bandwidth memory are controlled by a small number of suppliers. Energy availability and grid capacity are becoming critical limiting factors.

Together, these forces will create an environment where demand consistently exceeds supply. As organisations seek to scale AI against real capacity constraints, leaders must make deliberate choices about where agents deliver genuine value and where returns justify ongoing investment, while ensuring their workforce has the skills and confidence to use AI effectively.

From adoption to orchestration

To be able to truly unlock the competitive advantages of agentic AI, leaders must understand how to orchestrate AI agents effectively across their complex environments. This involves determining which agents are trusted for which tasks and most importantly, under what conditions. It requires the ability to shift workloads between models as pricing, performance and availability evolve. It depends on designing workflows that accelerate delivery while maintaining transparency and accountability.

In an environment defined by capacity constraints, flexibility becomes a strategic asset. Organisations that can adapt - both technically and operationally - are better positioned to sustain access and value over time.

Additionally, leaders must acknowledge that orchestration extends beyond the technology itself and is also an operating model challenge. This means that clear decision rights, escalation paths and ownership are essential elements when agents operate as part of the execution layer. 

This is becoming increasingly urgent as 81% of organisations expect AI agents to make impactful decisions within the next year, and 66% have given AI autonomous read and write access to core systems of record.

Designing AI with people at the centre

As the design and adoption of agents accelerates, leaders must define success in a way that strengthens their workforce, rather than sidelines it. 

The strongest outcomes come when agents remove friction from work. By automating repetitive and low-value tasks, they allow experienced professionals to spend more time on judgement, architecture, risk management and creativity - the areas where human capability creates the greatest value.

Leaders need to design intentionally for this outcome. The organisations seeing the strongest results are already taking this approach: 61% have redesigned roles around AI, while 24% have created entirely new roles focused on AI management.

Organisations that overlook this design work may see short-term efficiency gains, but risk longer-term erosion of capability, trust and operational resilience.

A leadership test, not a technology race

The agent era is well and truly here, but what comes next will be shaped by leadership choices: where AI is applied, how it is governed and how people are supported to work productively alongside it.

In a world where demand for AI exceeds the physical capacity to supply it, true advantage comes to organisations that design deliberately, orchestrate intelligently and keep human judgement at the centre. Turning AI investment into business value will depend on a workforce that is ready, skilled and empowered to use it effectively.