HFA × AI Aunties · Partnership Architecture
Scope Boundary & n8n Workflow Taxonomy
Based on AI Aunties Commercial Build & Deployment Handbook · Defines which n8n workflows serve the Auntie vs. which are downstream client territory
01 · Scope Boundary
02 · n8n Workflow Taxonomy
03 · Workflow Extension Documentation
04 · HFA Commercial Modes
The Architecture in One Sentence
The Auntie thinks. n8n stabilizes, routes, renders, logs, and executes. Nothing in the workflow layer rewrites, ranks, advises, decides, or alters what the Auntie produced.
AI Aunties delivers
Chat output
Advisory response · Analysis · Recommendation · Draft · Structured guidance
Downloadable artifact
Report · Word document · PDF · Formatted output · Client-facing work product
External brain + retrieval layer
Notion brain · Source library · Accepted/rejected outputs · Test logs · Use-case history · Structured knowledge base
+ Auntie Workflow Extension (if scoped)
Input triggers · Output rendering · Brain maintenance · Runtime governance — built by HFA, governed by AI Aunties handbook
Outside AI Aunties core scope
CRM updates
Slack / email notifications
Task creation in project tools
File movement and database updates
Dashboard creation
Operational workflows
Client-side automations
Third-party workflow orchestration
Who handles this?
Client does it themselves, or engages HFA directly. AI Aunties refers but does not govern. HFA sells as a separate engagement.
The strongest sentence from the handbook
"AI Aunties delivers the thinking product, the artifact, and, where scoped, the external brain and retrieval layer. What the client connects to those deliverables afterward belongs to the client or their automation provider unless AI Aunties has explicitly scoped and priced that workflow extension."
The cognition boundary rule
"Automation does not replace the Auntie, rewrite the Auntie, or make decisions on behalf of the Auntie. The Auntie remains the product runtime. The Auntie, operating through GPT, owns cognition, judgment, advisory reasoning, recommendations, language, and final client-facing content."
The implementation rule
"Technical partners may support implementation of the workflow module, but the automation remains governed by the AI Aunties Commercial Build & Deployment Handbook. The partner's role is implementation support, not product ownership or cognitive architecture."
Commercial responsibility matrix
Delivered by AI AuntiesDownstream — Client or HFA responsibility
Auntie-generated chat outputActing on that output inside client systems
Downloadable artifactRouting, filing, emailing, or distributing it
Notion brain + retrieval layerConnecting it to other tools or automations
Auntie package structureClient-specific workflow orchestration beyond scope
Scoped AIA workflow extension (HFA builds, AIA governs)Any automation not included in the AIA scope
Auntie Workflow Extension — Required Documentation
Every n8n workflow built as an Auntie Workflow Extension must be documented with these fields before it goes into production. AI Aunties owns the standard; HFA owns the implementation build tasks.
Required Field Description Example
Auntie Name
Required		 The specific Auntie the workflow supports. Must be a named, documented Auntie in the AI Aunties system. Renee (Commercial Real Estate Analytical Auntie)
Client / Use Case
Required		 The client context and specific business process being automated. Ardra (CBRE) — property analysis report delivery to client inbox
Workflow Purpose
Required		 The specific job this n8n workflow performs. One sentence. If it takes more than one sentence, split into two workflows. Renders Renee's analysis output as a formatted PDF and emails it to the client contact stored in the deal record.
Trigger Event
Required		 What starts the workflow. Must be a specific, observable event — not "when the Auntie is done." Webhook fired by GPT when Renee marks analysis complete; OR manual trigger by agent pressing "Send to Client" button in Notion.
Inputs Required
Required		 Files, data, forms, messages, or Auntie outputs the workflow needs to run. Be specific about format and source. Renee's analysis text (from Notion property page) · Client email address (from deal record in Supabase) · Property address for PDF filename
Actions Performed
Required		 Routing, rendering, notifications, updates, delivery, logging, and any other defined action steps. List each action. 1. Fetch analysis text from Notion · 2. Render PDF using template · 3. Email PDF to client contact · 4. Log delivery timestamp to Supabase
Systems Touched
Required		 Every system the workflow reads from or writes to. Include n8n, Notion, CRM, email, file storage, forms, or any other tools. n8n (orchestration) · Notion (Auntie output source) · Supabase (client record + delivery log) · Gmail (delivery) · PDF renderer
Cognition Boundary
Critical		 Explicit confirmation that this workflow does NOT rewrite, rank, advise, decide, or alter Auntie output in any way. If it does, it is not an Auntie Workflow Extension — it is a separate system that requires AI Aunties review. Confirmed. This workflow renders and delivers Renee's output verbatim. No text is modified, summarized, ranked, or altered. The PDF template applies formatting only.
Failure Behavior
Required		 What happens when the workflow cannot complete. Auntie output must never be lost or silently dropped. Define the fallback. If PDF render fails: alert sent to HFA and agent. Auntie output preserved in Notion. No silent failure. Manual delivery fallback via direct email.
Testing Requirement
Required		 How this workflow is validated before client use. Must include at least: happy path test, empty input test, and failure mode test. Happy path: full analysis → PDF rendered → email delivered. Empty input: no analysis text → workflow stops, alert fired. Failure: email API down → fallback alert triggered.
Owner
Required		 AI Aunties owns the product standard. HFA owns the implementation build. Both are named here. Product standard: AI Aunties (Sarita Myers) · Implementation: Human First Automation (Jerry Pender)
Cognition boundary — what n8n can and cannot do to Auntie output
✓ n8n MAY
✓Route the output to a person, system, or channel
✓Render the output as PDF, Word, or formatted email
✓File the output to Notion, Supabase, or Drive
✓Deliver the output via email, SMS, or portal
✓Log the output to the model registry
✓Trigger the Auntie with an input
✓Feed context from the Notion brain to the Auntie
✓Apply formatting templates (visual only, no text change)
✓Send notification that the output is ready
✓Run the QA loop (route for revision, not rewrite it)
✗ n8n MAY NOT
✗Rewrite or paraphrase Auntie advisory content
✗Rank or select between Auntie-generated options
✗Add recommendations not generated by the Auntie
✗Override the Auntie's judgment or conclusions
✗Summarize Auntie output with a different AI model
✗Make decisions about what content the client receives
✗Alter the Auntie's tone, voice, or language
✗Generate advisory content independently
✗Decide which clients receive which level of output
✗Run a second AI model on the output without AIA approval
HFA's Three Commercial Modes in the AI Aunties Architecture
HFA operates differently depending on where in the architecture the work lives. Each mode has different scope, pricing, governance, and client relationship dynamics.
Mode 01
Auntie Workflow Extension Builder
HFA builds n8n workflows that serve the Auntie product directly — input triggers, output rendering, brain maintenance, and runtime governance. These live inside the AI Aunties commercial scope and are governed by the AI Aunties handbook.
What HFA builds here:
· Input trigger workflows (form → Auntie)
· Output rendering workflows (Auntie → PDF/Word)
· Notion brain maintenance workflows
· QA loop workflows (Maxine/Lux)
· Model registry logging workflows
· Context retrieval workflows
Governed by AIA handbook Co-sold with AIA package
Mode 02
Downstream Automation Partner
Once the Auntie delivers its output, anything the client wants to do with that output in their own systems is outside AI Aunties core scope. AI Aunties refers the client to HFA. HFA sells this directly. No AIA governance applies.
What HFA builds here:
· CRM updates from Auntie output
· Task creation from recommendations
· Slack/email notification routing
· Database writes from artifacts
· Dashboard creation from reports
· File routing and operational workflows
HFA direct sale AIA refers, HFA closes
Mode 03
Independent HFA Engagement
Clients who don't have an Auntie but need automation. Pure HFA business. AI Aunties is not involved unless HFA identifies an opportunity to introduce the Auntie product — which becomes a referral in the other direction.
What HFA builds here:
· All workflow types from the catalog
· Industry-specific automation builds
· Solopreneur workflow subscriptions
· Full business automation systems
· Custom integrations and pipelines
· Everything in the HFA product catalog
Pure HFA business Reverse referral to AIA
The referral loop between modes
AIA client
needs automation
→
AIA refers to HFA
Mode 01 or 02
→
HFA builds
extension or downstream
HFA client
needs cognitive layer
→
HFA refers to AIA
Mode 03 → AIA sale
→
AIA delivers Auntie
HFA stays as ext. builder

Based on AI Aunties Commercial Build & Deployment Handbook — The Auntie thinks. n8n executes. HFA builds both ends of that relationship.