Overview
LuMay's architecture is a six-layer reference model with defined contracts between layers: Interface, Intelligence, Orchestration, Integration, Governance, and Analytics & ROI. Each layer has a single responsibility and a defined set of inputs and outputs, which is what lets you change a model, replace a connector, or add a channel without touching the rest of the stack.
For a US evaluation, the practical question is rarely "is this architecture good" - it's "where does this sit relative to what we already have, and what has to change on our side." The sections below walk through that mapping.
This is what 'platform, not project' means in practice.
The Six Layers
| # | Layer | Responsibility | Components |
|---|---|---|---|
| 01 | Interface | Channels agents are reached through | Voice (Twilio WebSocket), Chat, Microsoft Teams, Web portal, REST API |
| 02 | Intelligence | Reasoning and understanding | RAG, LLM reasoning (OpenAI / Anthropic), intent classification, structured outputs |
| 03 | Orchestration | Workflow execution and multi-agent coordination | Conversation flows, multi-agent delegation, HITL gates, versioning, NL builder |
| 04 | Integration | Connecting agents to external systems | Connectors (ticketing, email, docs), OAuth vault, MCP servers |
| 05 | Governance | Trust, compliance, and control | RBAC, Row-Level Security, audit logs, PII controls, guardrails, secrets management |
| 06 | Analytics & ROI | Measurement and value demonstration | KPIs, cost tracking, ROI dashboard, sentiment, quality scores, OTel traces, Prometheus metrics |
Mapping to Your Environment
Before scoping a pilot, most US evaluation teams work through this sequence against their own environment:
- Interface (01) - which channels do you actually need on day one (voice, chat, Teams, API), and which systems currently own those channels today?
- Intelligence (02) - which LLM provider(s) are already approved by your security team, and does your data governance policy allow the retrieval sources you plan to connect for RAG?
- Orchestration (03) - which actions need a human approval gate before they execute, and who is that approver in your organization?
- Integration (04) - which systems of record does the agent need read access to first, and which (if any) need write access?
- Governance (05) - what is your existing identity provider, and does your tenant-isolation requirement match row-level security at the database layer, or do you need a harder boundary?
- Analytics (06) - which business metric will you use to decide whether the pilot succeeded, and do you have a baseline for it today?
Illustrative Example
The following is an illustrative example of how a request might move through the six layers - it is not a specific customer deployment.
An internal assistant receives a request from an employee over chat (Interface). The request is classified and relevant context is retrieved via RAG (Intelligence). The orchestration layer determines the request requires an action in a ticketing system and checks whether that action is pre-approved or needs a human reviewer (Orchestration). If approved, the connector layer dispatches the action to the ticketing system using validated credentials (Integration), the action is logged with a correlation ID (Governance), and the outcome is recorded against the agent's containment-rate metric (Analytics).
Dependency Questions to Ask
- If the LLM provider has an outage, which layer degrades, and does the rest of the stack keep functioning?
- If a connector credential expires or a provider API changes, what is the failure path, and who is alerted?
- If an approval gate is skipped by mistake, is that visible in the audit trail after the fact?
- Who owns the decision to change a layer in production - engineering, security, or the business owner?
FAQ
Does changing LLM providers require re-architecting anything?
No - provider changes are contained to the Intelligence layer (02); agents, flows, and connectors are unaffected.
Can we deploy this entirely within our own cloud environment?
See Deployment for the specific models (SaaS, BYOC, private cloud, on-prem) and who manages infrastructure in each.
Where does tenant isolation get enforced?
At the Governance layer (05), via PostgreSQL Row-Level Security - see Security & Governance for detail.
What's the difference between this page and the business platform overview?
This page is the technical architecture reference. /platform covers the business case, pricing posture, and the four engines at a higher level.
Who should review this page before a pilot?
Typically a solutions architect or engineering lead, alongside whoever owns security/compliance sign-off for the systems being integrated.
Related
- Core Engine - the Intelligence layer (02) in detail
- Security & Governance - the Governance layer (05) in detail
- Deployment - how the architecture maps to infrastructure across deployment models
- Enterprise AI Framework - the broader operating model this architecture sits within
- US AI Platform hub - all 8 component pages