Work where AI played a real role.
And where I had to tell it to calm down. Four projects. Honest about what helped and what still required a human who knew what good looked like.
Consulting Project · 2025–Present
When a client has a mission but no brand, and you have to build it from scratch
The work
Led brand strategy and identity for a faith-based ministry client: positioning, voice, visual identity, brand guidelines. Then built the website myself — copy, IA, and full implementation. No agency handoff.
What I delivered
A complete brand system: vision, mission, values, tone of voice, logo direction, color palette, guidelines doc. And a live site that actually reflects the brand, not just describes it.
Discovery + Drafting
Used it throughout discovery and drafting: exploring positioning language, running through messaging alternatives, getting to first drafts faster. It also suggested some directions that were very confident and very wrong. Every judgment call about what the brand actually stood for was mine.
AI moves the drafting timeline way up. Brand clarity still comes from the client conversations, not the chat window.
Personal Brand Project · 2025
Same experience, different hiring lenses. Positioning is the real resume opportunity.
The challenge
A static resume was not telling the full story. The same background needed to read as GTM strategy, campaign management, marketing operations, or partner marketing depending on the hiring need. That is a positioning problem, not a formatting one.
What I built
Multiple role-aligned resume versions, each with a distinct positioning angle. A live interactive site covering experience, case studies, and capabilities. GA4 tracking to understand how people actually engaged with the content.
Pressure-Testing + Iteration
Used AI to pressure-test the same background through different role lenses and accelerate bullet refinement across versions. Fair warning: AI will tell you your draft is great if you let it. You have to ask it to be mean. The final calls on what was honest, strategic, and actually compelling were mine.
AI is a genuinely useful sounding board for positioning. It is also a yes-machine if you stop pushing. Keep pushing.
Retrospective · Cricket Wireless / AT&T
A year of data, four markets, three metrics, and the one part AI still cannot touch
The original work
Led a year-long brand and performance test across four submarkets using control-store methodology. Measured impact on gross adds, consideration, and brand perception. Translated a year of mixed signals into a clear recommendation for leadership.
What made it hard
Multi-market, multi-metric studies generate a lot of signal and a lot of noise. The hardest part was not the analysis. It was knowing what the numbers actually meant for strategy, and making that legible to stakeholders with competing reads on the same data.
Research + Synthesis
Research and setup: AI would have accelerated submarket profiling and competitive context significantly. In-flight analysis: Useful as a synthesis layer — spotting patterns, flagging anomalies, structuring interim readouts. The recommendation: AI would have helped with structure and narrative. But the strategic call still required people who knew the market and the organization.
AI would have made this project faster and the outputs cleaner. It would not have changed what the work required most: reading a year of mixed signals and deciding what they meant for the business. That is not a prompt.
Personal Systems Project · 2026 (Ongoing)
Building an Operating System for a Job Search
The challenge
A job search isn't a writing exercise. It's an operations problem wearing a cover letter. New roles show up daily, priorities shift, and every application needs research, positioning, tailoring, and follow-up — on repeat, dozens of times. Do that without a system and you end up spending real effort on the wrong roles and sounding like a slightly different person in every cover letter.
What I built
A repeatable operating cadence, built into an actual tool, not just a process in my head. A live tracker (HTML/JS, no backend) with status filters, fit scoring, and pay range logged per role. A "Job Search Agent" panel where I set target titles, industries, and must-avoid criteria, then generate a structured prompt I run through Claude to screen and score any posting on the spot. Same criteria, every time, so nothing gets judged on vibes.
The actual tracker. Fictional companies shown here, real tool underneath.
Two Tools, Two Jobs
Not one AI doing everything. ChatGPT scores each opportunity against my criteria and gives me the editorial pushback, the outside voice that tells me when a draft is flat or a "fit" is wishful thinking. Claude builds the actual output: tailored resume, ATS version, cover letter, and runs the screener prompts against the posting. Two tools playing different positions. I'm still the one deciding what's worth acting on.
Turns out the same operating principles run a job search and a product launch: clear intake criteria, prioritize the right work, build a repeatable process, measure what happens, improve it. AI makes the machine faster, and having two tools check each other makes it sharper. It still can't tell me what's worth building. That part's still mine, every time.