Product manager who builds with AI.
I use AI to run the program and to guide the build. By day I run a ten-team release train and co-build the requirements app it plans from. On my own time I ship products with AI agents: I decide what to make and what to cut, write it down, and let the tests referee.
10vendor teams planning from one board
5ten-week increments on one plan
2program tools built with AI agents
4products shipped with AI agents, behind 9,000+ tests
Selected work
All work10 teams · contract to test
The RTM tool
A digitally native requirements traceability matrix
10 teams · 5 increments
The PI dashboard
When the work lands
3 AI agents, one workflow
Job Synthesis
One profile, three AI agents, honest applications
119 effects · 2,400+ tests
Audio Edit & Tag
A desktop sound editor with five live stretch engines and 119 effects
In progress · 1,454 tests
idMLab
The modular sibling of Audio Edit & Tag: every engine a node on one canvas
How I use AI
The full methodTo run the program
- Requirements: every contract requirement traced to its epics, designs, rules and tests. Links the tool inferred stay marked as inferred.
- Planning: one board for five increments. A gate with no data shows a dash, never green.
- Reporting: delivery reports and roadshow numbers rebuilt from exports, and reconciled before anyone presents them.
- People: I am the AI resource for the program’s business analysts.
To guide the build
- I decide what to make and what to cut, and write it down before any code.
- Tests are the referee, not the model’s word. A 180-item manual became 180 tests.
- Agents work inside written rules: scope, git limits, and the lesson from each time one overreached.
- I check the agents’ numbers. A reported 682 tests turned out to be 384.
Who I’ve delivered for
Each one opens the work.
Meta
A+E Networks
SunTrust
Toyota
PGi
Unilever
Citi