I'm a content designer. Lately that means writing the words and building the AI systems that scale them. Same goal it's always been: clarity, usefulness, and respect for the reader's time.
Across very different products, the work tends to move through the same three gears. They're how I think about almost every problem I'm handed.
Meet the mess where it is. Get the language clear, the patterns consistent, and the basics documented so a whole team can move faster than I ever could alone.
Bring research and coordination in early. Build the systems and the shared understanding before they're urgently demanded — so the fire drill never has to happen.
Once the foundation holds, use it to do something ambitious. Lately that's meant pointing content design straight at the model — designing how AI itself communicates.
The Trade Desk is a premium platform for buying digital advertising across the open internet. A month after I started, the company announced Kokai — a full, CEO-vision-led redesign of the entire platform, built on the periodic table of elements, shipping in six months, and full of brand-new concepts and terminology.
It was happening with or without me. So I moved through it the way I move through most things — practical first, then proactive, then provocative.
New platform, new vocabulary, and a clear need to scale my own understanding fast. I wrote the foundational documentation that turned a sprawling rollout into something a whole company could learn from.
Cut basic training from a 90-minute CEO recording down to four foundational help articles and four content tenets adopted company-wide.
Adoption was low and understanding was scattered — users were overwhelmed. I built a system of centralized quick references that revealed the right information at the right moment, per tool and per page.
Wrote and shipped 30+ quick-help guides; Kokai adoption climbed from under 50% to 80% in roughly five months, with more revenue flowing through the new platform.
If progressive reveal eased overwhelm in the UI, the same principle — and the same surface — could anchor an agentic AI experience. I worked with new AI tools and interfaces to start building it.
Adoption reached 90%, and content now leads the company's AI task force, with three AI implementations underway built on Quick Help's location and core progressive-reveal idea.
That last step is where the job changed: I stopped designing only the screen and started designing how the model itself writes.
Today my work at The Trade Desk runs across the whole AI content stack as one connected system. I write the system prompts and styling standards that define how our AI generates content. I built an LLM-based quality judge, integrated into our agentic production pipeline, that holds generated content to those standards without a human reviewing every piece. I develop the prompt strategy underneath it, and I build the internal tools — PRD drafting, release-note generation, platform audits — that scale the whole thing across the org.
The pieces feed each other. The tooling makes the styling work scale; the styling work gives the tooling something worth running; the quality judge keeps the loop honest. It's a system that gets better as it runs — and it's still content design, just pointed at the model instead of the screen.
<img>.Collective Health is a third-party administrator — it sits between the employer, the insured member, the insurer, and the provider, working to make health insurance less miserable. Open enrollment is the highest-stakes window of the year, and content design touched every part of it.
I was new, learning, and it was a lot. I made the necessary fixes to our 2022 OE deliverables and refreshed whatever we could improve in the time we had.
100% on-time delivery — plus a research initiative spun up to actually understand how the program performed, so 2023 could be better by design.
We needed baseline research and far better cross-team coordination. I moved the OE kickoff to March instead of the usual September, and partnered with research to measure the program honestly for the first time.
100% on-time delivery for 7/1 OE clients — and research that surfaced the real opportunity: a 7% post-OE registration rate and low brand recognition, with enormous room to grow.
Without control of the program, measurement and brand consistency were impossible — and we barely communicated with members at all. I built a pilot to recreate OE around consistent member communication, responsibly, without burning out the team.
Won executive approval for the pilot and a more flexible, creative content ecosystem. The program lifted enrollment enough that it was expanded after I left, and contributed to winning the company's largest client to date.
The win wasn't a single screen. It was convincing an entire company that a member worth insuring is a member worth talking to.
Not every Grubhub order is perfect, and our care for customers was what set us apart. But as the company grew, that level of service got harder to deliver. The idea was a chatbot — and from my time at Facebook, I knew bots are tough to do well. Which is exactly why I made the case to our Chief Product Officer that I should co-lead it.
I started by doing what content designers do best: I wrote. The resulting playbook became our guide for selecting a vendor and building the bot. I led the technical and data requirement docs, mapped each outcome backward from its resolution, and wrote content that was honest about what the bot could and couldn't do — conversational, but upfront about its limits, handing off to a human the moment it hit one.
The 25% ticket reduction was the headline, but the more interesting result was what the bot taught us. The raw numbers couldn't tell us whether people got help or just gave up — so I read the individual conversations, used the bot's CMS to experiment with content in real time, and eventually migrated the whole experience into the product itself, where order data could make it genuinely seamless.
It's the through-line to everything I do now — I've been designing conversations between people and systems since long before the systems could talk back.
UX writing rarely lets me tell a story. So I've been lucky to work places that let me be me — satire, essays, the odd open letter — all aimed at shedding a little light, expanding a thought, and entertaining a few folks along the way.
Stakeholder management at Google was staggering in scale. Trying to laugh about it made it feel a little less brutal.
Read →I'm a nervous traveler. But I've never felt more enlightened by a work experience than I did on a research trip to Thailand.
Read →I got tired of watching people make more work for our contract coworkers by misusing the disposal bins. So the bins wrote back.
Read →A reaction to a truly absurd row about making cafeteria to-go boxes marginally harder to get. Told from the boxes' point of view.
Read →I wanted to give content strategy some color for the non-writers at Grubhub — so I made it a fun, informative series. (Link needs updating — the URL in your deck is broken.)
Add link →For about twenty years I've helped build and grow businesses through content — at small startups, agencies, and some of the largest, most-scrutinized companies on the planet. I've led teams and worked alone, mentored and been mentored, and tried at every stop to make the complicated feel simple.
The common thread is a stubborn focus on the person on the other end of the screen: is this clear, and does it actually work for them? That question got me from payment flows and chatbots into the AI systems I build today. The tools changed; the question didn't.
Underneath the work, I care deeply — and I've learned how far a little levity can go.