A practitioner's perspective on AI, cloud, and technology transformation — and the economics of making them work inside real enterprises. Every issue is archived here on my own site.
Where do you begin with enterprise AI? Start with one consequential decision — a trade break — follow how people actually resolve it, and map the context your systems don't share. Then give AI a bounded job and measure what changes.
Read the postAn AI system can retrieve the record, read the contract, query the database — and still give the wrong answer. The enterprise's real shortage isn't data; it's shared context: definitions, authority, business rules, and history.
Read the postFor 30 years, enterprise technology has tried to eliminate silos. It hasn't worked. Maybe the winning AI architecture isn't another platform that consolidates everything — but an intelligence layer that makes the systems you already have intelligible to each other.
Read the postYour AI tools work — they summarize, detect anomalies, answer questions. So why do executives keep asking where the ROI is? Perhaps because each tool is making a silo smarter instead of making the enterprise intelligent.
Read the postIf AI agents become how employees get work done, the agent becomes the interface between employee and enterprise — and owning that layer may matter more than owning the best model. The next great enterprise technology battle.
Read the postAn employee saves an hour — but an hour saved isn't automatically an hour of economic value. Why enterprise AI has to move from usage metrics to real business outcomes, and where ROI actually begins.
Read the postAI itself keeps getting cheaper — smaller models, falling inference costs. But putting it safely into production inside a regulated enterprise is where the real cost lives: data, integration, security, governance, controls, and human oversight.
Read the postEnterprises don't buy tokens — they buy work. How Microsoft, Google, OpenAI, and Anthropic are all racing up the stack from tokens to work, and why the real metric is dollars of useful work per dollar spent.
Read the postJane Street's ~$15B July loss and the Situational Awareness fund — a spectacular AI-trade reversal, and what it says about the gap between a real technology and a successful investment.
Read the postWhat if AI compute itself becomes something to be financed, securitized, and traded like a commodity? A look at the financialization of compute — and where the analogy breaks down.
Read the postNvidia and the hyperscalers are deploying hundreds of billions into AI infrastructure. But someone ultimately has to generate the return. A practitioner's view on the total cost of enterprise AI — and who captures the value.
Read the postToday's AI investment thesis assumes ever-growing centralized cloud inference. But open-weight models, smaller models, and AI-capable PCs raise a question: what if inference becomes increasingly hybrid — and what does that mean for the trillions being committed to infrastructure?
Read the post