The news that affects people who work in enterprise AI, distilled, opinionated, and honest about what matters. One email. Five minutes. Every Friday.
Three big moves this week. IBM Think laid out the most explicit "AI operating model" yet. Anthropic planted a flag in financial services with Claude Opus 4.7. And ServiceNow + Accenture made forward‑deployed engineering the new default for getting AI past pilot. The framing matters more than any single product…
Models commoditize. Prompts commoditize. The thing that doesn't is the labeled data, the test suite, and the human judgment loop that lets you swap a model out without losing your customer.
Salesforce, HubSpot, and a quiet new entrant each shipped relationship‑graph features in the same week. What they have in common, and what they don't.
Palantir wrote the playbook a decade ago. The whole AI category just rediscovered it. Why FDE programs are showing up on every vendor's GTM slide, and which ones are real.
The new pricing, the new admin console, and the new SLA. What's different, what's table‑stakes, and the one move that should make Microsoft nervous.
Finance, legal, healthcare, supply chain. A scorecard of who's shipped real product, who's still selling demos, and where the next consolidation wave is going to hit.
Orchestration across vendors, models, and protocols. Why MCP isn't the answer, why A2A might be, and what enterprises are doing in the meantime.
We talked to a dozen CIOs about what's getting cut, what's getting doubled, and the line items they're moving from "innovation" to "run‑the‑business."
Hugging Face is the exception that proves the rule. Three failed marketplaces, one observation about what actually makes infrastructure sticky, and what comes next.
One email. Five minutes. The signal in the noise of enterprise AI.