I've always been a context engineer.

My professional story runs through in-house enterprise creative ops—starting at Adobe more than 15 years ago.

Every role I've ever had has had me invent or re-invent everything.
How processes are run. How tools are used. How delivery happens.
The context that's always been required to define and build:

How. Work. Works.

This isn't a designer's portfolio. It's an operator's.
It's a memo about how I think and how I approach the work.
With receipts one click deeper.

In the early years at Adobe, we had no process—no intake, no resourcing model, no workflow, no tools, no tracking. We built those from nothing, bringing brand creative production back in-house at scale.

Nothing about it was revolutionary—it was simply the earliest and plainest version of the job I've been doing ever since.

The formula hasn't changed since. Break the work down, find the parts that repeat, write it all down—with the rules and edge cases—transform and modernize, build it into the tools, and lead the people responsible for running it and delivering every day.

In-house or agency, the work is the work.

The details have changed every time.
How people do work never has.

I've run this from both sides of the brand-agency arrangement—in-house at Adobe, then five years as the production partner at Merkle and Aquent.

In-house meant Studio operations and the production pipeline underneath it: dozens of agency partners, hundreds of projects a year across content marketing, demand, campaign, web, and video. Then operations and program management for the enterprise web team, where quarterly planning covered thousands of projects, tackled by cross-functional teams including UX, design, copy, engineering, and publishing.

Most recently, I ran an agency portfolio—ServiceNow, Google, Amazon, Walmart, Marriott—and the delivery organization behind it. The P&L, the forecast, the capacity reporting, and the margin on every engagement were my responsibilities.

And over the years, all of it global. In-house localization processes branched off core content projects with their own review and approval steps. Regional web teams came into one intake and one queue. Follow-the-sun agency teams stood up to answer always-on clients around the clock.

The measure I care about is whether people grow into leadership of their own.

At Adobe I built and led the studio managers who sat between strategy and execution. They drove the work, owned the stakeholder relationships, and had standing authority to raise a hand when creative missed the brief or the brand.

On the agency side, delivery ran through engagement leaders who owned their accounts outright—hiring, resourcing, performance, reporting. What they knew I wanted from them was a heads up before an escalation landed, not after.

Those leaders were mine to manage. The people actually creating and delivering the work almost never were—the creative teams, the execution teams, the stakeholders across every function.

That part of the job runs on no management authority at all.

Day to day I sit a click above the work—working with the leaders on the other side, who sit a click above it too. But the most strategic, highest-priority programs didn't get delegated. They stayed mine to drive.

At Adobe there were two in-house Studios—one for the consumer business, one for enterprise. I sat enterprise. A reorg handed us product ownership of the corporate marketing DAM, and our VP gave the go-ahead to split from the centralized model—to define our own processes and choose our own tools. What we defined became the content supply chain.

We chose Workfront for work management before Adobe acquired it. Suddenly a creative ops team inside a Marketing org was also a product organization: backlogs, user stories, external development partners, bugs to squash, users to onboard, KPIs to report.

There was no in-house engineering.

The whole thing was mine to personally deliver—three workstreams, three vendors, all of them mine to drive every day.

The processes came first—and not only ours. Acquisitions had folded standalone in-house teams into the enterprise side, each with its own way of working and its own tools.

Out of many ways of working, one had to be defined.

What came out of it was context engineering, years before the term existed. Sanitized replicas of the actual artifacts:

Only then can the building begin. Which at Adobe meant manifesting the newly defined processes in Workfront and AEM Assets.

Taxonomy work to have the tools call the same things the same thing. Exhaustive metadata capture not limited to a single step at the end, but spread out across the entire project lifecycle, so that every asset published to the DAM already carried all of its tags. User types identified and provisioned with the right level of access, from requesters to creators to publishers, both internal and external.

One platform gave all of marketing one point of intake and a single source of truth.

Asset volume tripled in a year and the same team absorbed it.

And this original stack became the shape Adobe later productized as GenStudio. We told this story on stage at Adobe Summit.

Since leaving Adobe, the approach hasn't changed—the same patterns, from dry-erase whiteboards to digital whiteboards to AI agents. What changed is the cost. Transforming how work happens—and building the systems underneath it—used to be the expensive part. It isn't anymore. I've become fully AI-native in my day-to-day work, and I don't feel like any skill is out of reach.

Most of what I've recently built has been simple but valuable—automations that tied billing, forecast, timesheet, and capacity data together across systems that had never talked to each other, into one place that could tell a single story. Some carried real agentic reasoning—reconciliations that didn't add up, patterns worth flagging, meeting transcripts that became action items.

Where a platform published no usable API, I connected to it anyway, reading the browser's own session. Here's the no-API pattern, published.

None of it took an engineering team or a development budget.

It took Claude Code, MCP servers, solid context engineering, and good judgment.

• Happens onceWhat genuinely happens only once and never the same twice. It's unrepeatable and run once deliberately with one named owner.
• Happens every timeWhat happens every time. It gets written down—rules, edge cases, context—and operationalized.
• Repeats without judgmentWhat repeats but doesn't require judgment. It gets automated.

Take a first-of-its-kind live event.

• OnceThe broadcast window happens once—run deliberately, with one named owner.
• Every timeThe intake questions, the go/no-go call, the regional hand-off happen every time.
• AutomatedThe reporting underneath repeats without judgment.

swipe to see all three streams →

Before any of this, I produced live television news. There are no second chances there.

Live is live.
The deadline is the deadline.

The broadcast airs on time, every time—and how we met that deadline, every day, was my job. Was the rundown set? Were the scripts written? Were the packages delivered? Did everyone agree to the plan?

Backup plans, and backup plans for the backup plans. Something was always going to go wrong; my job was to have already decided what we'd do when it did. Then to put on the headset and see it through, live on the air.

The newsroom also set my temperature. I don't get too high, I don't get too low, I don't get rattled. The me you get is the me you get—every time, in every circumstance, under every pressure.