AI-Powered Marketing Platform for Small and Medium Businesses

mowie · E-commerce · 2025 · source Agentic: no · promotional bias: low · source-backed · extracted with prompt 0.5.4 (schema v2)

Generated from extraction.json — do not edit by hand. Every claim cites verbatim source evidence; stated means the quote directly asserts the claim, implied means it supports it indirectly.

Problem

Domain: SMB marketing automation

Mowie wanted to automate practical marketing work for small restaurants, retailers, and e-commerce businesses whose owners bought marketing tools but lacked time and expertise to use them consistently.

Constraints:

• small-business owners have limited time to configure segmentation, email, and social tools
• business context is spread across websites, reviews, commerce systems, point-of-sale systems, weather, inventory, and local events
• marketing plans need weekly and quarterly cadence rather than real-time chat responses
• customers must remain able to approve, edit, decline, or stop scheduled content
• performance feedback is noisy because post-to-sale attribution is not exact

Architecture

Mowie builds a marketing operating loop around a business dossier. It ingests public and API data, including reviews, website/catalog information, commerce/POS signals, weather, inventory, and business calendars. LLMs and third-party APIs turn that context into loosely structured intermediate documents such as brand/business and audience dossiers. A hierarchy of documents is regenerated when upstream inputs change, and that context feeds quarterly calendars, weekly content plans, and nightly publish decisions. Customers receive approval emails and can accept, edit, decline, or stop queued posts. Performance and customer approval feed back into future planning.

Diagram confidence: reconstructed (reconstructed from prose).

flowchart LR
  third-party-api-connectors["third-party-api-connectors"] --> business-data-ingestion["business-data-ingestion"]
  business-data-ingestion --> brand-business-dossier["brand-business-dossier"]
  business-data-ingestion --> audience-dossier["audience-dossier"]
  brand-business-dossier --> document-dependency-hierarchy["document-dependency-hierarchy"]
  document-dependency-hierarchy --> calendar-planner["calendar-planner"]
  brand-business-dossier --> calendar-planner
  audience-dossier --> calendar-planner
  calendar-planner --> content-generator["content-generator"]
  content-generator --> approval-workflow["approval-workflow"]
  content-generator --> publishing-queue["publishing-queue"]
  approval-workflow --> publishing-queue
  approval-workflow --> performance-feedback["performance-feedback"]
  performance-feedback --> calendar-planner

Components

Data flow

A customer provides a business context. Mowie collects public web data, customer reviews, product/service catalog information, point-of-sale and commerce data, and daily signals such as weather or inventory. LLM and API steps summarize this into brand, business, and audience documents. When source inputs change, affected documents and dependent downstream artifacts regenerate. The planning layer creates quarterly calendars, weekly content, and nightly publish queues for email and social channels. Customers approve through email or calendar controls, and sales/performance signals feed back into future recommendations.

Why this architecture

Mowie's architecture is cadenced and document-centric because SMB marketing is recurring work, not a single answer. The dossier layer preserves business and audience context that can be reused across channels. A dependency hierarchy limits regeneration to downstream artifacts affected by changed inputs. Human approval protects customers from unwanted posts while keeping the owner workload low. Performance feedback is incorporated because the best evaluation for marketing content is whether it actually drives engagement or sales.

Models used: large language models · LLM-generated intermediate documents · templated prompts

Scale

• The founders' prior experience covered 45 customers, of whom only two actively used their paid marketing platforms.
  “we grew to 45 customers and we saw that out of 45 only two were really using their marketing platforms that they had paid for.”transcript, stated
• Mowie looks at about 80 artifacts to form a business and audience dossier.
  “about 80 artifacts that we take a look at to be able to create a what we call a brand and business dossier and an audience dossier around that particular company.”transcript, stated
• The planning cadence has quarterly, weekly, and nightly layers.
  “we do have a sort of a quarterly cadence for the overall calendar and then a weekly cadence in terms of creating the content itself and then a nightly cadence in terms of determining what needs to be published tomorrow”transcript, stated
• Daily data can include weather and inventory signals used to adjust marketing content.
  “So, we haven't really talked about it, but even things like weather or like inventory control, right?”transcript, stated

Alternatives considered (roads not taken)

Expect SMBs to use conventional marketing platforms themselves

Rejected because: The founders observed extremely low adoption even among customers who had paid for those tools.

Generate marketing calendars in real time

Rejected because: The system uses quarterly, weekly, and nightly cadences because campaign planning and publishing are scheduled workflows.

Force all intermediate reasoning into rigid structured data

Rejected because: Most intermediate documents are loosely structured except where customer editing or API data requires more structure.

Publish automatically without customer approval

Rejected because: The product keeps customers in the loop and supports rollback or stopping scheduled content.

Patterns

context-compaction (context)

Mowie compresses many public, review, commerce, and API inputs into intermediate dossiers that become reusable context for planning and generation.

batch-offline-processing (orchestration)

The platform generates marketing calendars and posts on quarterly, weekly, and nightly cadences rather than as a real-time chat interaction.

human-in-the-loop-gate (governance)

Weekly generated content requires customer approval, and customers can edit, decline, or stop queued content.

data-quality-loop (data)

The system detects when upstream documents or inputs change enough to regenerate dependent artifacts.

eval-driven-development (evaluation)

The product treats approval and performance feedback as evaluation signals, including whether marketing content affected sales or engagement.

Tradeoffs

Use recurring scheduled workflows for marketing planning.

Chose quarterly, weekly, and nightly cadences over real-time generation for every calendar decision.

Rationale: Marketing calendars and publishing are planned workflows, and scheduled cadence makes approval and queueing practical.

Keep humans in the approval loop.

Chose email/calendar approval with rollback over fully automatic publishing.

Rationale: Businesses need a low-effort way to confirm or stop posts when conditions change.

Use loosely structured intermediate documents.

Chose documents generated by LLMs and third-party APIs over rigidly structured representations for every inference step.

Rationale: The source material is varied and most intermediate reasoning does not need customer editing, so loose structure is more practical.

Gotchas (what bit them)

SMB customers often buy marketing tools but do not use them.

Root cause: They lack time and practical marketing expertise, even for basic segmentation and email workflows.

Resolution: Mowie shifts from tool provision to done-for-you planning and content generation with lightweight approval.

Open-web business context does not naturally fit a clean schema.

Root cause: Reviews, articles, business descriptions, and API feeds arrive with different formats and levels of structure.

Resolution: Use loosely structured intermediate documents, keeping structured data mainly where API data or customer edits require it.

Business changes can invalidate downstream marketing assets.

Root cause: A change high in the document hierarchy, such as a new location, changes dependent local context and customer segments.

Resolution: Detect changed inputs, regenerate affected higher-level documents, and cascade regeneration to dependent artifacts.

Post-to-sale attribution is not exact.

Root cause: Marketing performance depends on sales and engagement signals with noisy causal links.

Resolution: Use statistically meaningful inference and approval/performance feedback rather than claiming perfect attribution.

Evaluation

Mowie evaluates marketing output through business approval and observed performance signals, using customer acceptance and sales/engagement inference as feedback loops.

Metrics: customer approval of weekly content · post engagement · sales or segment-level sales impact · content decline/edit signals

Outcomes

• SMB marketing work is reframed from tool usage to automated planning and content execution
• business and audience context can be regenerated when upstream facts change
• customers can approve weekly content from email instead of operating a complex marketing platform
• performance and approval signals feed future marketing decisions

The transcript gives early-stage operating context rather than large production numbers: three target SMB categories, about 80 artifacts per business dossier, and quarterly/weekly/nightly planning cadence.