- within Wealth Management and Intellectual Property topic(s)
January 2026 was my fourth Davos, and I have learned that each one is shaped by themes not envisaged beforehand. This year was no different: geopolitics and AI predominated (in that order), but the official “spirit of dialogue” proved elusive.
The repurposed resort is in a valley flanked by two long mountain ridgelines. That means it also has two horizons, depending on whether you look at Parsenn-Weissfluhjoch to the north-east or Jakobshorn-Schatzalp to the south-west.
I know I should have just enjoyed the view, but the landscape reminded me of my favourite German philosopher, Hans-Georg Gadamer, who similarly embraced both a spirit of dialogue and the idea of two horizons.
For him, “horizons” represent the personal histories that form what we would now call “priors” or “preconceptions” – the place we are “coming from”. Although inevitably biased and fallible, Gadamer believed that “fusing” different horizons together through dialogue could generate new understanding about the world.
Not a bad summary of what Davos is all about.
A polarised AI non-debate
However, on subject of AI, that fusion didn’t happen.
Instead, two groups, each forming a horizon, largely spoke to themselves, or provided one-way transmissions to whoever would listen.
On one side of the “Davos valley” were the accelerationists. Their horizon centred on the frontier lab narrative, further amplified by consultants and start-ups. They were all about exponential progress, transformed futures, and ready-now capabilities. Only regulation, timidity, and legacy thinking could slow things down.
The other horizon was new to Davos: a practical-realist perspective shared by many mainstream business leaders, some researchers, and even a few big-tech companies. It acknowledged the long-term potential of AI but spoke of “pilot purgatory”, challenges in governing agents, frustration over elusive ROI, and the inability of LLMs to drive top-line growth.
Maybe there was a panel somewhere that really brought these two perspectives into meaningful dialogue. But if there was, I didn’t find it.
Why no fusion?
On reflection, it is not difficult to understand the gulf. The two camps are motivated and constrained by very different forces. Frontier labs are focused on attracting major investment and securing market share. These existential objectives are best served by a big, future-oriented narrative and rapidly evolving products. Dwelling on adoption difficulties or hallucinations would undermine the case for their long-term potential.
In contrast, most mainstream businesses are focused on here-and-now productivity, safe adoption, and growth. They are constrained by a profit imperative, finite access to capital, and the need to demonstrate measurable AI progress to shareholders and customers.
These are not just distinct horizons; they are different worlds.
Why fusion matters
The irony is that the future of both groups depends entirely on the other. Frontier labs need to build reliable revenue streams as they move towards IPOs and profits. This can only be achieved by meeting customer needs for measurable outcomes and safety.
Equally, business leaders need AI to revitalise their enterprises, unlock revenue potential, and increase productivity. To achieve that, they must build deep partnerships, take some risks, and move beyond experimentation.
Reasons for hope
Despite the Davos impasse, there are some reasons to be optimistic.
1.The rise of the practical-realists.The discernible voice of corporate users is progress in itself - and accelerationists can’t ignore them for long.
2.A growing body of high-quality adoption research. We are beginning to understand the problems of AI implementation throughrigorous studies– a necessary precondition to solving them.
3.New “fusion” roles.Forward Deployed Engineersare just one example of promising AI jobs that will link frontier labs and their users.
4.Pragmatic AI technology fusions. Even though frontier labs are going all out to scale LLMs towards “AGI”,othersare combining generative and old-style (neurosymbolic) AI to achieve better reliability for users.
I predict that enough fusion will occur in 2026 to produce several scaled productivity use cases focused on back-office functions and customer service. This progress will alleviate but not solve pilot purgatory as mass adoption of complex but safe multi-agent systems begins. Growth-oriented innovations are likely to be a phenomenon in 2027 or later.
Conclusions
Memories of Davos fade quickly, but the need for meaningful discussion remains. Maybe we can salvage the “spirit of dialogue” beyond big conferences and get AI working on the ground for the benefit of parties of all horizons.
If you’d like to hear more of the AI debate from Davos, you can also watch a panel discussion I participated in:Shaping the AI future we really want.
If you'd like to listen to an audio version of this article, please click here.
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.
[View Source]