AYAI systems portfolio Discuss a principal AI role
AI SaaS ProductLive product

PillarQuillAgentic Content OS

RESEARCHSTRATEGYCREATEPUBLISH CONTENT LOOP
My roleFounder, product, and AI architect
ContextCreators and agencies
System classAI SaaS Product
FocusDurable content and publishing state
Brief

The problem

Research, positioning, drafting, approval, scheduling, publishing, and learning were split across tools, so creators and agencies lost context at every handoff.

Open live studio

Constraints

A workflow that outlives a prompt

01The first product wedge is LinkedIn, but the domain model must extend to Instagram, YouTube, and X.
02Content state, calendar, analytics, and platform connection state must persist independently of any model provider.
03B2C creators and B2B agencies need different approval and multi-brand workflows.
04Agent output must remain editable, explainable, and connected to source research.
Architecture

The content lifecycle

Research, briefs, drafts, approvals, and publishing state remain durable product artifacts with ownership and history.

PillarQuill / logical architectureSanitized view
01ResearchSignals, sources, audience
02StrategyPillars, angle, campaign
03CreationDraft, critique, compliance
04OperationsCalendar, approvals, publish
05LearningAnalytics, feedback, memory
Decisions

Product decisions

01

Model the content lifecycle

Persistent content objects and workflow states make the product useful beyond a single generation call.

02

Agents collaborate around explicit artifacts

Research, brief, draft, review, and channel adaptation are inspectable outputs with ownership and history.

03

Publishing integrations are adapters

Channel-specific APIs are isolated from the core content model so the workflow can expand without rewrites.

04

Analytics closes the loop

Performance signals update future recommendations and content strategy through the same workflow.

Production

Reliable publishing

01Multi-tenant ownership and workspace boundaries.
02Asynchronous generation and publishing jobs with retry/idempotency.
03Connection state and token refresh for social integrations.
04Audit/history for drafts, approvals, and scheduled actions.
05Cost and usage tracking across agent workflows.
Product architectureAgent workflowsFastAPIMulti-tenant SaaSRedisCelerySocial integrationsContent intelligenceAnalyticsWorkflow stateFounder execution
Outcome

A live product for managing research, review, publishing, and learning as one connected content operation.

Relevant problem?

Turning an AI workflow into a durable product?

Discuss AI product architectureAll work
Next case study

Placement OS