My craft is building the systems that give people agency over their best work; AI is the latest tool in the kit.

Former colleagues facing layoffs needed a consistent, structured way to assess their market position, not one-off advice that varied by who they happened to ask. I designed a six-step analysis workflow producing a structured, citable PDF report for each candidate: market fit scoring, adjacent title mapping, live-searched skill demand ranking, cited learning paths, and line-level resume rewrites. A governing system prompt enforces tone, citation requirements, and output format across every analysis, functioning as the operational playbook for the program.

A 12-week career transition with no team, no manager, and no built-in structure needed the same rigor I'd apply to any ambiguous initiative. I structured it into six consecutive 2-week agile sprints, capped at 25–30 hours per week, governed by a 6-dimension personal rubric. Gemini served three distinct roles: structural planner, interactive learning coach, and deliverable auditor. NotebookLM organized large documentation volumes into queryable reference tools.

My resume tailoring process had settled into a reliable conversational pattern: classify the role, audit for real gaps before drafting anything, build the document, QA it. The question wasn't whether more of that could be automated. It was how much automation was appropriate, given that the entire premise rests on not fabricating or inflating anything. So before touching any tooling, I split the workflow into judgment steps that need a human in the loop by design, and mechanical steps that don't. That split became the constraint for everything after it: automate the mechanics, keep the judgment conversational.

Why it matters: The Skill cut average time per resume from about 45 minutes to about 8, which is what made it realistic to tailor 27 resumes in three weeks without shortcutting the audit step. That volume ran alongside a separate shift, from spray-and-pray applications to deliberate, JD-matched targeting.

As the first manager hired after an acquisition, I managed up to 22 reps. Calls were being monitored sporadically and scored inconsistently, manager to manager, classic variation waste hiding behind "it depends who grades you." I built a rubric that defined what a "good" call actually looked like and used it as the basis for regular 1:1 coaching, starting with my own team. Reps began requesting transfers to my team "because you always know what's expected with her." The results were visible enough that leadership created a standalone call-quality role off the back of it, and the model was eventually adopted across the company's other call centers.

Why it matters: Give me a team with no shared standard and I'll build one fast enough that the results make the case for scaling it.

My BAs reported to me but worked day-to-day on agile teams I wasn't part of. When performance issues came up, their team leads defaulted to working around a struggling BA rather than raising it, which left me without the specifics I needed for a real coaching conversation. I built a self- and peer-assessment matrix, adapted from the IIBA skills standard, that surfaced objective gaps between how a BA saw their own performance and how their team actually experienced it.

Why it matters: I build tools that outlast my presence in the room. This one's still shaping professional development, hiring and coaching decisions.

Leaders across Product and IT kept pulling me in informally to solve process and performance gaps that neither org owned, with no budget and no standing process behind the work. At the same time I was standing in as scrum master for a struggling UX content design team that couldn't reliably prioritize or estimate incoming work. I used that engagement as a live pilot: moved the team from kanban to scrum and out of email and spreadsheets into Jira, then tracked the outcome. Once the results were demonstrable, I built the business case and brought it to leadership. They funded a permanent Process Improvement Manager role, and I executed against that mandate from there.

Why it matters: I can see the gap before anyone's named it. But I'd rather quantify it and get it funded than freelance a fix without a mandate. That's the version that scales past me.

New hires, mostly incoming Product Owners and Business Analysts, were showing up without a laptop or system access because IT requests went untracked until someone remembered to follow up. Meanwhile hiring managers, some brand new to managing, were burning hours repeating the same live walkthrough for every incoming hire: where to find the holiday schedule, which sites to bookmark, how the team actually worked. I built one shared artifact that both people used instead of two separate, disconnected processes.

Why it matters: Cut hiring managers' admin time by roughly half and gave new hires the structure to ramp up on their own terms, less oversight, more ownership from day one.