Bring the funnel into the product
See how one product context can power landing pages, email, analytics, user research, and voice-agent interviews without rebuilding every handoff.
Ben Sufiani, Lydia Penkert, Patrick · The article: Vibe Code Your Marketing Funnel
The funnel no longer has to live beside the product
Ben describes the old stack as a chain of landing-page builders, forms, email tools, and CRM records that did not know whether the person was already a customer. Every step created another handoff and another partial view.
Inside one product context, the same record can carry the source, the signup, the product activity, and the revenue. The page and the follow-up can respond to what the person has actually done.
“The worst thing was that there was the marketing world and the product world, completely separated.”
Context turns a one-off funnel into a repeatable skill
The product files hold the design, story, audience, and existing journey. Ben can create a lead magnet and its follow-up without restating all of that context in a fresh tool.
After a funnel works, he turns the process into a skill. The next landing page inherits the same rules and can still be corrected in conversation.
“Whenever I create such a motion, I ask the skill creator to create a skill around that.”
AI can widen access to research without inventing evidence
Lydia uses AI to run heuristic reviews, challenge assumptions, and suggest lean ways to test an idea. Those tools help a team begin research earlier and make a specialist's method available in the product workflow.
She keeps a firm line around synthetic users. A generated persona can pressure-test a thought, but it cannot reveal something a real user has not already put into the evidence.
“Synthetic users are not real ones. They are based on training data or whatever you give to them.”
Voice agents make interviews scalable – the analysis makes them useful
Patrick's team built voice agents that conduct customer interviews and a second layer that evaluates the conversation. It clusters themes, tracks sentiment, and exposes where the interviewing agent missed an opportunity.
A separate model judges call quality rather than letting the interviewer grade itself. The product is not simply a talking bot; it is the evidence pipeline around hundreds of conversations.
“The idea is not the voice agent itself, but the analytics layer we built on the conversation.”
Dive deeper
Read the complete stack for building landing pages, follow-up, and attribution inside your product.
Read the insightSee how builders are connecting product, growth, and AI in working systems.
Register freeUse the Builder Codex to move from a separate landing page to one product journey.
Open the Codex