Article · September 22, 2025

The GenAI bottleneck was never the model

Two slides, pointing in opposite directions. One: 95% of generative-AI pilots are failing. The other: people getting real work out of the same tools.

The first was the headline that MIT NANDA’s report The GenAI Divide (opens in a new tab) set off that summer. The second took one focused task and almost no organisational apparatus. Same technology, opposite results: the difference sits in the organisation around the model.

That was Denis’s opening at the first GenAI Enterprise Day, an evening from GenAI Enterprise Hub (opens in a new tab) hosted by ekona.ai in Basel during Swiss {ai} Weeks. Five speakers then took the question apart.

Opening slide contrasting a headline on enterprise AI returns with examples of practical adoption.
The opening slide: one side of the argument.

Five places AI stalls

#
  • Adoption. Katarina Granath, SIX, on the structures that make it stick: an AI circle, a community, job rotations, mentorship and management backing.
  • Experimentation. Patrick Hammer, Nezasa, on getting every role and department to try the tools on its own work.
  • Workflow and trust. Sam Khalil, ekona.ai and the evening’s host, on what it takes to become “AI-agents native”.
  • Responsible use. Anne-Marie Buzatu, ICT4Peace Foundation, on ethics, human rights and security when resources are tight.
  • Procurement. Sandro Stich, Alfred, on moving AI contract talks from pushback to partnership: liability and the use of data.
Diagram grouping corporate transformation, SME adoption, implementation, ethics and security, and legal concerns.
Five perspectives, one question.

The five returned for the closing panel, Rethinking Business in the Age of GenAI, moderated by Denis: what had moved beyond experimentation, what had stalled and why, and how each organisation was structuring itself around the question.

A full room of guests talking over drinks and food in ekona’s glass-walled hall in the evening.
After the panel, the discussion moved to the apéro.

What it means for your project

#

None of these is fixed by a better model, and they echo what we meet in our own delivery work – see five practical obstacles to useful AI.

A plan that covers only the technical system covers a third of the problem. It also needs the workflow, and the organisation that will run it.

Use the five as a checklist before your next AI investment: adoption, experimentation, workflow, responsible use and procurement. We work through it with your team. See workshops & training, or Start a conversation.

Download as PDF

Have a project in mind?

Tell us what you are building or improving. We'll help you choose a practical next step.