I get AI into production with controls leaders can trust.
At work, I built an AI assistant at JPMorgan Chase and took it firm-wide. On my own time, I run five working systems built with AI agents. Both are below.
01 Professional
Twelve years making operations work better, now with AI
Financial services, healthcare and education. Today I'm a Senior Process Improvement Analyst at JPMorgan Chase.
- 70% less time to create a process model, with the AI assistant I built and scaled firm-wide at JPMorgan Chase
- 43% lower Claims team losses at PNC, forecast and reduced with executives
- 12% higher quality scores at Cardinal Health, plus 4 points of NPS
- +7 points of NPS at McGraw Hill from a customer insights program I launched
02 Personal
Five systems I built with AI agents
Each one is in real use, and each runs under the rules I'd expect of any team: clear scope, written limits, and independent checks.
- Product ยท in daily use Livi Home The kitchen tablet and phone app my family uses every day. It costs $0 a month to run.
-
Forecasting
FormulaAnalytics
A Formula 1 forecast that runs itself and publishes its own track record.
-
Decision support
Trade Finder
Five conflicting ranking sources joined into one answer on a phone.
03 More detail
The longer version
At JPMorgan Chase, 2022 to now
Creating a process model took long enough to be a longstanding barrier to doing it at all. I developed an AI assistant that cut the time to create one by up to 70%, then scaled it to firm-wide adoption. It earned executive recognition for removing a barrier that had been there for years.
- Governance. I established a center of excellence for process management and built automation for process model governance that cut manual work for clients.
- Senior alignment. I ran workshops with senior leaders and cross-functional teams to turn their priorities into execution plans.
- Delivery. I drove the migration from iGrafx to Signavio and hit its milestones early.
Work history
I led the Knowledge Management and Change Control teams at McGraw Hill, and a speech analytics community of practice at Cardinal Health that got leaders using customer sentiment data.
- JPMorgan Chase 2022 to now Sr Process Improvement Analyst
- Cardinal Health 2020 to 2022 Business Intelligence Consultant
- McGraw Hill Education 2017 to 2020 Process Improvement Analyst
- PNC Financial Services 2015 to 2017 Business Analyst
- Amend Consulting 2014 to 2015 Process Improvement Consultant, Lean Six Sigma Green Belt
How I manage AI work
- Define success before the work starts. The pass mark is written down before any test runs.
- Review independently. A second agent grades the output without seeing the reasoning behind it.
- Enforce the rules that matter. Permissions and automated checks hold them, so nothing depends on an agent following instructions.
- Keep a record. Every project has a log of verified facts and dead ends, so no session repeats an old mistake.
What each project shows
- Livi Home
- Value: a product in daily use that costs $0 a month to run and keeps family data away from outside AI.
- Skills: product ownership, delegating to agents under written rules, release and rollback, building to a cost limit.
- FormulaAnalytics
- Value: a forecast I can audit, because every prediction is scored in public, including the misses.
- Skills: probabilistic modeling, backtesting, automated pipelines, directing coding agents through pull requests.
- Trade Finder
- Value: one answer in place of five tabs, with the age and coverage of every source on show.
- Skills: data integration, data quality rules, managing a paid API budget, designing for the person deciding.
- Vault
- Value: ideas get captured in seconds and are findable later, with no filing work from me.
- Skills: running agents unattended, permission design, audit logging.
- Usage band
- Value: no more hitting a usage limit in the middle of a task.
- Skills: AI cost monitoring, plugin development, TypeScript with tests.