Is the Future of AI Value Bigger Models or Smarter Systems?

We are starting to see in research coming out indicators of the truth of where the value proposition in AI is likely to sit. For much of the AI revolution, success appeared to be a race towards ever larger and more expensive models. However, recent research demonstrates that the greatest economic value may not lie within a single model, but within the intelligent ecosystem that surrounds it.

A study by MIT tracked software developers work before and after they adopted AI tools. Importantly, they measured this at several different levels, from the amount of code written, to the number of discrete files edited, to the number of projects or features worked on, to actual releases of new software. They found an explosive impact at the top of this funnel that in turn shrunk fivefold to a roughly 30%  uplift in the number of full software releases. Sounds good yes? However this productivity was not reflecting in any sizable uptick in marketplace uptake of apps or services. So a 30% productive increases in software production leading to little evidence of increased consumption sounds like a hidden cost of 30% more ‘ghost code’ to manage and vulnerable contact surface potential. Sounds more like AI jobs for the boys, rock in Mythos and its Cyber AI models.

A then there is a new approach is emerging around open-source AI agents perform the majority of day-to-day work, while selectively calling upon frontier models as expert advisors only when higher levels of reasoning, validation or specialist knowledge are required. This mirrors how successful organisations operate, with capable teams handling routine activities and experts engaged only when necessary.

The implications are significant. Organisations can achieve near-frontier performance while dramatically reducing operating costs, increasing transparency and retaining greater control over their intellectual property and data. Rather than purchasing intelligence as a premium commodity, they orchestrate it as a strategic capability.

The lesson extends beyond AI engineering. As intelligence becomes increasingly abundant, competitive advantage shifts away from the model itself and towards system design, orchestration, governance, proprietary data and domain expertise. Value accumulates in the architecture that coordinates intelligence, not merely in the intelligence engine.

The winners of the AI era are therefore unlikely to be those with access to the largest models. They will be those who build the most effective ecosystems around them, combining open innovation, specialised expertise and trusted governance into a scalable and economically sustainable operating model.

In the age of abundant intelligence, the model becomes a commodity. The system becomes the differentiator and success will not be determined by who owns the best model or churn the most software but by who builds the best system around it. The health warning to the inflated AI company valuations is that if I am correct, the economic value will no longer accumulate in the model but in the system delivering the competitive differentiation.

As I eluded to in my last blog, transformation I believe is the key word when it comes to tapping the true value and potential of AI and as promised I will address this later this weekend.