The part of AI work that begins after the demo: helping people understand what a system can do, build against real work, and keep using it when the novelty wears off.
SOURCE-BACKED RECORDS
Since January 2025, I have led or co-led more than 80 source-backed workshops, trainings, webinars, masterclasses, speaking sessions, and community-education events.
The rooms vary. Business owners, operators, real-estate and title professionals, professional-service teams, and people getting serious about AI for the first time. The job stays the same: make the work real enough to use.
Completed a live AI security/platform exercise in 18 minutes, first among approximately 75 developers and engineers in a session allotted up to five hours; selected as one of five participants to present findings as an executive-level technical briefing.
A recurring series for business owners and operators on AI Employees, practical follow-up, and systems that give teams time back.
A practical session on Claude and applied-AI workflow education for business operators and AI adopters.
A session on practical tool selection and the work of deciding where AI is actually useful.
A speaker session on applied AI and business AI for a mixed operator and professional audience.
A conference session on AI and technology for more efficient note and loan-management work.
The problem. A capable model is not automatically a reliable business role. The real work is defining what it owns, what it can access, what requires approval, and what evidence says it is ready.
The approach. I work with a Runtime Shell → Client Forge model: establish a reusable operating foundation, then shape the role, context, tools, and onboarding path around the actual work.
The rule. Structural checks, live authorization/read-back, and client acceptance are separate gates. A tool being available is not proof that a capability is ready.
The problem. A connector list can make a system look more capable than it is. A useful integration has to run under the correct identity, with the right account, clear permissions, and an understandable approval moment.
The approach. I design for scoped access, named account selection where ambiguity matters, safe read-back verification, and a deliberate distinction between learning about a workspace and taking action inside it.
The rule. Trust is designed into the handoff. It does not arrive after deployment.
THE UNDERLYING DEPLOYMENT WORK IS PRIVATE. THESE NOTES DESCRIBE ARCHITECTURE, GATES, AND DESIGN TRADE-OFFS—NOT CUSTOMER CONFIGURATION OR CONNECTED ACCOUNTS.
Seattle-based. Working across technical architecture, AI adoption, founder/operator enablement, and practical education.