Building an AI-First Product Marketing System Around the 3D Framework

This case study was created from an episode of the Product Marketing Adventure Podcast. It has been adapted into Dusted's case study format to preserve and showcase the guest's real-world experience. Written to preserve the voice of the true author, Steven Schuler.

I use a 3D product marketing framework: data, decision, and delivery. Data means collecting market and customer information; decision means turning it into messaging, positioning, and go-to-market strategy; and delivery means bringing those decisions to life across assets such as one-sheets, web pages, emails, and events.

As AI accelerated product and engineering teams toward weekly releases and six-month roadmaps, I saw an opportunity to modernize the way product marketing worked. Rather than using AI simply to generate more content, I wanted to build an always-on system that could collect and synthesize information, surface important trends, and continuously improve our marketing knowledge base.

Product and engineering teams were shipping faster, while product marketing was often still operating on a quarterly rhythm. At the same time, valuable customer intelligence was fragmented across Gong, Slack, Zendesk, Notion, Zoom recordings, and Google Drive.

The increasing power to produce content quickly also created a quality risk: teams could end up creating sloppy, generic AI-written content faster. The real goal was not to increase content volume. It was to reduce redundant work and shift more of my time toward the strategic decision-making at the center of the 3D framework.

I built the system in five steps:

1. Map the knowledge foundation. I catalogued where information about customers, competitors, products, and market trends lived across the organization.
2. Map stakeholders and their needs. I identified what product, sales, demand generation, brand, solutions engineering, and customer success needed to know, along with where each group worked and what outputs would be most useful to them.
3. Create the golden documents. I consolidated the core sources AI should rely on: persona and ICP documents, brand voice, vertical and industry insights, product documentation, messaging, PRDs, and examples of both high-performing and poor content.
4. Build focused agents and feedback loops. I started with small, useful applications rather than trying to create a perfect system all at once. I reviewed and pressure-tested AI-generated insights myself, validated them with peers, and codified the decisions back into the knowledge base.
5. Market the marketing. I shared early wins, demonstrated new capabilities, gathered feedback, and placed each agent where its users already worked. Small wins built trust and helped create momentum for broader adoption.

The system used a retrieval-augmented generation approach: AI retrieved information from the curated knowledge base before generating an output. This kept the system grounded in our actual messaging, products, customers, and market knowledge while ensuring that I retained ownership of the strategic decisions.

The knowledge foundation drew from sources including Gong sales-call transcripts, Slack, Zendesk Suite support tickets, Notion, Zoom Workplace recordings stored in Google Cloud Platform, product documentation, PRDs, persona documents, and messaging materials.

For an early real-time sales enablement agent, I consolidated years of competitor battle cards and sales training—including objection handling, trap-setting questions, competitor tactics, and technical positioning—into Gemini Google Notebook LM , Google’s free knowledge tool. Sales representatives could ask questions during live prospect calls and immediately retrieve relevant guidance.

I also used AI tools such as Claude to generate drafts from the golden documents, identify patterns across large volumes of customer conversations, and evaluate examples of good and bad content. The system was designed as a RAG-powered knowledge base rather than a collection of disconnected prompts.

The first sales enablement agent made previously buried information immediately usable. Instead of searching through decks for a specific battle-card slide or promising to follow up after a competitor came up, sales representatives could ask for the right response during the call. This helped move opportunities from stage one to stage two more quickly and accelerated deal velocity.

A launch agent could produce a first draft of approximately 12 launch assets—including three social posts, a blog post, a press release, and web-page copy—instead of requiring roughly two weeks of initial writing. That gave the team time to socialize the work with product and gather feedback about a month before launch, rather than scrambling the night before.

An analyst-relations agent combined solutions-engineering knowledge with product documentation to accelerate responses to analyst RFIs, including questionnaires with as many as 200 questions. Across these use cases, the main result was not simply faster content production. We reduced repetitive work, improved consistency and cross-functional alignment, and created more time for strategic decisions, stakeholder relationships, and market-informed product marketing.