Dead reckoning is how you fix your position when there’s nothing to sight. You know your speed, your heading, you know how long you’ve been underway. So you commit to a mark on the chart knowing it’s an estimate you’ll correct as you go. That’s the honest description of AI strategy inside most large institutions right now, and it’s what this newsletter is for.
I’m Andrew Welch. I lead Cloud Lighthouse, a specialist AI lab working across 15+ countries that acts as outside counsel to international organizations, governments, and regulated enterprises trying to scale AI without betting the institution on any one vendor. We build the model they use to do it, we build with them in our lab, and we represent their interests wherever AI decisions get made. What we don’t do is sell an LLM or an implementation army, which is what lets us tell clients things they don’t want to hear. I also lead the Center for Trustworthy AI, our open research and education arm, and convene the AI Trust Forum, which puts the people wrestling with these questions in the same room a few times a year.
Before this I was global CTO at one of Europe’s larger systems integrators, and before that a long stretch in enterprise data, software, and infrastructure. Twenty years on that side of the table is where I came to believe that advice is worth less when the firm giving it also sells the thing being recommended. Cloud Lighthouse is built the other way around. Before all of it, the US Coast Guard, where I first learned to plot a course.
What this is
Essays on AI, written from inside the institutions that have to answer for it. Not commentary from the outside, and not a roundup of what shipped this week. I’m interested in what actually changes for a CIO at a development bank, a program lead at a public sector or United Nations agency, or a board that’s been told to have an AI position by next quarter. I write in the gap between what gets announced and what gets built. I’ll share essays every few weeks or when there’s something worth saying. Each is a piece you forward to a colleague with “read this before the meeting.” You suspect you will disagree with some. That’s the point.
I try to avoid the breathless register that has taken over most writing about AI. I don’t write about tooling for its own sake, and I don’t pretend to certainty I don’t have. The mark on the chart is an estimate. Again, that’s the point.
The rest
Swedish-American, living in Valencia, Spain down the street from the United Nations Digital Hub. Hometown is Provincetown, on Cape Cod. Ana’s husband. Alexandra’s dad. I wrote a novel, Field Blends, in 2020. I like wine more than is professionally advisable and I am most at home on a boat, inspiration for the name Dead Reckoning.
Reply to any edition and I’ll read it. I’m also on Bluesky at @andrewdwelch.com.
