What IT Gets Wrong About AI Maturity
Episode Overview
In the first Hot Takes of 2026, CCO Sean Hoyer and COO Ari Stowe take on the growing debate around AI hype, maturity, and real-world impact on IT. They hash out why expectations for agentic AI outpace reality and why it should be treated like an employee instead of a silver bullet. They also unpack human-in-the-loop models, AI reasoning’s (considerable) limits, and why better outcomes hinge on rethinking operational maturity!
Key Takeaways
- “AI maturity” isn’t about flipping a switch to fully autonomous “see/think/do/fix” systems. It’s actually about adopting the best currently available tech in tightly scoped use cases, as well as building trust over time.
- The biggest adoption failure mode is expecting general copilots/LLMs to deliver deep, domain-specific operational expertise; agentic AI has to be applied inside real business processes and guardrails.
- Human-in-the-loop is an adoption and trust-building mechanism, where AI starts by gathering context + recommending actions before graduating to higher-impact execution.
- Many IT teams are now solving problems they created: hyper-instrumented monitoring produces overwhelming noise, so the real challenge becomes finding signal and optimizing for customer/employee experience outcomes.
- AI reasoning has meaningful limits on net-new problems; the practical path is using AI to correlate context, assist decision-making, and excel in patterns you’ve seen before, not to magically solve unprecedented incidents.
FAQ
Q: Are enterprises actually “boiling the ocean” with agentic AI expectations, or is the hype backlash overcorrecting?
A: The hype backlash can overcorrect. While some buyers have unrealistic “it should build workflows on the fly and touch critical infrastructure” expectations, most teams are trying to start slow and scope agents to realistic, high-value work rather than expecting a silver bullet.
Timestamp: 2:40–5:56
Q: Why is “just buy Copilot/ChatGPT/Claude” not a real strategy for IT operations?
A: General LLMs can be great at generic knowledge work, but they don’t automatically bring the domain-specific context required for complex operational environments (like your specific network, configurations, and rules). The value comes from putting agents into domain-specific workflows and connecting them to the systems and processes where work actually happens.
Timestamp: 8:01–10:25
Q: What do massive AI infrastructure investments actually mean for enterprise teams?
A: It likely points toward an “AWS-for-AI” model: core vendors provide models and AI services as infrastructure, while companies build domain-specific products on top. That can accelerate builders, but it also introduces real cost/complexity tradeoffs (power, cooling, and operational “hidden taxes”).
Timestamp: 36:48–45:52
