Across business units and functions of a ~250-person product org, from field research with sales agents to portfolio-level investment strategy
Team
Solo IC who built a working team for each initiative: head of design, senior UX researcher, senior product designer, staff product manager, head of engineering
Industry
Agriculture
Too busy running a business for our tools
I was asked to get more data from the few sales agents already on the platform. I found most were too busy running their businesses to use it, and turned the roadmap toward helping them run those businesses.
$50Ma year, allocated through the planning process I redesigned around sales agents’ jobs
Research, a jobs framework, a new practice and a new planning process, each reinforcing the others, changing who the roadmap served, what got measured and where $50M a year went.
The call I made
Engagement was low, and the roadmap was aimed at the sales agents already on the platform, with features built to collect more of their data.
I changed the question to why most sales agents weren’t there. The commercial team’s plan to merge agencies was turning them into business managers with no time for field tools. So I looked at the whole territory and the agency they run.
The business still needed farmers’ data, and farmers were wary of sharing it. My bet was that they would share it more comfortably as their relationships with their sales agents grew stronger, with more trust and loyalty.
The hard part
My first answer, journey management, didn’t fit. Sales agents’ days are run by weather, disease, commodity prices and a phone that can’t go unanswered. So I switched to jobs-to-be-done: design for the outcomes they want, with solutions that flex with an unpredictable day.
With no dedicated budget, I was asking an organization used to planning from its software to start from the sales agent’s day.
What it changed
I founded a Design Strategy Center of Excellence to work a year ahead on who we were building for and why. I also redesigned annual planning around the jobs sales agents are trying to get done, so funding decisions started from their needs.
The year’s investment white papers were approved. They turned the roadmap toward the sales agent’s whole territory and toward running the agency as a business. The aim was to give sales agents their time back.
Without a mandate, I spread the method through allies I mentored and a pilot Persona Experience Team that my head of design and I set up. So the sales agent’s view showed up in rooms and studies I never touched.
Engagement later closed at 44% against a 35% goal, and adoption at 52% against 50%.
What I learned
People rarely move through the world in straight lines, but software tends to be built as if they do. It’s tempting to force that messy reality into a tidy artifact. The mess is how real change unfolds inside a system. Part of the work is helping people get comfortable enough to work with it.
Booz Allen02Booz Allen · Military health careSenior Design Strategist
Role
Senior Design Strategist
Scope
Eleven military treatment facilities and their leadership, from hand-tagging patient comments to network-level priorities
Team
With a design strategy partner and two data scientists we brought on, alongside the client’s chief experience officers
Industry
Military health care
Killing our own project
I inherited a printed booklet to help patients find their way. The bigger problem was that leaders could only act on the few patient comments someone had time to read. I set the booklet aside, and we built the tool that let eleven military treatment facilities hear all of them and see where to act first.
Adoptedacross the facilities, outlasting our engagement; escalated to the Defense Health Agency
One tool put 100,000+ patient comments into a single view for leaders across eleven treatment facilities, showing each facility against the others and moving them to act on what mattered first.
The call I made
I joined a project for eleven military treatment facilities around Washington, D.C., partway through. I inherited a printed booklet meant to help patients find their way through the hospital.
In an interview, a chief experience officer scrolled through the patient survey for us. It held over a hundred thousand open-ended comments, in a raw Excel export. Leaders could only act on the few someone had time to read, and critical problems sat in the rest.
I pushed us to set the booklet aside. It asked more of patients who were already in pain. My bet was that fixing the problems behind their experience would do far more over time. It would give the network a way to hear the whole story and know where to start.
The hard part
I had to bring our client with us. With my design strategy partner, I built the case and a prototype made from the client’s own patients’ comments. The prototype won it.
This was before large language models were everywhere. We brought on two data scientists. I hand-tagged the sentiment of thousands of comments so the natural language processing had something to learn from.
What it changed
We built a dashboard where every new batch of comments arrived sorted by topic, sentiment and critical issue. It showed each facility’s performance beside the others’. In the prototype, the deputy director saw bathroom cleanliness among the top ten pain points and acted on it right away.
It was adopted across the facilities, outlived our engagement, and was escalated to the Defense Health Agency as something to fund and scale across all of its facilities.
What I learned
Do the primary work as well as it can be done, without becoming so focused on it that you miss a bigger opportunity beside it. When new evidence points somewhere bigger, have the nerve to leave the path you were on.
Amazon03Amazon · Delivery pickup and returnsSenior UX Designer
Role
Senior UX Designer
Scope
The whole pickup service, digital and physical, in the US and Spain, from ride-alongs with drivers to the organization’s definition of its customers
Team
With my manager, and partners in product, engineering, UX research, marketing and business development
Industry
E-commerce customer experience
One definition of the customer, across the whole service
A sliver of the checkout page raised a bigger question: who were our customers? The answer became the organization’s shared definition of them.
9customer archetypes, adopted as the organization’s shared definition of its customers
One small slot on the checkout page opened up lockers, stores, drivers’ routes and returns, and surfaced customers the service wasn’t planning for.
The call I made
In a meeting about our section of the checkout page, it became clear we were guiding a customer’s choice between pickup and home delivery on assumptions rather than evidence.
Pickup was being built one feature at a time, and our picture of who used it, and why, was blurry. I moved from feature work to service design for the whole pickup network. I gave up quick feature wins to find out who our customers were and what they were counting on.
The hard part
Service design had no charter on the team. I made room for it with my manager, and by building relationships across product, engineering, research, marketing and business development.
The world was bigger than the one we pictured. We found lockers customers were afraid to visit, addresses given as landmarks, and expectations that differed by culture.
What it changed
A wall-sized service blueprint connected the Last Mile and access point teams and let people see how the whole system fit together.
I defined nine customer archetypes and what each one needs, the Durable Customer Outcomes, to show how the service should serve each. The organization adopted them as its shared definition of its customers.
My director approved a multi-city research proposal, aimed where the business was already expanding. He called it the department’s most important research for the next year. With my manager, I co-wrote a three-year plan for the design team. Both were approved, and layoffs stopped both before they could start.
What I learned
The leverage point is different in every organization. At Amazon, a proposal moved when the evidence tied a customer outcome to investments the business was already making. I was improving the return on something already committed, rather than asking for something new.
A governed workspace where AI agents do real work against my files under written rules. Every decision that governs future work gets an ID, a date and a quote, so a rule made once is still in force a week later. The agents read the register before they act.
Assumptions get expensive when they become roadmaps.
I help organizations make better decisions about complex human systems.
Beyond the work
Five questions
What sent me upstream?
Early in my career, I was part of a product launch that failed, then failed again with each iteration. I saw the defeat that leaves on an organization and its people, and I wanted to get good enough at designing services and products to prevent it next time.
What am I actually interested in?
I love to untangle the tangible and intangible systems that dictate how experiences unfold. I'm especially interested in how organizations make decisions, which is really their epistemic operating model. The best external outcomes for customers start with making better decisions internally, and that's a hard thing to get right.
Why does this matter more now?
In this new age where AI is rapidly accelerating the building and output cycles, it matters more than ever to identify the right problems early, because the downstream effects compound. Coherent systems of knowledge and context become the most valuable currency inside an organization, and building those is the part I want to be doing.
How do I work with AI?
Mostly as something that needs governing. What I build are systems of connected information with rules about what is true, who owns which fact, and what has to be checked before anything gets acted on. Inside that, the work is closer to coaching than prompting: guiding, critiquing, running feedback loops, and noticing the small things that separate adequate output from good. The noticing is the part it still cannot do. It fills real gaps in my capability, mostly technical, and it lets me see a dozen executions in the time one used to take. The value there is not that it produces more. It is that it produces things I can reject quickly, and getting to the right answer is mostly a matter of discarding the wrong ones sooner. But it is confidently wrong often enough that the interesting design problem is the system around it rather than the prompt itself.
Where am I less useful?
Hand me a settled roadmap and one screen to perfect, and you'll get competent work from someone quietly going flat. The further I am from the decision, the less I'm worth, which is a strange thing to advertise, but better said now than discovered together in month six. I like repairing very large systems, and I'm not the person you want polishing a small one.
Tell me where you're trying to go. I'll tell you if I can help you get there.