“When security lives in a separate silo, every modernization project becomes a negotiation,” Benito says. “When it lives in the same budget line, the trade-offs are visible and somebody is accountable for all of them.” Over the past several years, that single budget has shed $1M a year through modernization, resource optimization, and vendor renegotiation—a 25% reduction in expenses—without anyone being able to argue that security paid for it, because security was at the table deciding. The newest line in that budget is AI consumption. I treat it the way he treats cloud: a managed spend with usage-based controls. “Not a science experiment that shows up on an invoice.”
The night the pager went quiet
When I started, availability sat at 94%. To an outsider, the number sounds fine. It isn’t. Ninety-four percent is roughly three weeks of downtime a year, spread across the moments a business can least afford.
His team moved 65% of on-premises services to cloud and hyper-converged platforms. The spreadsheet version of that story is a 55% cut in total cost of ownership, worth $2.2M a year. The version I tell is about the pager. Availability rose from 94% to 99.99%, and the team became 20% more efficient—not through headcount, but by letting the platform watch itself. AIOps anomaly detection catches drift before a human would. Predictive capacity management buys hardware before it’s needed rather than after. Automation handles the repeatable fixes at three in the morning. “That was the moment the organization started believing technology could be a source of calm rather than a source of surprises,” he says.
One set of numbers
Every company has the same meeting. Finance brings a number, marketing brings a different one, and operations brings a third. The first twenty minutes go to arguing about whose spreadsheet is right. I wanted that meeting to end. The answer was a single enterprise data foundation: an eventhouse for real-time events, a lakehouse for everything else, and streaming ingestion feeding both. More than 100 source systems across booking, member, and supplier domains now land in one governed layer. Redundant data marts that had accumulated over the years were retired, and report build time fell 40%. The part that took longest wasn’t the plumbing. It was agreeing on definitions. What counts as a booking? When is a member active? The team standardized the semantic layer so finance, marketing, and operations report from the same figures. “Now the meeting starts at minute one with the question that matters: what do we do about it?”
A paved road for AI
“The fastest way to kill AI in a company is to say yes to everything. The second fastest is to say no to everything.” I did neither, and I built an operating model.
Any business unit with an idea goes through the same intake and triage. The idea receives a risk tier, a human-in-the-loop requirement matched to that tier, and a set of evaluation gates it must clear before production. Every model and every agent lives in a maintained inventory. The whole system is certified to ISO 42001, meaning an outside auditor has confirmed that the company runs it the way it says it does. For business units, the result is a documented path from idea to production, with the guardrails already built in. Where has it landed? Service-desk triage and resolution, where AI now handles the first pass. Product-defect detection. Supply-chain and travel-logistics optimization, where the math is hard and the payoff is immediate. Each case is funded against a defined P&L payback and tracked after launch, not just before. Under the hood, it’s a multi-model architecture. “I’ve watched too many companies bet the house on one vendor,” he says. “We route each workload to the model that fits it, and we can swap.” The delivery engine is the same one that runs everything else: SRE, DevOps, and engineering teams that build AI-first from the start.
From no certifications to a sales advantage
When I arrived, the company held no security certifications. Enterprise prospects would send a security questionnaire, and the deal would stall for weeks while answers were assembled from scratch.
He built the security program from zero. Today, the company holds PCI DSS 4.0, SOC 2 Type I and Type II, and ISO 27001 certifications, sustained through annual audits. “Compliance stopped being a cost center the day it became a contract prerequisite,” he says. “The certifications don’t sit in a drawer. They unblock enterprise deals.” The foundation is zero trust on Microsoft Entra ID: multi-factor authentication everywhere, conditional access that evaluates context before granting anything, and privileged access management so that admin rights are borrowed for a task rather than owned forever. Least privilege is the default posture across the estate. Day to day, the security operations team runs 24x7 detection and response on Microsoft Sentinel and Defender, with AI-assisted triage so analysts spend their time on the alerts that matter. Continuous vulnerability management and third-party penetration testing keep the program honest. GDPR- and HIPAA-aligned frameworks, disaster recovery, and incident-response readiness mean the plan exists before the bad day does.
I present risk to the board in business-impact terms. “Nobody on a board wants a vulnerability count. They want to know what could stop revenue and how long it would take to recover.” He owns the cyber-insurance strategy for the same reason: it forces the conversation into dollars. And he governs third-party risk across payment processors, travel-supplier integrations, and SaaS vendors, because “the next incident is as likely to arrive through a contract as through a firewall.”
