As a Coinbase CX/AI Content Strategist, I use data-driven insights to pinpoint friction points and improve CX. I work closely with machine learning partners, process specialists, legal teams, and product managers to design agent workflows that are AI/LLM compatible and write Help Center articles.
AI Conversation Design: I design end-to-end conversational and operational workflows that bridge the gap between human needs and machine logic. With a background in high-stakes FinTech, I specialize in building modular, LLM-readable knowledge systems that drive self-service and reduce agent friction. From designing branch-logic diagnosis paths to training AI for empathy, I transform complex data into seamless customer success stories.
Support Automation: By collaborating with machine learning partners and agent process specialists, I've designed and implemented AI-powered solutions that automate customer support, boosting agent efficiency and empowering more customers to find their own solutions through self-service.
Content Governance: Customer satisfaction, safety, and ease drive my strategic content enhancements. After discovering a legal review gap creating risk, I developed and launched a consumer content legal review process to ensure consistent content governance.
Prompt Engineering: I led a project to train the customer support AI to be more empathetic and accurate. I'm worked with machine learning, CX operations, and legal teams to optimize agent scripts by writing the prompts that guide AI and creating guardrails for sensitive topics. This ensures the AI's iterative responses in our agent platforms are not only helpful but also clear, low-risk, and consistent.
Cross-functional Trust: Fostering partnerships across departments and gaining stakeholder buy-in is one of my greatest strengths. When I engage partners, I share well-researched, organized, and value-driven plans. When I communicate with partners, I prioritize clear, kind, and energetic communication.
Help Center Content: I'm an expert in creating and optimizing content for Help Centers and internal knowledge bases. I use data-driven strategies to write and edit articles, ensuring the content adheres to style guides and provides the best possible user experience. My work helps customers find the answers they need quickly and easily.
Case Studies
Study 1: The "Machine-Readable" Blueprint
Conversational Design & LLM Synthesis
The Challenge: The "Clarity Gap" in AI Retrieval. Legacy knowledge articles often lack the granular structure and decision-tree logic required for modern automation. This structural deficit leads to two primary friction points: LLM hallucinations during retrieval-augmented generation (RAG) and increased cognitive load for human agents navigating ambiguous processes.
The Solution: Machine-Readable Content Governance. Engineered a new content framework designed for dual-consumption: structured enough for an AI to parse without "hallucinating" and intuitive enough for human agents to execute high-stakes tasks. By automating low-risk procedural steps, we enabled customer self-service and freed agents to focus on complex, high-emotion resolutions.
The Results: Increased CSAT (Customer Satisfaction) score 5% after launching new LLM-readable knowledge articles.
Strategic Action Plan:
Framework Development: Authored comprehensive LLM knowledge workflow formatting guidelines to standardize data ingestion, based on Voice of the Customer (VoC) data.
CMS Architecture: Reconfigured Contentful (CMS) components to support embedded metadata and Informational Data Points (IDPs), ensuring content was modular and machine-readable.
Logic Mapping: Partnered with Process Specialists to translate legacy text into granular workflows with exhaustive logic branches, accounting for all edge-case scenarios.
Automation Identification: Collaborated with Consumer Content teams to isolate automatable actions, such as account reviews and automated (low risk) flag removals.
Integrated Communications: Built "Communication Option" (Comm Option) components within Contentful to provide agents with contextually relevant, pre-approved scripts.
Cross-Functional Alignment: * Machine Learning: Verified that the AI Co-pilot accurately iterated through workflow logic within the agent dashboard.
CX Ops: Ensured the customer-facing chatbot maintained high precision when retrieving data from knowledge workflows.
Legal & Risk: Established robust AI guardrails to ensure compliance and mitigate enterprise risk.
Mass Transformation: Migrated 350+ agent knowledge articles into the optimized machine-readable format.
AI-Driven Quality Assurance: Developed a custom Glean AI Agent to automate the quality-checking and formatting of knowledge workflows, ensuring 100% adherence to the new standards.
Note: These are visual mock-ups only. Original Coinbase Knowledge articles are proprietary and cannot be shared.
Before
Old Knowledge Workflow
After
New LLM Readable Knowledge Workflow Metadata
Before I reformatted:
Lacks complete steps agent needs to take within tool
Does not include next steps for all scenarios
Not LLM readable
Does not include agent scripts for
After I optimized for LLM:
Includes metadata on backend that accounts for inclusions and restrictions
Automates process so AI can resolve (low risk) issues, including (low risk) flags
After
New LLM Readable Knowledge Workflow Steps
After I reformatted for LLM:
Informational Data Points [IDPs] (represented in brackets) include granular, embedded steps for AI and human agent use
Logic branches account for every scenario so there’s no guessing or hallucination
Embedded Comm Options contain agent scripts for email, chatbot, and phone agent use
Workflow communications have been legally reviewed to mitigate risk and improve accuracy
Workflows contain “Jump to” links that route to other scenarios
