“Deep technical expertise in AI/ML — and seamless translation into clear tradeoffs for business leadership.”
$5M+ manual operation → reusable AI accelerator.
I led SIGNAL end to end for a top-five U.S. wealth-management firm in the independent-advisor channel: found the case, prototyped the solution, scoped the work, led enterprise delivery, transformed the operating team, and expanded the account.
The winning model is peer proof plus tailored delivery.
Traditional services are under pressure when they sell generic transformation instead of what peers are actually doing. Static products are easier to replicate than ever. The durable model is: show a working solution, tailor it to the client, transfer ownership, then help with the next strategic problem.
The operating model was too manual to protect advisor trust at scale.
The organization wanted lower service cost, faster visibility into why advisors were calling, better agent-assist knowledge, and no degradation to CSAT. The current model could not see emerging problems quickly because it was built around manual reports and keyword-bound taxonomy.
We separated near-term proof from long-term value capture.
The long-term value case was cost-of-service reduction and trust preservation. The short-term proof was whether we could replace the manual contact-driver workflow with an AI-enabled operating capability the client could actually run.
I won the room by turning stakeholder language into a North Star.
Copilot captured the meetings, Excel organized the verbatims, and Claude helped compress the thread into a prototypeable product concept. The hardest stakeholder's own words became the solution spec, the demo, the change order, and the requirements foundation.
The prototype made the strategy feel real.
SIGNAL was a single-page app that ingested call transcripts, mapped calls to contact drivers, surfaced emerging issues every 15 minutes, detected knowledge gaps, and let leaders ask questions across structured and unstructured data.
Top contact drivers
Account transfer status18.2%
RMD withdrawal timingNEW
Statement / tax doc11.4%
Portal login / MFA6.1%
Knowledge gaps
312 calls · 0 articles
coverage stale
Every call became a signal; every gap became routable work.
The pattern was intentionally enterprise-ready: semantic search with confidence scoring, human review for low-confidence matches, knowledge-base lookup, gap flagging, and an analytics layer for proactive and reactive questions.
AI accelerated the artifact work; scope judgment protected delivery.
I used the prototype, meeting notes, MSA, SOW language, and prior change orders to draft scope quickly. Then I made the hard choices: non-prod first, production gated, explicit client dependencies, ARB and AI governance, and a lean delivery pod.
The team moved from manual reporting to AI experience operations.
This was not a software handoff. The contact-driver analytics team stopped manually tagging calls and shifted into a higher-value operating role: using SIGNAL and chatbot logs to identify signals, tune prompts, adjust guardrails, and find the next automation opportunities.
- slow spike diagnosis
- taxonomy blind spots
- limited business leverage
- prompt tuning
- guardrail refinement
- executive signal reports
Cadence was not status theater. It was altitude management.
Daily scrums kept the client and project team unblocked. Leadership check-ins escalated cross-functional dependencies. Monthly SteerCo connected delivery to call-abatement goals, AI-enabled transformation, and the next commercial opportunities.
Delivery scaled because ownership was designed into the team.
I sourced and interviewed the technical and functional leads, set weekly performance connects, and used one-week Kanban-style sprints so team members owned tasks, leads owned workstreams, and I owned the outcome.
The delivery win became a growth flywheel.
Because SIGNAL was prebuilt in our AWS environment, delivery was smoother and the asset was reusable. The client relationship opened adjacent personalization and AEP work, while the accelerator created credibility in airline and bank pursuits.
I use AI to accelerate the work — and upskill teams by doing.
I treat AI as a developer, product manager, project manager, and commercial acceleration partner — then teach teams by doing. At Deloitte, that became Claude training completed by 1,000+ practitioners. At PwC, it became sales and delivery standards for AI-age dealmaking, plus partner and Managing Director trainings anchored to real pursuits.
We stopped only telling clients about AI. We let them experience it.
I pioneered an RFP response pattern where the response includes an HTML experience embedded into the deck, partner letters beside it, and an inbound AI phone number with the full RFP context. It turns the proposal into a working demo of the capability we are selling.
Same delivery muscles. Different risk surface.
The same playbook applies to enterprise AI programs: find the business case, build a tangible prototype, govern the risk, drive adoption, and package the learning. Internal rewiring moves faster; customer-facing agents demand tighter evaluation, observability, rollback, and brand controls.
- faster iteration
- behavior change is the work
- lighter external blast radius
- evals and HITL
- observability and rollback
- trust and brand risk
Questions?
The point of the story: I can run enterprise delivery, build the technical system, transform the operating team, expand the account, and lead the accelerator-to-delivery model I have proven across PwC and Deloitte.