.jpg)
An evening for the people deciding what their organisation should build with AI, and who needs to be capable of building it. Presentation of the paper, a panel with both co-founders, then open discussion.
Across organisations, some people are using AI to complete existing tasks faster. Others are using the same technology to redesign how the work itself happens, building systems, workflows and agents that affect hundreds or thousands of people. That difference is the subject of CFTE's new whitepaper.
In May 2025, The AI-fication of Talents identified a hidden trend. System Thinkers were acquiring new leverage because AI was collapsing the distance between an idea and its implementation. Cursor was one of the clearest early signals, a small team reaching a scale that historically required far more people, capital and time.
Thirteen months on, the pattern is easier to see. Small AI-native teams produce organisation-scale outcomes. Professionals inside established institutions build systems that previously required a formal technology project. Experts turn their methods and judgement into executable AI capabilities that thousands of other people can use.
When the ability to build becomes widely available, the constraint moves. The question is no longer who can build, but who knows what should be built, how the parts should work together, and where human judgement must remain.
.jpg)
The clearest results came from defined tasks close to a real user need, where AI released capacity without removing human judgement.
AI added to a process built around manual checks and legacy systems produces little. The end to end workflow has to be mapped first.
Colleague facing assistance and autonomous customer facing action were treated as different categories, with review and escalation retained for high impact decisions.
Token, licence, compute and assurance costs may decide which use cases survive at scale, so feasibility demonstrations are no longer enough.
Once systems plan and act, ownership, permissions, auditability, incentive alignment and resilience matter as much as model performance.
The strongest cross cutting finding. Professionals lack practical understanding of what AI can do, where it fails and how to redesign work around it.