Sub-100-person organizations have different AI procurement constraints than enterprises, tighter budgets, less vendor leverage, less in-house governance, often more sensitive data. Here's the procurement frame that actually fits the shape of those orgs.
GTC 2026 anchors the week. Vera Rubin, a $1T order book through 2027, and a partnership map from Uber to Disney to Eli Lilly. Mistral ships Small 4 and announces Forge for training-on-your-own-data. AI hits 25% of March layoffs. The EU's child-safety amendment lands.
You've inherited a 1,500-line values.yaml. The fix isn't refactoring it in place, it's the six-step migration to a DaC shape: catalog, cluster, identify, push down, version, ship. Here's the walkthrough, concrete enough to mirror.
Every vendor is shipping a personal AI now. Almost none of them can answer the basic ownership questions, who owns the model, the memory, the patterns. Without that story, personal AI is a marketing label on a vendor relationship.
Atlassian and Block stack the largest AI-cited cuts of the cycle, the Challenger numbers put AI at 25% of US March layoffs, Anthropic sues over a Pentagon supply-chain-risk designation while Google quietly takes the contract, and NVIDIA opens GTC with $1T in Blackwell-plus-Rubin orders booked.
The 2026 audit conversation about AI usage has gotten sharp. The questions are sophisticated, the evidence asks are specific, and most shops can't produce what's being asked for. Here's what the conversation actually sounds like and where the gap sits.
A year ago I wrote about treating each AI interaction as its own bounded unit, own context, own audit, own memory boundary. Twelve months of building inside that pattern is enough material to grade what held up, what didn't, and what I missed entirely.
MWC Barcelona puts agentic AI in every booth, GPT-5.4 ships mid-week, Oracle telegraphs the largest layoffs in its history to fund AI capex, and the EU Council adds non-consensual deepfakes to the AI Act's prohibited list. The week the cuts started being about funding compute, not replacing workers.
MCP won the integration layer. That doesn't mean every tool integration in 2026 should be MCP. The honest architecture is MCP for 80% of cases plus a small set of deliberate escape hatches for the 20% where the protocol shape doesn't fit. Worth being specific.
MWC Barcelona pushes on-device AI back to the front of the stack, AMD and Nutanix sign a multi-year agentic-AI deal that makes the on-prem enterprise stack real, HyperNova 60B drops on Hugging Face, and the state-AI-law map keeps moving while the federal preemption clock ticks.
A year of using AI in daily work. Concrete cases where it earned its keep, concrete cases where it didn't, and the pattern that explains the difference. Worth being honest about both halves of the practitioner experience.
The playbook for a single-tenant to multi-tenant migration done well, additive schema, identity backfill, dual-path policy enforcement, row-level security, observability for both paths, traffic shift, deprecation. What I wish I'd known on day 1.
