Approach
I study how people and systems behave, then build the structure to improve them
My operating model
Observe the workflow
I start with what actually happens, not the documented process. That means sitting with the people who do the work, tracing each handoff, and looking at the data the workflow already produces.
Find the friction
Friction shows up as waiting, re-keying, rework, unclear ownership, or users who quietly give up. I log each point with its evidence and cost so priorities come from data, not from whoever is loudest.
Align people and systems
Most fixes are part process, part tool. I get the right functions to agree on who owns each step, what good looks like, and where a person must stay in the decision.
Test the intervention
Changes go out in small, testable pieces with quality checks and test cases, including AI outputs, so we learn quickly without putting users at risk.
Measure and improve
A small set of operating metrics, reviewed on a regular rhythm, shows whether the change worked and what to do next. Documentation and training keep the improvement in place.
What Product Operations means in practice
Product Operations is the work that lets product teams make good decisions repeatedly: clear workflows, reliable data, defined handoffs, quality checks, and documentation people actually use.
In practice that means I spend as much time on how work moves between people as on the product itself. A feature that ships into a broken workflow rarely delivers what it promised.
Connecting user evidence and operational evidence
User evidence tells me where people struggle: observation, funnel data, support contacts. Operational evidence tells me why the organization struggles to fix it: handoffs, ownership gaps, manual steps, missing data.
I put both side by side. In the tuition calculator research, analytics showed where people left and session recordings showed why; in application monitoring, recordings explain what the weekly numbers can't. Together they make the case for change much stronger than either alone.
Where AI creates value in workflows
AI is most useful on high-volume, repetitive work where a person can check the output quickly: extracting data from documents, summarizing a situation, drafting a first pass, or watching many signals at once.
It is least useful as decoration. If a simple rule or a clearer form would solve the problem, I recommend that instead.
Responsible AI and human oversight
Decisions that affect someone's money, record, or access stay with a person. AI prepares; people decide. Reviewers see the source next to the AI output, can correct or reject it, and their corrections become quality signal.
I design the human side first — who approves, who can say no, where cases escalate, and what stops the system — and label every project honestly as a concept, a pilot, or launched work.
How documentation, QA, and measurement support adoption
People adopt what they understand and trust. Clear documentation, test cases that cover messy real inputs, and a small set of metrics reviewed on a steady cadence are what turn a launch into lasting change.
Templates and playbooks
Sanitized versions of tools I use. Free to download and adapt.
Weekly product intelligence brief
The structure I use to turn funnel and technical signals into a short weekly brief that teams act on.
Download template : Weekly product intelligence brief (Markdown file)Workflow friction map
A template for mapping steps, handoffs, and friction before proposing any tool or automation.
Download template : Workflow friction map (Markdown file)AI task suitability & human-review checklist
Questions I answer before automating a task: is AI a fit, where must a person decide, and what could go wrong.
Download template : AI task suitability & human-review checklist (Markdown file)
Have a workflow that should work better?
Bring me the messy one. I'll start by watching how the work actually happens.