Artificially intelligent

Artificial intelligence has brought clear gains in efficiency and productivity. Corporates are cutting workforces, consultancies are changing pricing models, junior hiring is freezing. All in the name of cost rationalisation. But are we sure efficiency should be the north star? The question deserves more scrutiny than it usually gets, from both an economic and a business perspective.

As humans we tend to value what is inefficient. Switzerland has around 700 watchmaking firms. The world probably does not need them all. If efficiency ruled, one would be enough, perhaps a single quartz producer. Or better still, a sport watch packed with sensors and data points to optimise how we live and sleep. More precise than a mechanical movement. More functional in every measurable way.

But we value the magic, the complexity, and even the scarcity of great craft. That is worth understanding before we hand the creative work over to machines.

A recent paper proposed a four-stage view of AI's role in innovation. First, AI as a tool. Second, AI as an interactive support agent. Third, AI as a fully equivalent member of the innovation team. Fourth, AI as an independent orchestrator of innovation efforts and teams.

The first stage is normal by now. If a product is advertised only as AI-powered, that is no longer a differentiator. The second stage, agentic AI, raises harder questions. A project management system that autonomously allocates resources and identifies bottlenecks. Many organisations would welcome this and point to a "logical actor" to justify decisions. Or AI-driven innovation systems that could scout a company's external environment and "autonomously manage and execute innovation projects without significant human input and oversight."

The question is whether this is what we actually want.

The issue with machine learning technology applied to innovation management is that innovation is future-oriented. Ernesto Gismondi had a useful rule: do not look to markets first when developing new products. "We make proposals to people," he used to say. If we develop services only by analysing thousands of data points, we may miss the point. Sometimes the apparently inefficient choice is the right one, but only once you understand its purpose and meaning.

Ursula K. Le Guin imagined something relevant in The Dispossessed. In her ambiguous utopia, a central computer allocates activities to citizens who can still refuse them. The parallel with agentic AI is uncomfortable. With AI, the real question is not what we can do with it. The question is what we want from it.

The tedious chores can go. The thinking, the variety, the reading and socialising: those are worth keeping.

Links: Liquid innovation, Learning loops and innovation, Meaning drives innovation beyond tech and business models