When we decided to rebuild Tech CRG's operations around AI — not as a feature, but as the foundational layer of how we work — I expected certain outcomes. Faster response times. Lower cost per incident. Better throughput from our NOC and Service Desk.
What I didn't expect were the second-order effects. The things that changed that I didn't plan for.
What I expected
The case for AI in managed IT services is straightforward: AI can monitor more devices per operator, correlate events faster than humans, and surface actionable alerts instead of noise. The economics are obvious. The operational improvement is measurable.
We got those outcomes. Mean time to detect dropped. Alert fatigue dropped. Capacity per operator increased. The math worked.
What I didn't expect: the talent shift
The biggest surprise was what happened to our engineers. When AI handles L1 — the routine monitoring, the first-pass alert triage, the pattern matching — the humans in our operations center stop doing L1 work. And when that happens, you find out very quickly who your actual engineers are.
The people who were good at following runbooks and closing tickets had a hard adjustment. The people who were genuinely curious about complex problems — who wanted to understand why something was happening, not just how to fix the symptom — those people thrived. AI made their work more interesting, not less.
We changed how we hire as a result. We stopped looking for people who could follow a process and started looking for people who could think through a problem. That's a harder interview and a more expensive offer — but the quality of our team is fundamentally different.
What I didn't expect: the metrics changed
We used to measure our operations center primarily on volume metrics: tickets closed per day, first-call resolution rate, average handle time. Those metrics are still relevant — but they became less interesting as AI compressed them.
The metrics that started to matter more: mean time to detect (MTTD), mean time to contain (MTTC), and — most importantly — the number of problems we prevented versus the number we responded to. An AI-native operations center shifts the economics from reactive to proactive. Proactive is harder to measure, but the business impact is larger.
What I didn't expect: clients asked different questions
Before we rebuilt around AI, the questions we got in sales conversations were mostly about credentials — certifications, references, pricing. After, we started getting a different category of question: how does your AI actually work? What does it see that a human analyst would miss? How do you handle false positives?
Those are better questions. They reflect a buyer who is thinking about outcomes, not just inputs. And they led to better client relationships — because the clients who asked those questions were the ones who would measure us on the right things.
The honest caveat
AI did not solve everything. The problems it didn't help with were the ones that require human judgment, relationship, and context — client communication during an incident, architectural decisions that involve business tradeoffs, escalations that require someone to make a call under uncertainty.
AI handled the volume. Humans handled the judgment. That's the right division of labor — and understanding where that line is has been the most important operational lesson of the last two years.
Juan M. Delgado is CEO of Tech CRG, an AI-native technology company headquartered in the US with legal presence in 12 countries across the Americas.