The Unglamorous Systems Work That Funds Your AI Strategy

Before I led agentic AI programs, I spent over a decade inside the systems that run enterprises — GTM and CRM platforms at Fastly, enterprise architecture at F5 and GE, Fortune 500 consulting at Oracle. The most valuable thing that decade taught me has nothing to do with models: the boring systems work is what funds, and ultimately decides, your AI strategy.

Lead-to-Cash: the cycle nobody tweets about

Lead-to-Cash runs from the moment a lead is generated to the moment payment lands. It crosses every core system a company owns — CRM, CPQ, contracts, ERP, billing — and every handoff between them is a place where revenue waits.

At F5, we shortened lead-to-revenue operations from three months to two weeks — not with new headcount, but with active notifications, redesigned approval processes, a proper subscription engine, and a deal cycle made visible through self-serve, real-time reporting. Deployments went from hours to minutes. None of it was glamorous. All of it showed up in cash flow.

Here's the connection to AI that most strategies miss: that same work — clean integrations, automated handoffs, visible state, fast deployment — is precisely the substrate agents need. An AI agent can only act on systems that are reachable, automatable, and observable. The Lead-to-Cash work is the AI-readiness work.

Three transfers from systems engineering to AI strategy

1. Cycle time is the metric that matters. We measured Lead-to-Cash in days removed, not features shipped. AI programs should be held to the same standard: how much faster does the lease get signed, the ticket get resolved, the quote get approved? "Agent deployed" is an output. Days removed from a cycle is an outcome — and it's what a CFO will fund again next year.

2. Integration debt compounds; so does integration investment. At Fastly I was responsible for integrations across dozens of platforms — marketing automation, CPQ, contract management, revenue intelligence, support. Painful, perpetual work. But every integration built then is a tool an agent can use now. Companies that skipped that work are discovering their shiny new agents have nothing to hold onto. Your integration map is your agent capability map.

3. Discipline scales; heroics don't. The practices that turned around those platform teams — CI/CD, behavior-driven development, mob programming, automated QA growing from 5% to 30% coverage in a quarter — are the same practices that make AI systems shippable: versioned prompts, eval suites as regression tests, continuous deployment with feature flags so a misbehaving agent can be rolled back in minutes. Teams that never built deployment discipline for ordinary software will not magically develop it for autonomous software.

The career version of this argument

There's a generation of executives who came up entirely on the AI wave and treat enterprise systems as someone else's plumbing. And there's a generation of systems leaders who treat AI as a feature to be procured. The leverage sits with people fluent in both — who know what a CPQ approval chain actually looks like and what an agent needs to traverse it safely.

That intersection is where I've chosen to operate, and it's where I'd advise any technology leader to position their organization: fund the systems work without apology, measure it in cycle time, and let your AI strategy stand on infrastructure that can actually carry it.

The companies that win with AI won't be the ones that skipped the unglamorous decade. They'll be the ones that made it compound.