Operating Intelligence Platform
The platform is a client-specific operating intelligence layer over the workflows, systems, documents, metrics, decisions, and approved data paths that matter. Source inventory, canonical entities, semantic and truth definitions, document intelligence, permissions, query, tools, observability, and the agent layer that sits on top — in that order.
Architecture
Sources are governed before agents are built. Permissions sit between query and source. Every answer carries citations. The full thirteen-layer model lives inside five reading groups so an architect can scan it in under a minute.
01
Sources & access
Source inventory
Every system, report, and document set the workflow depends on, with owner, status, and access path.
Data contracts
Source-to-canonical mappings, freshness expectations, exception rules, and reconciliation logic.
Source access profiles
What the platform may read, retrieve, export, or compute from each source. Permitted and prohibited actions are explicit.
02
Canonical entities & semantic truth
Canonical entity model
Stable platform identities for property, lease, tenant, work order, vendor, KPI, and the entities your workflow depends on.
Standardized data
Curated views and datasets aligned to the canonical model, with lineage to source.
Semantic & truth layer
Definitions, formulas, manual adjustments, reconciliation rules. Every metric carries a truth profile.
03
Documents & knowledge
Document & knowledge layer
Governed unstructured sources (leases, asset plans, board materials, SOPs) with classification, retrieval rules, and citation requirements.
04
Permissions, query, tools
Permission & governance layer
Who may view, query, retrieve, draft, approve, export, trigger, or administer. Agents inherit scoped, auditable permissions.
Query layer
Approved users ask questions against governed sources. Answers return with source references, freshness, definitions, confidence, and escalation paths.
Tool & action layer
Approved APIs, workflow actions, retrieval tools, and write-back paths. Each tool has a typed contract.
05
Observability, agents, AgentOps
Observability & audit layer
Traces, source references, tool calls, model calls, permission decisions, approvals, incidents. Required for compliance, debugging, and trust.
Agent layer
Controlled AI capabilities operating on top of the governed platform. Permissions and behaviors are scoped, evals are required, traces are mandatory.
Managed AgentOps layer
The operating model after deployment: eval regression, source freshness, change control, incident response, expansion or retirement.
Minimum Viable Foundry
The first platform build is always a Minimum Viable Foundry: one approved workflow, one or two source systems, a small canonical entity model, a limited semantic layer, a governed document set, queryable views with cited answers, and one or two agent-ready use cases. Expansion is a separate decision under change control.
First wedge scope
Workflow boundary
One approved domain (asset management, leasing pipeline, property operations, capital projects, finance close, investor reporting, ESG, vendor intelligence, or similar).
Source systems
One to two primary sources. Source inventory metadata for the rest.
Canonical entities
Only the entities the first workflow depends on. Defer the rest.
Semantic layer
Definitions and truth profiles for the metrics this workflow uses.
Document set
Governed retrieval index over the policies, SOPs, plans, and memos this workflow cites.
Permission model
Scoped, observable, revocable. Agents do not inherit broad human access.
Queryable views
Specific, evidence-grounded questions an asset manager or executive would actually ask.
Pilot agent capabilities
One or two low-risk agent capabilities (read, summarize, classify, draft, recommend) on the governed layer.
Data access
Discovery and the first scaffolds run on metadata, redacted samples, and synthetic examples. Production credentials cross only after approval, only into your environment. Each stage has its own purpose and its own controls.
| Stage | Title | Data used | Purpose | Controls |
|---|---|---|---|---|
| 01 | Metadata scaffold | System names, source inventory, owners, field lists, workflow maps, KPI definitions, synthetic examples. | Build the operating model without sensitive records. | No production credentials. No bulk data. No live integrations. |
| 02 | Redacted samples | Redacted exports, limited sample records, sanitized documents. | Validate schemas, mappings, query behavior, evals, and platform assumptions. | Written approval. Data minimization. Retention rule. |
| 03 | Client-controlled environment | Approved source connections inside the client cloud or tenant. | Build the real operating intelligence layer. | Security review. Identity. Audit. Network. Logging. Retention. |
| 04 | Production operating layer | Approved production sources and governed documents. | Make the approved domain queryable. | RBAC/ABAC. Monitoring. Lineage. Access review. Incident path. |
| 05 | Agent layer | Governed sources, tools, semantic layer, retrieval layer, evals. | Deploy agents safely on top of the platform. | Tool contracts. Guardrails. Traces. Eval thresholds. Revocation. |
Example queries
Each answer carries source references, freshness, the definition used, confidence, and an escalation path. The platform refuses, qualifies, or escalates when truth is disputed, fragile, or unknown.
Query
“Which properties are below budget this quarter, and why?”
Query
“Which lease expirations create the most risk in the next two quarters?”
Query
“Which capital projects are over budget and missing owner commentary?”
Complementary stack
The Operating Intelligence Platform is a governed semantic layer over your existing systems, not a new system of record. It is designed to be deployed inside your cloud, follow your identity model, comply with your security review path, and complement the platforms you already operate.
Cloud & data platform
Real estate operating systems
Enterprise & operations
Documents & content
BI & analytics
What the platform does not do
Where this lives
The Operating Intelligence Platform is the technical asset delivered by Operating Implementation, after the Operating Diagnostic decides which workflow belongs and which controls it requires. After launch, Managed AgentOps operates the platform and the agents on top of it.
Next step
Start with one operating area where better intelligence could protect NOI, reduce risk, improve throughput, or make portfolio signal visible sooner. We will sketch the first wedge with you in 30 minutes, and you will leave with three candidate workflows and a no-pressure decision packet.