Reasoning Layer for Enterprise AI Agents

It is the infrastructure that sits between an enterprise's systems and its AI agents and provides the structured model an agent needs to make precise decisions rather than guess from documents. A reasoning layer returns the chain of decisions, owners, triggers, and blockers that bears on a question — not a list of links the agent has to read. Naboo's reasoning layer is built on a Decision Graph: a queryable model of who decides what, what triggers each decision, what blocks it, and what depends on it.

A Decision Graph is the data structure that powers the reasoning layer. Unlike a knowledge graph (which represents entities and relationships) or a vector index (which represents text similarity), a Decision Graph represents decisions as first-class nodes. Each node has an owner, a trigger, blockers, dependencies, and supporting evidence from the systems where the work lives (tickets, PRs, code, Slack, internal services). Agents traverse it to answer questions search cannot — for example: 'what's blocking the Q3 ship of checkout v2, who owns each blocker, and what's the next decision in the chain?'

RAG (Retrieval-Augmented Generation) retrieves the document chunks most semantically similar to a query and hands them to a language model. The reasoning layer returns the structured chain of decisions, owners, triggers, and blockers the agent needs to act. RAG can find a document that mentions 'feature flag' or 'deploy'; the reasoning layer can answer 'what's blocking the deploy, who owns it, and what triggered the current state' — because no single document contains that answer. In benchmarks Naboo is 97% more accurate than RAG, uses 90% fewer tokens, and is 10x faster.

Enterprise search platforms index documents across SaaS and return the most relevant ones plus a generated summary. They make search smarter for human knowledge workers. The reasoning layer models how the company itself decides and ships — who decided, what triggered, what blocks — and exposes that model to AI agents. Search returns links. The reasoning layer returns the chain.

Two patterns recur. (1) 'What's blocking [X] from shipping?' — the agent traverses the chain of open decisions (security review, infra capacity, customer comms, etc.), returns the owners, statuses, and triggers, and surfaces the next decision. (2) 'Why did [X] happen?' — the agent walks the chain of decisions that led to a current state (rollout, incident, policy change), returns the decision-makers and the supporting evidence (PRs, tickets, Slack threads). Both questions are unanswerable by document search alone because the chain of joins is not written down in any single source.

In two tiers. The Topic Graph indexes the entire enterprise dataset across tools, links documents by explicit and implicit references, resolves entities so a person or repository maps to a single identity everywhere, and enforces document-level permissions on every node. The Decision Graph sits on top: a Forward Deployed Agent — a specialist in ETL and data science — sits with the customer to encode the organization's private definitions of decisions, owners, triggers, and blockers across the systems involved. Both tiers are continuously updated.

Two surfaces. GraphQL for agents and applications that want to issue structured queries directly. An MCP (Model Context Protocol) server for agents that prefer to express intent in natural language and let a translation layer construct the structured query. Both return structured answers — decisions, owners, statuses, supporting evidence — that an agent can act on directly.