# AI Flagship Architecture — Milestone 34

Vexlora treats AI as an application-wide operating layer, not a standalone novelty page.

## AI surfaces

- AI Command Center dashboard: adoption, agents, approvals, domain impact and model/runtime posture.
- Global AI Copilot: workspace-scoped assistant with explicit context selection and agent handoff UI.
- AI Agents: catalog, builder, autonomy policy and execution history.
- CRM Revenue Copilot: deal scoring, account evidence and next-best action workflows.
- Product AI Studio: product naming, descriptions, bullets, category, attributes, tags and SEO preparation.
- Product Image Studio: background removal/replacement, cleanup, enhancement, lifestyle scenes and marketplace variants.
- SEO Copilot: keyword intent, metadata, SERP preview and optimization recommendations.
- Support Copilot: ticket summary, sentiment, priority and grounded reply direction.
- Analytics Investigator: anomaly explanation, evidence, confidence and follow-up analysis.

The existing AI Chat, Writer, Code, Image Generator, Summarizer, Translator, Prompt Library, Models, Requests, Tokens, API Usage and Cost screens remain available.

## Cross-module experience

AI entry points are intentionally embedded into existing admin workflows:

- Product Create/Edit → Product AI Studio, Product Image Studio and SEO Copilot.
- CRM Customer 360 → CRM Revenue Copilot.
- Support Ticket Details → Support Copilot.

This makes AI feel like an operating layer across the admin rather than a disconnected app.

## Integration boundary

Milestone 34 is frontend/sample/API-ready. It does not bundle provider API keys, model SDK secrets, vector databases, Prisma, autonomous write execution or server-side agent orchestration. Buyers can connect OpenAI-compatible, Anthropic-compatible, Gemini-compatible, self-hosted or custom AI gateways behind these UI contracts.

High-impact actions are represented with human-approval UI to encourage governed production implementations.