The press release reads like a LinkedIn thought-leadership post written by committee. "Enterprise AI stack is being rewritten in real time." "Decision-makers are starved for objective, defendable data." "Architectural blueprints, actual deployment barriers, back-end infrastructure realities." That's three paragraphs of analyst-speak for "we're going to ask vendors uncomfortable questions about why their shit doesn't work in production."
Strechay's focus areas: cloud infrastructure, advanced data infrastructure, platform engineering, DevOps orchestration, observability, and the intersection where AI and enterprise security collide. Translation: the messy middle where proof-of-concepts go to die.
He's already published on enterprise GPU utilization — the compute waste sitting inside AI infrastructure. Also reviewed the AI Infrastructure & Compute survey before it went live. The monthly VB Pulse surveys track five adoption areas: agentic orchestration, agent reliability/evals, agentic security/identity, AI infrastructure/compute, and context layers including RAG.
June's agentic orchestration report surveyed 145 enterprises. Two-thirds hedged their model strategy instead of committing to one provider. Smart move — the Anthropic Claude outage that same month proved why. Nothing validates multi-vendor strategy like your primary model going dark during a demo.
The VB In Conversation video series gets Strechay as host. Promise: "unvarnished look at which tools perform under production-grade pressure." I'll believe it when I see a vendor admit their vector database falls over at 50 QPS on camera.
Strechay's quote: "VentureBeat has built an audience of enterprise builders and technology buyers that any analyst would want to serve. My goal is to use deep empirical metrics and VentureBeat's proprietary tracking data..."
The sentence cuts off in the source. Fitting metaphor for most enterprise AI projects.
What actually matters: can this research arm call out the gap between vendor slide decks and 3 AM pages? The Pulse surveys are a start — proprietary data beats analyst opinions backed by "trust me bro." But the real test is whether they'll publish the uncomfortable numbers: utilization rates, failure rates, actual vs. projected ROI.
Enterprise buyers don't need more "architectural blueprints." They need to know which vendors ghost you after the PO gets signed.
Next MCP server runs on Android →