Your AI can do more than your organisation knows how to ask for

Enterprise AI capabilities have outpaced enterprise deployment, and the distance is growing every quarter. The industry calls it the “capability overhang”. After four days with the leadership teams building frontier AI, here is what we learned about why organisations get stuck and what the ones pulling ahead are doing differently.

There is a growing gap between what frontier AI systems can deliver and what enterprises are deploying. The industry calls it the “capability overhang”, and after four days in San Francisco and Silicon Valley – meeting directly with leadership teams at OpenAI, Anthropic, Google, AWS, Perplexity and Zoom, alongside investors at Andreessen Horowitz and Menlo Ventures – we believe closing it is the defining strategic challenge for 2026.

We made the trip with CEOs and board-level leaders from Cochlear, Commonwealth Bank, Hays, MinterEllison, Nine, Origin Energy, Qantas, Suncorp and Ticketek. The conversations ranged widely, but the gap between what AI can do and what enterprises are doing with it came up in every room we walked into.

AI is capable of far more than most enterprise applications are asking of it. The bottleneck is no longer the technology. It is the bridge between what AI can do and what people and organisations have learned to ask of it.

As an OpenAI leader put it, “Yesterday’s model can do everything you can dream of tomorrow.”

The capability is already here, but the enterprise gap is widening, not closing.

The scale of the gap is visible in how enterprises are spending. Menlo Ventures reports that AI now captures 6% of the global SaaS market, growing faster than any software category in history. Enterprises are buying productivity tools and code assistants while the infrastructure for agentic deployment, autonomous workflows, and large-scale process change sits largely untouched.

What keeps organisations stuck

Most organisations are stuck for one of two reasons, and often both.

The first is familiar. Proof-of-concept programs run successfully in contained environments but stall at the point of scaling into production. Dozens of POCs are evaluated and written up, with very few crossing the threshold into live deployment. The enterprises that have broken through are compounding their advantage rapidly. Yet Anthropic reported that the majority of its 300,000-plus business customers are still operating near the base of the deployment curve.

The second, in our experience, is more consequential. Most organisations limit AI access to small pilot groups. The logic feels sound. But organisations that roll AI tools out to everyone generate the usage signal that tells leadership where the real value sits. Broad deployment shows leadership who is using AI most, where unexpected use cases are emerging, and which teams are building habits others need to learn from. Selective pilots suppress that signal. Broad deployment creates it.

Why the most engaged feel furthest behind

The most sophisticated clients, the ones who have gone furthest and seen the most, consistently describe feeling furthest behind. Not because they are failing, but because serious engagement reveals how much remains untouched. The organisations that have not yet started tend to feel the least urgency.

If your leadership team feels comfortable with where you sit on AI deployment, that comfort may itself be the warning sign.

The gap is not shrinking. Every quarter of under-deployment compounds the distance, not in missed productivity alone, but in the usage data, high-value applications, and organisational capability that accumulates only through committed use. The frontier labs are not waiting.

We will share more of what we learned in the next edition. If your organisation is working through similar questions, reach out to our team.