By Teresa Torres · producttalk.org · @ttorres on X · LinkedIn
This episode of Just Now Possible features Mowie's CEO Chris O'Connor and co-founder Jessica Valenzuela discussing how their AI marketing platform for small and medium businesses grew out of a hands-on concierge service. The founders explain how Mowie builds a context 'dossier' about each business by crawling the web, infers customer segments and marketing pillars, and generates quarterly content calendars with channel-specific posts. They describe moving from rigid schemas to looser markdown-based intermediate documents, and using customer approvals, edits, and regeneration requests as their main evaluation signal. The conversation is useful for anyone building AI products that must synthesize messy real-world data while keeping humans appropriately in the loop.
01Key takeaways
- Do the manual work yourself before automating, so you understand what good output really looks like.
- Use a layered context document structure so generated content stays grounded in a consistent picture of the business.
- Prefer flexible intermediate formats like markdown over rigid schemas when processing messy, unstructured real-world data.
- Make recommendations traceable so users can see which inputs shaped them, which builds trust.
- Treat customer approvals, edits and regeneration requests as lightweight evaluation signals for AI output quality.
- Connect marketing outputs back to business outcomes like point-of-sale data to prove real attribution.
02Key sections
- From concierge service to AI product
- The founders describe how managing marketing manually for overwhelmed business owners revealed the problem, and how they validated demand before automating. Their early work was done by hand to learn what good output looked like.
- Document hierarchy and business context
- Mowie builds a dossier about each business by crawling public sources and layering context documents. This structure lets generation draw on a consistent understanding of the brand and its customers.
- Schemas versus loose markdown
- Early rigid schemas for structuring messy data gave way to looser markdown intermediate outputs. Flexibility made the system more robust to unstructured real-world inputs.
- Calendars, campaigns and traceability
- Quarterly calendars combine public events, business-specific events and recommended campaigns, with traceability showing which context documents influenced each post. Customers can see the reasoning behind recommendations.
- Feedback as evaluation and what's next
- Approvals, edits and regeneration requests serve as lightweight evals for quality. The team plans deeper attribution to point-of-sale data and expansion into omnichannel and digital out-of-home displays.
03From the post
“Listen to this episode on: Spotify | Apple Podcasts What if your small business could have a full marketing team—automated content calendars, customer segmentation, and channel-specific posts—without the headcount? In this episode of Just Now Possible, Teresa Torres talks with Chris O'Connor (CEO) and Jessica Valenzuela (Co-”
04Frameworks mentioned
Summary and takeaways written by PM Atlas; quotes are short excerpts. © the original author.