The work
Fifteen years of making the operating model change, not just the tooling.
A decade around AI-era service platforms — Einstein and Service Cloud groundwork, then chatbot foundations, then enterprise GenAI and RAG. Operating AI directly since 2023, agentic products since 2026.
Timeline
Roles and scope.
Remote, Seattle based. Full employment history, with company names, on LinkedIn.
Scope and accountability
What I own, stated plainly.
Decisions
Owns go, defer, and stop decisions where the workflow keeps a human in the loop and the blast radius is contained.
Escalates through governance review when autonomy, data exposure, or customer impact crosses the line. The gate is a judgment about risk, not a spend limit.
Financial accountability
Owns the business case. Builds and defends the investment case, tracks realized value, and reports on it. Spend sits with function leaders.
People
Direct line into Product and Engineering, with a team and extended delivery capacity. Previously held the line for Customer Support, Customer Success and Customer Advocacy.
Outside that line: coaches eight professional services teams on solution design, workflow fit, feasibility and tool selection, and drove agentic QA automation without a QA organization to own it.
Previously hired, built, and led a 16-person organization.
Case studies
Four operating systems and decisions I have led.
Building the intake and prioritization machinery for an enterprise AI CoE
Global enterprise software portfolio. Founding member of the group. Built the intake and prioritization machinery first: discovery workbooks, maturity assessments, prioritization models, governance guidance, ROI dashboards, implementation playbooks, reusable templates.
Helped move leaders from experimentation to structured execution through phased roadmaps, prioritized use cases, named executive sponsorship, and quarterly accountability. Enablement reached 40+ business units via executive workshops, role-based learning, AI literacy programs, and champion networks.
Transforming customer support with AI in the workflow
The failure mode is starting with an enterprise programme. Established the baseline first, then picked the single workflow where the pain was already measurable — knowledge research and customer email drafting — and tested whether AI improved it before touching anything else. Measured against the baseline, not against expectations: knowledge research time fell 25%, answers got more consistent, workload came down.
The retrieval layer underneath it is a unified knowledge foundation spanning the ticketing, issue-tracking and knowledge systems the organization already ran on — one place to answer from instead of four to search. Troubleshooting time fell 40%.
From there, agentic workflows took load off the queue rather than just speeding up typing. The part that compounds: every closed case now produces a knowledge article, which feeds that retrieval layer, which makes the next case faster to resolve. The system improves by being used instead of decaying between refreshes.
Across the transformation, four independent measures moved: operating expense down 25%, throughput up 50%, backlog down 60%, and NPS from 65 to 100. Four different readings of the same change is the part that matters — a single number can be an artifact, four moving together is an operating model that shifted.
That closed-case-to-knowledge workflow became the standard playbook for turning support issues into system knowledge.
The organizational half mattered as much as the technical one. I created the company's first AI-at-the-core customer-facing team — process and workflow designed around AI from the start rather than retrofitted onto it — then scaled that workflow across four other support teams. That is the difference between a support team with an AI tool bolted on and a support operation whose shape has actually changed.
Onboarding 80+ engineers to AI tools in one week
Amazon Q Developer, now Kiro, and Microsoft Copilot, across engineering and QA. Run with the engineering managers and team leaders who owned those teams, on groundwork already laid: those managers had been through AI workshops first, which is why one week was even possible.
Four things did the work: real engineering problems that already mattered, low-risk starting points where mistakes were easy to catch, respected engineers sharing wins with their own peers, and office hours to remove friction as it appeared.
The goal was never blind trust. It was safe, informed challenge. Adoption is not a communications problem. It is a product experience problem — and not a one-person problem either. Managers who already believed it carried it further than any programme could push it.
On the harder question underneath it: rather than promising the team that AI would not change their jobs, which would not have been honest, we hired a college intern supported by AI tools and showed a different entry point into the role. Faster onboarding, meaningful entry-level work, a new talent pathway. AI lowered the experience barrier without lowering the quality standard.
Designing the human-authority model for a risk-scoring product
Traditional reporting shows the data. The product connects fragmented operational sources into one decision context, detects meaningful signals, and scores risk so teams know where to focus first. A product and engineering team built and shipped it.
My contribution was the shape of the decision: domain professionals set the scoring weights, validate the score, and choose the action. AI prioritizes attention. People decide what happens next. That constraint is what separates a risk score people act on from one they learn to ignore.
Foundational leadership, 2017–2023
Before AI was the headline.
Two industries, two rebuilds, and the play that repeats: stand up the function that does not exist yet, then make the platform underneath it carry the load.
The Technical Account Manager playbook, built twice
Built the Technical Account Manager programme from the ground up at a cybersecurity SaaS company, where the role did not exist before. Stood the same role up a second time at a regulated healthcare platform, in a different industry and a different compliance regime, to bridge the technical org and the customer.
Same play, different context. That is what makes it a playbook rather than a one-off.
Support organizations rebuilt across three regions
Led Support and Technical Account Manager teams across North America, India and Japan for an enterprise platform covering network security, risk analysis, and compliance. Took the organization to a 24x5 follow-the-sun model through workflow redesign, automation, and portal modernization, with the team producing 200+ knowledge base articles along the way.
At the healthcare platform: hired and built a 16-person organization, rebuilt support into a two-tier structure with clear escalation ownership, and led the build of a self-service knowledge and triage platform delivered in six months by the team.
The AI groundwork, from 2017
Migrated the support ticketing CRM onto Salesforce Service Cloud, then laid the groundwork for Salesforce Einstein chatbot services on top of it. Did the same at the healthcare platform, laying the groundwork for its AI chatbot and moving manual work into automation.
Groundwork, stated as groundwork: the platform decision, the data structure, and the sponsorship a later chatbot needs to work at all. Hands-on AI delivery starts in 2023.
Speaking
Where this gets said out loud.
Translating GenAI, agentic automation, governance, and change management into operating models that executives, product leaders, and delivery teams actually use. Keynoted Volaris Quadrants twice, in 2024 and 2026 — the employee leadership conference Volaris Group bills as the largest gathering of vertical market software leaders in the world.
Keynote — Volaris Quadrants 2026, Copenhagen
Operationalizing agentic AI platforms: moving from experimentation to execution. Agent loops, planning, tool use and retrieval. Multi-agent workflows coordinating specialized roles. MCP connectivity across enterprise systems, with human review and governance checkpoints in the path. Evaluation, monitoring and production readiness — the parts that decide whether a demo becomes a dependency.
The through-line: turning service platforms from systems of record into systems of action. Worked examples across intelligent triage, case summaries, root-cause analysis, bug creation and knowledge generation, spanning the ticketing, issue-tracking and knowledge systems a support organization already runs on. Then scaling those patterns as reusable assets across business units, and measuring adoption, throughput, quality, customer impact and ROI.
Keynote — Volaris Quadrants 2024, Cancún
Real-World AI Adoption: Transforming Customer Support with Augmented Intelligence. Augmented intelligence rather than replacement: AI supporting human judgment across transcription, case summaries, and email drafting. Covered early experimentation, measured business impact, why grounding and retrieval matter when hallucination is the failure mode, and the SPARK adoption framework — Start small, Pilot, Analyze, Refine, Keep growing.
UC Berkeley Executive Education — guest lecture and fireside chat
From AI Experiment to Operating System. The catalyst for changing the first question from what can AI do to what business problem is worth solving. Then the pilot, frontline adoption, the shift from chat to systems of action, spec-driven engineering, and the four-gate executive playbook.
S.E.E. Annual Conference & Trade Show 2026, Grapevine, Texas
2026 Trends and Best Practices in Fleet Software. Co-presented on the Transportation track: how AI and agentic AI are reshaping utility fleet operations, moving past predictive maintenance to the unglamorous part — getting data and workflows ready before anyone presses automate.
The Southeastern Electric Exchange's annual conference and trade show, which the organizer bills as the premier electric utility industry event.
EDGE 2025 — Volaris Smith portfolio, Oceanside, California
Smarter, Faster, Better: Using AI to Redefine Customer Support. Customer support at a turning point, argued as augmentation rather than replacement: AI should carry the repetitive work so the team stays on judgment, empathy, and the complex cases only a person can hold. Backlogs, response times, and operational friction come down as a consequence of that trade, not instead of it.
Led a session at the portfolio's annual talent development event, which brought 200+ leaders from 30+ vertical market software companies together under a growth, profitability and AI theme.
Volaris Virtual Innovation Summit, 2025
From Idea to Impact: Building Practical Innovation Pipelines That Deliver Measurable Value. Addressed business unit leaders from across the Volaris portfolio on how an idea becomes a funded, owned, measured thing rather than a demo that impresses a room and then stops.
The spine of the series: guardrails agreed with Legal, IT and Compliance before the build rather than after it; every initiative tied to a measurable business outcome; an executive sponsor and a named business owner on the hook; AI literacy and champion networks carrying adoption into the business units; adoption, cost savings, throughput, CSAT and time to value all tracked; and the pilots that work turned into reusable playbooks, workflows and operating models instead of one-off wins.
AI Champion Enablement Programs, 2024–2025
Executive workshops and maturity model sessions building AI literacy, champion networks, change readiness, and sustained adoption capability.
The champions shipped, under a problem-first rule: start from a real pain point or do not start. A retrieval chatbot over internal documentation, multi-agent infrastructure automation, an AI-powered hosted operations suite, a release-note generator, and DevOps automation over MCP servers — built by engineers and by colleagues who do not write code, and demonstrated at a GenAI at Work: Real Projects, Real Impact showcase.
Sustained through an internal enablement community, where the Microsoft 365 Copilot and Copilot Cowork material passed 8,000 views.
What I speak about
- Moving AI from experiment to operating system
- Operationalizing agentic AI platforms
- Frontline adoption and building trust
- Agentic workflows and human oversight
- The executive four-gate playbook
- Spec-driven engineering when code gets cheap
- Knowing when not to use AI
Executive education
Where the operating discipline came from.
- University of Chicago Booth — Chief AI Officer ProgramExpected Nov 2026
- Stanford Graduate School of Business — Digital Transformation and AI PlaybookJan 2026
- UC Berkeley Haas — Technology Leadership ProgramDec 2025
- UC Berkeley Haas — Data StrategySep 2025
- UC Berkeley Haas — Digital TransformationDec 2024
- UC Berkeley Haas — AI Business Strategies and ApplicationsJun 2024
Returned to UC Berkeley Executive Education as a guest speaker and fireside chat participant: From AI Experiment to Operating System.
Get in touch
Conversations I am always open to.
- Conference talks, panels, and fireside chats
- Executive education and workshop sessions
- Advisory conversations on AI adoption, governance, and operating models