Last November, a few colleagues and I attended the Gartner Symposium in Barcelona.
This conference always manages to be exhausting, slightly overwhelming, occasionally superficial -and genuinely excellent- all at the same time. A bit like tapas: many small plates, not all memorable, but the overall experience is worth it.
What I particularly enjoy is the high-end mix between:
– research-based insights showing where technology topics actually stand,
– unexpected keynote conversations,
– and the familiar (and slightly ritualistic) “top IT trends” narrative.
This year’s speakers included an eclectic and memorable lineup:
– Margrethe Vestager former Executive Vice-President of the European Commission (digital & competition policy)
– Charles Duhigg bestselling author (The Power of Habit, Smarter Faster Better)
– Dr Michelle Dickinson science communicator and keynote speaker
– Bear Grylls adventurer, TV host, author, professional reminder that corporate life is comparatively comfortable
– Jo Malone perfume entrepreneur, probably the most unexpected talk of the event — and one I enjoyed immensely. As distinctive as her fragrances.
And of course, plenty of AI discussions -because 2025 without AI would be like Barcelona without Gaudí.
Interestingly, something from the 2023 conference resurfaced in my mind during this year’s edition -the growing gap between expectations and reality around AI.
Several points stood out.
AI value: still early days
Gartner presented data suggesting that around 20% of AI initiatives achieve clear ROI, while only about 2% generate truly disruptive value.
That is both encouraging and sobering -AI delivers value, but not automatically and not everywhere.
The CFO perspective
Predictably, CFOs appear even more cautious than CIOs (which is saying something).
Translation: “Interesting technology – now show me the numbers.”
GenAI reliability (the part nobody loves)
One figure that sparked discussion: generative AI error rates can still be around 25%.
That’s perfectly acceptable when drafting ideas or summaries. Less so when 100% accuracy is required.
Which leads to practical implications:
– define formal quality metrics
– implement two-factor error checking
– think explicitly about the “good-enough threshold”
– invest in AI literacy for staff
In short: AI still needs adult supervision.
AI ? fewer people (at least not yet)
A widely held assumption remains: deploy AI ? reduce headcount.
Gartner’s observations suggest something more nuanced.
Implementing AI takes time, requires new skills, and creates additional work before efficiency gains appear. The immediate effect is often redistribution of effort, not reduction.
Workforce reality
Another structural point raised: global workforce growth is expected to stagnate, and in Switzerland this could happen around 2033.
This reframes AI entirely -not as a cost-cutting tool, but as a capacity-sustaining one.
Final thought
None of this suggests AI is overhyped or unimportant -quite the opposite. Organizations should move quickly to explore where AI creates real value.
But AI is probably not the universal solution to every operational problem. More likely, it is a powerful tool that rewards disciplined, realistic adoption.
AI will not replace strategy -but it will expose the lack of one.