Product Manager · Discovery-Led Product & AI

Roman Novak

I run structured discovery — JTBD, user research, deliberate scoping — to ship products people actually adopt, including AI.

Cvent · 2021–2026 5 yrs Product Discovery & Delivery 8+ yrs prior Support & Client Services

Over five years I've run discovery-to-delivery on eight products from a blank page — grounding every one in JTBD research and user studies, then shipping deliberately, including three built on AI.

8+
products shipped to GA / beta
15+
features delivered across 5 years
3
AI products shipped from scratch

Three products, three phases of scope

Case 01 — AI · 2026

Shipping a conversational AI analytics product

Problem

Planners wanted answers from their data, not more dashboards to learn — and there was no internal playbook yet for building or evaluating an LLM product.

Role

Product lead — defined the tool suite, wrote the eval framework, and drove the product from proof-of-concept to open beta.

Result

Live in open beta today, with a 21-tool natural-language agent suite spanning registration, session, attendance, feedback, and revenue data — now driving the eval score program for the beta-to-GA transition.

What I did
  • Defined the full 21-tool suite spanning five data domains
  • Wrote eval question sets and built the evaluation scoring program using third-party LLM observability tooling
  • Coordinated with the internal Data Science team on a companion forecasting capability, and authored the internal FAQ for Sales, CS, and Product Leadership
  • Shipped the chart-generation layer that lets the assistant answer in visuals, not just text
Case 02 — Scale · 2025

Taking analytics from one event to a full program

Problem

Power users managing 200+ events a year had outgrown per-event dashboards. No infrastructure existed to group or compare events at the program level.

Role

Lead PM across all three milestones — new data infrastructure, UX patterns, and coordination across Content, Engineering, and Customer Success.

Result

Delivered a portfolio-level analytics experience across three milestones in nine months (GA May 2025) — plus the team's first shipped AI feature, an automatic feedback-sentiment summary, in the same window.

What I did
  • M1: event group creation and management
  • M2: program performance landing page with smart group suggestions
  • M3: group-to-group and period-over-period comparison, plan vs. actual
  • Led field research at a major customer event that surfaced benchmarking and sharing needs now on the 2026 roadmap
Case 03 — Foundation · 2021–2023

Building the first in-product analytics surfaces from zero

Problem

Planners had no in-context view of registration pacing, session engagement, or feedback — everything required manual report runs, and no product existed to build on.

Role

Product Manager, owning the problem space end to end from greenfield discovery through GA and measurement.

Result

Shipped the first versions of what are now the session-performance and registration-performance dashboards (GA Aug 2022) — Cvent's primary per-event analytics surfaces — plus a PM/UX discovery process the team still uses.

What I did
  • Ran discovery and defined both dashboards from a blank problem space
  • Took the session- and registration-analytics MVPs to GA
  • Built the measurement framework that surfaced early under-discovery signals, which shaped later UX research
  • Established the PM/UX discovery process adopted as the team standard

Competencies, in evidence

DiscoveryProblem reviews, PRFAQs, JTBD-grounded research
OwnershipProblem definition through post-launch measurement
AIRequirements + eval frameworks for LLM features
Cross-teamEngineering, Content, CS, Data Science coordination
MeasurementCohort & funnel frameworks, behavioral-change metrics
StrategyMulti-year roadmap, per-event → AI platform

Everything shipped, by capability

Capability 01

Conversational AI analytics & custom dashboard builder

A natural-language AI assistant that lets event teams query registration, attendance, engagement, and feedback data through purpose-built agents — generating charts on demand and assembling fully personalized, multi-source dashboards, so the platform becomes the primary place to get insight rather than a data source you export from.

Delivery history (4 items)
YearMilestoneStatus
2026Custom, multi-source dashboard builderIn Progress
2026Companion forecasting capabilityIn Beta
2026On-demand chart & visual generationIn Beta
2026Natural-language query assistant (open beta)In Beta
Capability 02

Event program intelligence

A portfolio-level experience for organizing events into logical groups, comparing performance against a baseline or prior period, and surfacing KPIs — registration pace, attendance, revenue trends — across an entire event program, with export and external publishing for stakeholders who don't have platform access.

Delivery history (6 items)
YearMilestoneStatus
2025–26Group-to-group comparison & external publish/export (M3)Released
2025Group sharingReleased
2025Period-over-period comparisonReleased
2025Funnel improvementsReleased
2025Portfolio landing page with smart group suggestions (M2)Released
2025Event grouping (M1)Released
Capability 03

Registration performance analytics

A real-time, in-platform dashboard for monitoring registrations, revenue, and conversion at a glance — with flexible filtering by type, segment, and custom field, side-by-side comparison against past events, and shareable exec-ready reporting, all without leaving the platform.

Delivery history (5 items)
YearMilestoneStatus
2024Combined comparison + filteringReleased
2024Custom-field filteringReleased
2023Revenue tracking depthReleased
2023Event-to-event comparisonReleased
2022Real-time registration & revenue tracking dashboard (MVP)Released
Capability 04

Session performance analytics

A filterable, in-product dashboard giving planners a unified view of session enrollment, attendance, engagement, and feedback — with cross-event benchmarking and an AI summary that turns open-ended attendee responses into an executive-ready sentiment read, so nothing has to be compiled by hand across reports or spreadsheets.

Delivery history (4 items)
YearMilestoneStatus
2025Cross-event benchmarkingReleased
2025AI-generated sentiment summary from open-ended feedbackReleased
2024Additional filtering and cross-event compareReleased
2022Unified session enrollment, attendance & engagement dashboard (MVP)Released

Let's talk

Open to Product Manager roles — discovery-led product, 0→1 delivery, and AI.