01
AI has information.
RAG, search, databases, and enterprise systems can retrieve what the company knows.
Intent infrastructure for enterprise AI
AI can access your company.It still doesn't know what your company is trying to accomplish.
Nucadia gives every AI system a governed answer, grounded in your business strategy and current situation, to what matters and what it should do next.
Organizational intent · current context · constraints · capabilities → decision-ready Action Context
SCROLL TO RESOLVE
The problem
More data doesn't mean better context.
Every AI application reconstructs organizational intent independently.
RAG + INTENT
Retrieval finds information. Nucadia determines what should govern the decision.
RAG, search, databases, and graphs can still supply candidate information. Nucadia evaluates that information against the current objective, authority, task, and situation before it becomes decision context.
02
AI has tools.
Agents increasingly have access to dozens or hundreds of operations across enterprise systems.
03
The counterpoint
AI still has to infer what matters.
Goals, priorities, constraints, prerequisites, and appropriate actions remain scattered across strategy documents, policies, workflows, prompts, and people's heads.
One decision, different context
What matters for this decision, right now?
A customer-success agent has permission to read account data, communicate with the customer, prepare a renewal, and offer a discount. But permission alone does not tell the agent what makes sense now.
ACTORCustomer Success Manager
TASKHandle renewal
SITUATIONAcme Corp
STATERenewal in 14 days · Strategic account · Critical service case open
Active intent
Retain the strategic account while protecting operating margin.
Mandatory constraint
Resolve the critical service failure before commercial negotiation.
Resolved capabilities
support_case.readrecommendedcustomer.emailapplicablecase.escalateconditionalspecialist.scheduleconditionalrenewal.generateblockeddiscount.offernot applicableThe agent had permission to offer a discount. Nucadia gives it the context to understand why that is not the right action yet.
The control plane
Turn organizational intent into infrastructure.
Nucadia creates a governed layer between enterprise strategy and runtime AI behavior.
In practice, Nucadia turns business strategy and operating rules into a reviewable specification, lets your organization decide what is authoritative, and applies that specification whenever AI needs to make a decision.
ARCHITECTURE PRINCIPLE
AI helps create and maintain the specification. The Resolver applies it. Runtime AI can assist with bounded interpretation without becoming the authoritative source of organizational intent.
AUTHOR
AI helps structure the specification.
Strategy, policy, role definitions, workflows, schemas, data catalogs, and tool descriptions become proposed Intent Objects, Task Schemas, context relationships, source-authority rules, prerequisites, and Capability Objects.
GOVERN
Authority stays with the organization.
AI-generated proposals do not automatically become authoritative. Nucadia keeps approval state, provenance, version, scope, and effective time explicit before configuration becomes active.
RESOLVE
The Resolver applies the specification.
At runtime, Nucadia evaluates the actor, task, situation, current state, and authorized capabilities against the governed specification and produces a traceable Action Context. Runtime AI may help classify requests, extract entities, or retrieve candidates, but those interpretations remain subject to governed resolution.
Runtime resolution
What matters for this decision, right now?
Nucadia compiles the intent, information, constraints, unknowns, and available actions that should govern the current decision into a versioned Action Context.
REQUESTINTENTCONTEXT + PREREQUISITECAPABILITIES
Runtime / resolve
What applies to this decision, right now?
RES-0001482
ACTORCustomer Success Manager
TASKHandle renewal
SITUATIONAcme Corp
Active intent
Retain the strategic account while protecting operating margin.
Mandatory constraint
Resolve the critical service failure before commercial negotiation.
Selected context
Account tierStrategicCRM / current
Renewal14 daysCRM / current
Service caseP1 · OpenServiceNow / current
Blocking prerequisiteblocking
Remediation state
Current remediation outcome has not yet been established.
Applicable capabilities
support_case.readrecommendedRequired to establish current remediation state.
customer.emailapplicableAppropriate for status communication and expectation setting.
case.escalateconditionalUse if the service case remains unresolved or SLA risk increases.
specialist.scheduleconditionalUse if specialist intervention is required.
renewal.generateblockedCommercial negotiation is blocked by unresolved remediation state.
discount.offernot applicableNot appropriate as the first response to unresolved service failure.
SAFE HANDLING
When context is insufficient, Nucadia does not have to guess. A missing prerequisite can be represented as blocking or nonblocking. The Action Context can direct the downstream system to clarify, abstain, defer, escalate, or continue only with explicitly permitted subtasks.
TASK SCHEMA
Required decision fields, dependencies, unknowns, and stopping conditions
COVERAGE
Included, irrelevant, withheld, unavailable, stale, or conflicting
SOURCE STATE
Authority, freshness, provenance, and current version
REPRODUCIBILITY
Bundle, intent, resolver, and source versions with re-resolution triggers
DELIVERY
Canonical model-agnostic context for copilots, agents, workflows, and applications
Agents + capabilities
Give agents the capabilities that make sense now.
Existing IAM, policy, and application systems remain authoritative over what an agent may access or invoke. Nucadia conditions the downstream agent's planning on intent, current state, prerequisites, constraints, and the already-authorized capability set.
AUTHORIZATION ANSWERSWhat may this agent do?
Existing IAM, policy, and application systems remain authoritative.
NUCADIA ANSWERSWhat makes sense to do now?
Resolve applicability against intent, prerequisites, and current state.
Nucadia governs the decision environment and conditions for action. The downstream agent plans and sequences the action.
Nucadia surfaces the best-fit capabilities for the task at hand, a high-recall, low-noise set the downstream agent can select and sequence from.
POTENTIAL EFFICIENCY BENEFIT
Narrower context and capability sets may reduce unnecessary input tokens, irrelevant tool schemas, retrieval, and failed agent loops. This has not yet been validated in production.
Evidence
Decision quality comparable to an ideal expert brief.
92.8%
Nucadia decision context
action correctness
41.9-47.5%
Raw + conventional context
action correctness
91.7%
Ideal expert brief
action correctness
ACTION CORRECTNESS
Did the AI choose the right next action given the intent, constraints, permissions, and information it should have used?
Behavior correctness: 95.6% · reversal accuracy: 97.5% · null stability: 100% · explicit unauthorized context exposure detected: 0 · three acting model families.
Responses were evaluated by model families other than the model that produced the response.
Responses were evaluated by model families other than the model that produced the response.
CONTROLLED TECHNICAL VALIDATION
This was a sealed synthetic enterprise benchmark across three model families, conducted under the earlier working name IntentSpec. Nucadia is the current public brand; the underlying condition and results are unchanged. It provides evidence for the context-resolution mechanism under controlled conditions. It is not production customer evidence, a security certification, or proof of business outcomes.
The larger system
Change strategy once. Let AI behavior follow.
Nucadia can propagate governed business strategy into runtime decisions, observe how AI behaves under that intent, and surface where execution begins to diverge.
Local intent, exceptions, tool use, and outcomes can create review signals without automatically rewriting authoritative organizational intent.
Design partners
Help shape the intent layer for enterprise AI.
We are working with a small number of teams building enterprise agents, copilots, and AI platforms to test Nucadia against real decision workflows.
01One agent or copilot
02One task family
03One intent structure
04One capability catalog
05One or two enterprise context sources
The goal is not a broad platform rollout. It is to test whether governed intent and context materially improve one consequential AI workflow.