The gap between what AI can do, and what companies actually ship
Every week I publish one article exploring an idea about AI, agents, and their diffusion into the enterprise and people’s lives.
AI is already better than most humans at most knowledge work. So why isn’t it adding trillions to GDP yet?
Three things sit in the way. Diffusion: the technology can’t reach the people who would actually use it. Coordination: 60-80% of work time goes to “work about work,” which no model has been trained to do. Traces: the decisions and judgments that drive value live in people’s heads, not in any document.
Those three bottlenecks are the next decade of enterprise AI. This publication is about seeing them, naming them, and shipping through them.
Free, weekly. 5 to 10 minutes per post.
Who is this for
If you’ve shipped, or have to ship, AI agents inside a bank, fintech, card issuer or merchant acquirer, this is for you.
If you’ve sat in a steerco where the slide says “AI initiative” and the room says “another pilot,” this is for you.
If you want to be at the forefront of where AI evolves and its impact in the real world, this is for you.
If you want to be on the top 1% (if not 0.1%) of professionals with actual understanding on all-things AI, this is for you.
Who I am
I’m Angel Salinas, author of this publication. I design and build agentic systems, both professionally and on my own time.
By day, I’m Director at Visa Advisory, where I lead the AI Labs. We deploy custom agents inside banks, fintechs, payment processors, and merchants across Europe. Most of the projects I see across the industry stall in the same three places (the ones at the top of this page). I write about those
On my own time, I build agentic systems to find the places the technology actually breaks. The day work and the side work feed each other: what fails inside large enterprises shapes what I prototype on weekends, and what I prototype on weekends sharpens the questions I bring back to enterprise work.
All views here are my own and do not represent Visa Inc. or its affiliates. I won’t write about Visa clients, deals, internal numbers, or payments competitors.
What you’ll get here
One email a week. An exploration of one useful idea around AI, Agents and their diffusion into companies and people’s lives.
If you read something here that lands, or that you disagree with, reply to the email. I read every reply.
Where to start
If you’re new here, these will give you the clearest picture of what this publication is about:
7 reasons why your company AI efforts fail → A taxonomy of the reasons enterprise AI initiatives stall, written for readers stuck in the 95% of companies with a pilot graveyard and nothing in production.
AI & Enterprise Value → The founding manifesto. The real economic value of AI in the enterprise will not come from giving everyone a copilot. It will come from custom agentic systems deployed on the right business processes. A framework for picking which.
On diffusion, coordination and traces → The thesis post. The three bottlenecks named at the top of this page, in their full version. The frame the rest of the publication builds on.
Agentic Coordination → Knowledge workers lose 60% of their time to coordination overhead. Agents promise to reclaim it, and to create entirely new coordination layers. The end state is becoming a conductor of infinite minds. Very few professionals are building that skill yet.
What is then left? →The personal piece. If synthetic intelligence keeps scaling, what remains valuable for knowledge workers? Seven traits I think will matter more, not less, as the models improve.
Stay in touch
Free, weekly, easy to unsubscribe.
(Or find me on LinkedIn)
