Six case studies — the strategy on top, the AI underneath. Every number is audited and traces to the record.
Recruited to stand up the enterprise sales motion from scratch for an AI-powered workforce-analytics platform serving 9,500+ global brands — packaging, competitive positioning, and value-selling that moved real revenue.
ActivTrak had a self-service product with strong PLG numbers and almost no enterprise infrastructure. No packaging built for larger accounts, no value narrative that translated to sales conversations, no competitive positioning that held up in deals. The job was to build all of it.
The first move was packaging. The Essentials Plus tier was designed and launched specifically to give the field something to sell into mid-market and enterprise accounts — a defined offer with a clear value prop, mapped to the economic buyer's concerns rather than the product feature list. That tier hit $2.7M in new ARR against a $675K goal — roughly four times the target — in the first full cycle. Since launch, Essentials Plus has added roughly $12M in run-rate ARR to the enterprise segment; the packaging, positioning, and GTM are mine, and the tier's cumulative result is the motion's.
The second move was competitive. The sales team was discounting to close deals: average competitive discount ran at 18%. That number exists because the field lacked the positioning to hold price. The work here was repositioning against the two or three competitors that showed up most in deals, building battlecards reps would actually use, and training on the objection patterns that led to discounting. The discount rate dropped to 4%. Against identified competitors in the $30K+ enterprise segment, the win rate has held around 74% (no-decision deals excluded), sustained through continuous re-positioning off a monthly win-loss report I run myself. That discipline is what lets the number survive a shifting competitive field.
The third move was infrastructure. Customer Education and Technical Writing were brought under the same function as Product Marketing — creating a tight loop between positioning, enablement, and documentation. That structural consolidation is what allowed the AI layer (described in the next case study) to compound: one team, three functions, one coherent knowledge base.
The competitive infrastructure work also included a ~10–12% ARPA lift from packaging improvements, and expansion ARR of $947K against a $626K goal — the enterprise motion, once started, ran across both new and existing accounts.
Two more pieces of the motion. I introduced ROI and value-selling into the org and built a value calculator now deployed to the entire sales team. And I drove Schedule Adherence, a net-new product, from concept to launch, owning the research, definition, positioning, and GTM, and acting as PM across the development lifecycle alongside engineering. The positioning work bleeds into product when the product is technical enough to demand it.
A production agent network spanning competitive intelligence, enablement, education, and content — self-improving by design.
The team is three people. The output runs like a larger team because roughly ten production AI agents handle the infrastructure layer — the work that would otherwise be the ceiling on what a small function can do.
The agents are not demos. They are in daily use: the competitive scanner runs weekly against ~24 competitors and posts verified summaries without manual intervention; SARA (Sales AI Resource Assistant) handles objection handling, battlecard retrieval, and value prop lookup by ICP; the Account Intelligence agent turns CRM notes and call transcripts into MEDDIC analysis; the General Helper gives reps company-specific, grounded answers from a RAG knowledge base built on internal documentation.
The infrastructure also includes three custom MCP servers — built against the Obsidian, Productboard, and Smart Connections APIs — that wire the agent layer into the tools the team already uses. The Python API client for ActivTrak's own product (pagination, retry/backoff, type-safe models, UTC handling) was written to give the agents reliable data access.
The output numbers: delivery time roughly 50% faster, output roughly double — with no added headcount. The team also ships 50+ discrete AI deliverables, including apps (ROI calculator, GTM tracker, enterprise packaging launch dashboard) and the weekly automated release-status and competitive-intelligence documents that previously required manual coordination across functions.
The 2024 AI Marketing Innovation Award from the Product Marketing Alliance was a recognition of this body of work — not a product feature, but a production system built by a marketing team, in a real organization, doing real revenue work.
Replaced static battlecards reps ignored with an always-on system: ~24 competitors, adversarially verified, auto-posted.
The starting condition was a battlecard deck that was months out of date and largely unused. Competitive positioning was a manual research task — someone had to find the changes, summarize them, update the deck, and route the update to the field. That loop broke constantly: the research was slow, the updates were infrequent, and by the time a rep needed the information, it was stale.
The replacement is an automated intelligence pipeline. Every week, the scanner runs against approximately 24 competitors: it pulls pricing pages, G2 reviews, product update announcements, and positioning language, then runs adversarial verification passes to check for hallucination before generating a structured summary. The output is formatted and posted automatically — no human in the loop between the run and the post.
The system is built in Python using a multi-agent architecture: one agent per competitor domain, with a verification agent checking each claim against its source before the summary is finalized. The infrastructure runs unattended. When a competitor changes its pricing or drops a major product update, the scanner catches it in the next weekly cycle.
The downstream effect: reps stopped asking for updated battlecards because the information was already there. The competitive discount rate dropped from 18% to 4% — the scanner is one part of that story, not the whole story, but it is the part that made competitive positioning durable instead of episodic.
Try a live demo of this approach ↗
That opens VibeCI — my Kaggle "AI Agents Intensive" capstone (submission in progress): a standalone, clickable demo of the claim-vs-reality competitive-intelligence approach. It is a separate demo, not the internal production scanner described above.
Built the brand and the self-service conversion engine for a JetBrains video SaaS, from launch to ~$22M ARR.
Wave.video launched as a product without a market position. The job was to build everything: brand, messaging, acquisition channels, the conversion funnel, and the affiliate and community infrastructure that would make self-service growth compound over time.
The growth motion was product-led. The organic channel was built to 96% of traffic — 160K monthly visitors — and the conversion funnel was optimized end-to-end, from search and social discovery through to trial activation and upgrade. The affiliate program scaled to 1,200+ affiliates; the social audience to 200K+; the email list to 1M+. None of those numbers were bought — they were built through positioning and channel strategy.
The team was ten people across five countries. Coordinating that team across time zones and a fast-moving product roadmap — while maintaining consistent positioning across every channel — was as much of the job as the marketing strategy itself.
Over three years, ARR went from $0 to ~$22M. LTV increased 78%. The PLG motion ran without a traditional sales team: the product, the content, and the conversion infrastructure did the work.
Repositioned an early-stage content-curation tool into an enterprise customer-journey and personalization platform — and moved the company upmarket.
Roojoom had a capable product and the wrong market. It was sold as a content-curation tool to SMBs — a crowded, low-ARPU space with little room to grow. The opportunity was in the enterprise: large organizations needed a way to orchestrate and personalize customer journeys, and the underlying technology could do exactly that. The job was to reframe what the product was, who it was for, and what it was worth.
The pivot was a repositioning, not a rebuild. New positioning, new packaging, and a value narrative aimed at enterprise buyers turned a content tool into a customer-journey platform. The upmarket move changed the economics: instead of chasing many small SMB accounts, the motion targeted a smaller number of high-value enterprise contracts.
The repositioning moved deal economics from roughly $6K a year in the SMB segment to as much as $1M a year in the enterprise — landing a $1M/year contract with a large European enterprise, with the largest deals reaching $1.6M in ARR. Sales-qualified-lead conversion ran at 25%, and the traction helped secure a private-equity capital infusion.
"Solo Pilot → Squadron Leader" — 14 SCORM modules taking every employee from first prompt to building agents, deployed company-wide.
The enablement problem at ActivTrak was not that people were skeptical of AI. The problem was that "use AI" is not a skill — it is a directive. The gap between "we encourage everyone to use AI tools" and "people are actually building things with AI" is a curriculum gap, and closing it required a structured program, not a nudge.
"Solo Pilot → Squadron Leader" is a 14-module SCORM course built in-house and deployed through the existing learning management system. The arc runs from first prompt to multi-agent orchestration: basic prompting, then iterative refinement, then tool use, then building simple agents, then chaining agents together. The final modules cover the infrastructure concepts — RAG, MCP, evaluation — that employees need to understand in order to work alongside the production agent network.
The course was built by the same team that built the agents. That matters: the curriculum reflects what the infrastructure actually does, not a generic introduction to AI concepts. The examples are drawn from real workflows; the exercises use real tools.
The result: ~1,600 active users with a 61% completion rate. Company-wide badging was added to mark completion at each level. The program also covers accounts representing $13.9M + $2.6M ARR in terms of adoption — which is reported as coverage, not revenue attribution.
That's the work. If it maps to a role you're filling, let's talk.