Product work across regulated growth, platform modernization, enterprise automation, and AI workflow design.

Challenge

Costco Travel Insurance was a major sales channel, but a large share of transactions still depended on an in-house call center. That model was expensive, limited to business hours, created wait-time friction, and made it hard to scale demand. Costco and Travelex needed a self-service digital experience that could preserve Costco's brand trust, privacy expectations, and compliance requirements while moving customers through quote, enrollment, and purchase without agent support.

Product approach

I started by mapping the existing call-center workflow, reviewing call drivers and customer friction, and interviewing Costco account managers and operations teams. From there, I translated the service-heavy process into a digital quote-to-purchase journey, defining the PRD, success metrics, sprint priorities, and UAT plan across Travelex, Costco, design, engineering, QA, compliance, and analytics.

Execution complexity

The product had to simplify a regulated insurance workflow without hiding what was legally or operationally required. The team had to align Costco's partner expectations, data privacy, analytics, accessibility, quote logic, enrollment fields, payment behavior, and production-readiness testing while shipping fast enough to reduce call-center dependency.

Outcome

We launched the self-service white-label site in about six months, enabling Costco members to purchase travel insurance online 24/7. The site grew to $23M+ in annual sales, drove 150% growth, reduced call-center volume nearly in half, and became a model for future partner integrations.

Challenge

Costco required member verification before users could access the insurance experience. The launch requirement used Costco member ID and last name, but many users did not have their physical or digital card ready. That made verification a major early-funnel blocker before users could even see a quote or understand value.

Product approach

I treated verification as both a compliance requirement and a conversion problem. I evaluated smoother alternatives such as single sign-on, but that required Costco-side engineering and would have delayed the six-month launch. The MVP decision was to launch with the available verification API, instrument the funnel, and use real drop-off data to make the case for a better phase-two experience.

Outcome

The initial launch moved the business online and protected partner requirements, but the data made the conversion cost visible. We recovered about 10% conversion through iteration and created evidence for a stronger long-term verification strategy, including SSO or a redesigned verification step after users are more committed to the quote journey.

Challenge

The broader platform challenge was migrating direct and partner travel insurance sites from Drupal to Adobe. The team could not simply copy the old pages. We needed to reimagine content, quote flow, product display, validation, compliance, analytics, and partner-specific configuration while keeping multiple brands and rules consistent.

Product approach

I helped turn a complex migration into a product modernization effort. A key decision was to avoid hard-coding business rules into the front end. Instead, product rules, eligibility, pricing behavior, field-level validation, and display logic needed to come from APIs and configuration so the backend could act as the single source of truth.

Execution complexity

Each site variation could affect eligibility, content display, quote behavior, validation, checkout fields, or downstream fulfillment. QA had to validate not just the page experience, but also the API responses, product configuration, business rules, compliance text, and partner-specific edge cases.

Outcome

The migration created a more scalable foundation for direct and partner sites by separating experience presentation from product-rule ownership. It improved the team's ability to launch and maintain partner experiences with clearer configuration, more consistent validation, and less front-end rule duplication.

Challenge

Mobile traffic was high, but checkout completion lagged desktop. Analytics and customer feedback pointed to manual credit-card entry as a major source of friction, especially for returning users buying travel insurance on a phone.

Product approach

I led the cross-functional work with UX, IT, compliance, and payment-gateway teams to introduce Apple Pay as a one-tap payment option. The product work focused on PCI constraints, gateway behavior, Figma review, adoption metrics, load-time risk, QA, and rollout coordination across U.S., India, and China teams.

Challenge

The Canada D2C site needed a compliant cookie consent experience, updated privacy approach, and clear user controls. Legal required strict first-visit consent behavior, while UX, sales, and marketing needed to preserve trust, conversion, and analytics visibility.

Product approach

I initiated the project, researched market and compliance expectations, facilitated stakeholder workshops, and scoped a OneTrust banner integrated through GTM. The banner used clear consent categories, a preference center, and a full-screen pre-consent overlay required by compliance, without relying on dark patterns.

Challenge

Operational cost reduction was a team OKR, but insurance documents could not be optimized only for print cost. Some pages needed to remain single-sided for readability, legal, marketing, or customer-comprehension reasons, and policyholders could miss critical information if the wrong documents changed format.

Product approach

I analyzed the end-to-end document flow from content assembly and data integration through print and mailing. Using document volume and usage data, I grouped documents by reach and customer impact, then worked with business units, compliance, marketing, engineering, and product teams to decide which documents were safe candidates for double-sided printing.

Challenge

In the early REaCH build, stakeholders needed visible progress and fast demos, while the product also needed a durable dashboard foundation for future sensor data, alert patterns, and incident-command workflows. The technical discussion centered on whether to build quickly first or invest earlier in reusable structure.

Product approach

I first sought to understand the tech lead's concern: speed mattered because early demos kept stakeholders engaged. Then I built credibility by learning the front-end tradeoffs and proposed a practical middle path: preserve fast iteration while using reusable global styles and material-design components where they would accelerate later sprints.

Challenge

Underwriters often had to interpret long broker emails, chat history, and notes before deciding what quote change or policy action was needed. The work was slow, text-heavy, and dependent on human review, but it still required transparency and control because insurance decisions carry compliance and fairness risk.

Product approach

During Zurich's internal AI hackathon, I helped design a prototype workflow using LLMs and rules to ingest broker conversations, summarize intent, extract key business details, and recommend underwriting quote updates. I treated the AI as a co-pilot, not an autonomous decision-maker. The product flow kept output editable, source-aware, and reviewable so the underwriter stayed in control.

Product learning

The most important product decision was not just whether the model could summarize text. It was where to put user control, how to expose confidence and source context, and how to measure whether the assistant actually reduced underwriting effort without increasing decision risk.

Challenge

Local-service businesses lose urgent leads when owners are in the field, calls arrive after hours, or customers submit vague requests and move to another provider. The product question was how to capture enough context quickly, classify urgency, and hand the owner a clear next action without forcing customers through a long form.

Product approach

I framed one reusable workflow pattern: capture the request, classify the service type and urgency, collect location and timing, then produce an owner-ready handoff. The prototype applies that pattern across plumbing, HVAC, roofing, lawn care, and cleaning, with each route using service-specific copy, urgency signals, and handoff details while sharing the same interaction model.

Execution complexity

The important design constraint was making the prototype feel specific without rebuilding the product five times. Each workflow needed different language, urgency logic, and service details, while the underlying system needed to stay simple enough to iterate quickly with AI-assisted webcoding.

Outcome

The result is a live Cloudflare prototype that demonstrates fast product framing, reusable workflow design, AI-assisted build execution, and a practical local-business automation use case. It is intentionally positioned as a prototype, not a production revenue claim.

Problem

Many small service businesses know they need a website, but the existing options feel risky: agencies are expensive, DIY builders take time, and many finished websites still do not clearly capture calls, quote requests, or follow-up context. The product question was how to create a lower-risk offer that helps a business see value before committing money.

Personal insight

This came from observing a repeated gap in local-service businesses: owners are good at the service work, but not always at digital lead capture, follow-up process, measurement, or website operations. I wanted to test whether a productized website service could combine practical PM discovery, webcoding speed, and simple automation into an affordable offer.

Solution

LeadFlow Studio packages the work as a preview-first service: build a clean website direction first, let the owner inspect the value, then launch only if the offer makes sense. The MVP offer includes a practical website, lead form, alert workflow, simple reporting, and an optional Notion-backed lead database or lightweight CRM.

Challenge

The hard part is not only building websites quickly. It is turning a broad service into a repeatable product without losing customization. The offer needs standard packages, clear scope, simple pricing, fast asset collection, low maintenance burden, and enough reporting to prove value without becoming a heavy agency operation.

GTM plan

The 0-to-1 motion is narrow: target local-service businesses where lead capture is obvious and the purchase decision is simple, such as contractors, cleaners, lawn care, HVAC, roofing, and small professional services. The first wedge is a free website review and preview-first build. The 1-to-scale plan is to turn repeated needs into templates, route-specific intake forms, Notion database schemas, monthly report templates, and vertical packages.

From 0 to 1, then 1 to scale

From 0 to 1, the goal is validation: confirm pain points, close the first pilots, measure whether owners understand the preview-first model, and learn which service categories repeat. From 1 to scale, the goal is productization: reusable site templates, repeatable intake forms, Notion-backed backend workflows, automated monthly reporting, standard onboarding, and a clearer GTM channel for local-service niches.

Story

Before moving fully into product management, I led operational and digital adoption work at Panda Restaurant Group. I managed a multi-unit business, launched UberEATS adoption, improved KPI ranking from #187 to #7 out of 245 areas, and learned how product decisions show up in frontline execution.