Agency is not the ability to work harder. It's the ability to shrink the distance between an intention and its execution. In a world built to reward specialization, the most valuable competitive advantage isn't a deeper skill set — it's a better system. True agency arrives the moment you stop being the engine and start being the architect.
The Multi-Track Friction
In my current role, my daily reality has looked like this: mornings in Operations at an urban mobility company — answering immediate client and customer needs, checking staff schedules, clearing the urgent before it becomes a crisis. Afternoons in Business Development — outbound calls, pipeline reviews, persuasion at scale.
Two roles. One person. Opposite demands.
The Ops version of me is fixing the present. The Sales version is hunting the future. In the middle sits a bottleneck: me. Every hand-off between the two lives in my head, which means every context switch costs something — time, energy, cognitive sharpness.
I tried to manage my way out of it. Better calendar systems, stricter time blocks, smarter prioritization. None of it worked at the level I needed. The problem wasn't efficiency. It was the ceiling of human agency: I could optimize myself to the edge of burnout and still not escape the fundamental bottleneck of being the connector between my own roles.
I've Been Here Before
This friction isn't new to me. It's a pattern I've run before — just with different tools.
In 1998, I built a production facility from a green site in the food packaging sector. One ongoing task, hundreds of sub-tasks, and constantly evolving requirements as the business grew. My solution was the same instinct I'd reach for today: think in systems. Make every process as repeatable and boring as possible. Delegate. Supervise outcomes, not steps.
By the time I walked away, that approach had produced three binders — ISO 9001, ISO 22000, and ISO 14001 — and a production facility running on its own momentum.
Before that, I coordinated operations for a language school: one administrator and three sales staff managing the front office, me running the backend with twenty teachers. Different industry. Same architecture. Hold the system, delegate execution, supervise outcomes.
Those binders were my harness and my evals. The teachers were my agents. I just didn't have that language yet.
The method hasn't changed. The constraint has. Building that kind of agency used to require a whole team — and human teams are expensive. They have bad days. They misalign. They leave. Finding a team of consistently high-agency people is one of the hardest problems in any organization. Today, the constraint is configuration, not headcount.
From Automation to Orchestration
The way out started as a shadow project — something I was building in the margins, not yet sure what it would become.
It started with automation. If-This-Then-That rules: auto-log CRM updates, route support tickets by keyword, trigger follow-up sequences. Useful, but narrow. Linear automation handles clean inputs and predictable steps. It breaks down at the messy, high-context handoffs — a sales call where a client made an unusual request, an ops issue with no clean category.
The pivot came when I stopped building automations and started building an orchestration layer — a system of AI agents that reason across context rather than execute rules. And that required a different kind of thinking: not "which AI tool is best," but "which AI is right for this specific job."
For BD, I use Claude to analyze call outcomes and build systems, and Gemini to research prospects and scrape the web. For the Business English learning platform I'm building in the margins, I chose Claude as the senior developer and Gemini as the curriculum expert — specifically because Gemini's training made it a stronger EFL resource. Two different models, evaluated for their strengths, are working in the same system.
Here's what that looks like in practice: a complex sales call ends on Teams. My orchestration layer reads the transcript, identifies every operational commitment I made, cross-references current support load, and drafts a hand-off summary — before I've opened another window. No manual note-taking. No context lost in translation between my Sales and Ops selves. The system holds the thread, so I don't have to.
That same shadow project eventually became the platform itself — AI-powered, built for adult English learners, running alongside the day job. The same orchestration principle applies: lesson generation, student feedback analysis, and curriculum sequencing — all running through a system I oversee rather than manually operate. The two tracks no longer compete for the same hours. They run in parallel, connected by a layer that understands the intent behind both.
The Multi-Track Operator Without a Lever
The multi-track life is becoming the norm. Day jobs alongside companies in progress. Creative work alongside operational work. The single-track career is a 20th-century artifact.
But the multi-track life is a trap if you remain the primary laborer on every track. Without a technical lever — an orchestration layer that understands the context of your different roles — you're not running multiple tracks. You're just exhausted.
The old path to this kind of agency required a team: coordinators, administrators, sales staff, and production crews. Human teams are expensive, inconsistent, and hard to scale. Agents don't have bad days — once they're properly configured. The new constraint isn't headcount. It's clarity of intent and quality of setup.
The next generation of professionals won't be defined by how much they can do. They'll be defined by how effectively they can orchestrate: building systems that carry their intent across contexts, hand off work intelligently, and multiply output without multiplying hours.
I'm not managing Sales and Ops anymore. I'm overseeing the system that connects them — and using the same architecture to build something of my own on the side.
The goal isn't to be a better cog. The goal is to build the machine, and then step back and architect what comes next.