Owned Analytics, CSAT, and Data Platform. Shipped 30+ features, drove $2.5M+ in annual savings, and shaped company strategy directly with the cofounders — from redefining Ada's north star metrics to co-authoring the MCP platform vision.
Championed the evolution of Ada's north star from Automated Resolution alone to Five Star Resolutions — a composite of resolution quality and customer satisfaction. Drove the strategic argument that resolution without delight is insufficient, backed by data showing 27% of "resolved" conversations still left users dissatisfied. This became the company's primary success metric.
Led the rollout of ClickHouse-powered real-time analytics, eliminating multi-hour data delays and delivering fresh metrics on every page load for all generative customers. A year-long, cross-functional initiative spanning 12+ contributors.
Redesigned the homepage dashboard from scratch with new views: cost savings, CSAT score distribution, survey completion rate, and popular topics — giving customers a single-pane-of-glass view of AI Agent performance. Iterated directly with the CEO and chief architect on design and information hierarchy.
Led the MCP Server from EAP through GA — coordinating across engineering, design, legal, enablement, and GTM. Then co-authored the v2 platform vision with the chief architect: closing the full feedback loop where AI analyzes agent performance, diagnoses issues, implements fixes, and measures results. Drove the public skills repo and marketplace launch strategy.
Successfully championed the inclusion of CSAT in Ada's north star metrics. Shipped NPS & CES survey questions, enabled CSAT by default for new bots, and built the CSAT block for generative — driving a 30–50% increase in survey response rates.
Migrated both Conversations and Messages endpoints to ClickHouse, eliminating 24–48 hour delays. Increased rate limits 3x, added autocapture variables, and shipped meta-messages for AI agent reasoning visibility.
Advocated for Ada to invest in its own AI agent the same way it asks customers to — with a dedicated AI Manager. Took on the role myself, identifying a significant performance gap and pushing for resources. Increased both AR and CSAT by 20% through hands-on coaching and optimization, proving Ada's own product could deliver the results it promises.
Discovered that customer-facing teams were spending 1+ hour per customer manually reformatting Topics data into slide decks for executive briefings — because enterprise customers viewed the in-product Topics reporting as unusable. Drove the investigation, aligned leadership on the severity, and led the redesign effort that became a top-priority initiative.
Stood up DAU/WAU and DAU/MAU dashboards in PostHog, benchmarked against industry-standard SaaS engagement targets. Built feature-level usage tracking for reports, dashboards, and new analytics features — giving leadership visibility into adoption for the first time.
Led the Recommendations feature through EAP with 7 customers. When analysis showed 70–80% of suggestions overlapped with existing knowledge bases, recommended pausing GA despite strong customer excitement — and redirected investment into ML quality improvements. Prioritized long-term customer trust over short-term launch optics.
Conducted onsite and virtual discovery sessions with 8+ enterprise customers — LeoVegas, Super.com, Empower, Miles Mobility, GoFundMe, Klaviyo, Clearcover, Dashlane — going beyond surface-level feature requests to understand root challenges. These insights directly shaped the roadmap.
Added Topics, Playbooks, Coaching, Language, Channel, and more filters to the MCP Server for deeper conversational analysis.
MCPAnalyticsCompleted the real-time migration — both Conversations and Messages endpoints now powered by ClickHouse with sub-second freshness.
APIClickHouseNew report section surfacing search queries with zero matching knowledge articles. Requested by 10 customers representing $5.9M ARR.
AnalyticsKnowledgeShipped Ada's MCP Server enabling natural language analytics via Claude Desktop and ChatGPT. OAuth and API key auth supported.
MCPAIPlatformFull search and filtering for Playbooks management, supporting customers with 100+ playbooks.
UXPlaybooksRenamed Actions Usage to API Usage, added HTTP Request block tracking across Greetings, Processes, and Handoffs.
AnalyticsAPIAligned reporting language from "bot" to "AI agent", "chatters" to "customers". Improved attribution tooltips across CSAT reporting.
CSATUXClickHouse-powered real-time analytics rolled out to all generative customers. Year-long cross-functional initiative eliminating multi-hour data delays.
ClickHouseAnalyticsPlatformLed rollout of fine-tuned 4.1 mini for Automated Resolution assessment. More accurate than 4o (o3-as-judge favored it 65% of the time), saving ~$1.4M annually.
LLMCost SavingsUpgraded the LLM for Inquiry Summary and Reason for Classification, replacing Google Translate with LLM translation. ~$765K annual savings.
LLMCost SavingsRedesigned the CX-Product feedback loop. Freed up meeting time for deep-dive discussions on the most impactful customer pain points.
ProcessCXRedesigned the homepage with cost savings view, CSAT distribution, survey completion rate, popular topics, and modern visualizations.
DashboardAnalyticsAdded Net Promoter Score and Customer Effort Score as optional survey questions — critical quality metrics used as north stars by many customers.
CSATSurveyLaunched toggle for AI-generated Topics and optimized the matching prompt, reducing failed matches by 50%.
TopicsAIExpanded AR accuracy assessment to include Processes, Actions, Coaching, and Company Description — previously only Knowledge was used.
ARAccuracyEnabled conversation summaries for non-English conversations, helping agents get up to speed across languages for clients like Grab, Loblaws, and Sophos.
i18nHandoffsUpdated labels and descriptions across the Conversations Breakdown report to reflect AI-first language and multi-channel support (chat, voice, email).
AnalyticsUXMerged over 100 pull requests using Devin, Claude Code, and Cursor — 3x more than any other person at Ada. Became the #1 contributor on the app repo in the last month and #2 on product-docs over 6 months. Inspired peers to adopt AI-powered development — multiple colleagues cited this as a direct inspiration in performance reviews.
Used Replit and AI tools to build interactive design prototypes on the fly during feedback cycles with leadership — turning abstract ideas into clickable mockups in minutes instead of days. The CEO asked me to share the approach company-wide, calling it a game-changer for how Ada builds and provides feedback on product.
Built a GitHub Actions pipeline that automatically picks up bug tickets from Linear, has Claude investigate root cause across Ada's repos, and opens a draft PR with a fix — all before a human even looks at the ticket.
Created an automated alerting system that monitors customer bot metrics (Containment, Volume, AR, CSAT) and sends Slack alerts on anomalies, with optional Claude AI analysis of flagged conversations.
Started a cross-functional learning channel and curated monthly highlights of the best 5-star CSAT comments — which became a key marketing asset for the company.
Built a working prototype for gradual rollout / A/B testing of Playbooks, enabling customers to test changes on a percentage of real traffic before full deployment.
Prototyped a natural-language report builder using Claude to translate plain English into structured analytics queries — with multi-turn conversation and entity-aware context.
Your closeness to customers, taking frequent calls, working with CSCs, and reviewing Genni conversations provides invaluable insight into real problems. You move quickly to respond and jump in wherever needed, which accelerates progress.
Adam has done an excellent job building intuition for Ada's customers' problems. He listens to context from a wide range of people and sources and distills it into meaningful initiatives that will make Ada a fundamentally better product.
One of Adam's best qualities is that he will always go "see for himself" before making any judgements. His decisions are always rooted in data. I see Adam growing as a great product manager.
Seeing how effectively he gets around his team's time limitations by prototyping first with AI has been really cool, and inspired me to do more vibe-coding too. Adam's problem-first framing, openness to feedback, and creative use of AI make him a strong collaborator.
I was very impressed with how you sourced feedback from the CX team on how to improve the in-app analytics experience, and how you shaped this into priorities and a roadmap. Love having you on the team!
Adam is calm, friendly, and easy to work with. He brings a sense of control even when things are unclear or stressful. After meetings with him, I usually feel more focused and confident in what needs to happen next.
Great bias toward shipping product. Always pushing to iterate quickly and ship fast. Good at finding opportunities to leverage AI to move faster. He was always calm and never made things worse with panic. He always gave me the time I needed to work through the issue.
Read any of Adam's documentation, and you will see evidence of a thoughtful, well researched, well organized, innovative and iterative process. He reduces timelines and friction in many ways, including but not limited to self-starting AI experiments and spinning up proof-of-concepts through clever and iterative prompt engineering.