Ontology

One executable model of your business

A live model of how your business runs. It defines the levers that drive your results and lets people and agents act on them safely.

Objects and relationships

What your business runs on and how it connects, mapped from your systems.

placesdelivered toships asleaves from

Expressions and ML models

Calculations and predictions computed on your objects.

p(on time) = f(carrier, weather, dock capacity)
.5.710.73

Governed actions

The changes people and agents can make, with approvals, write-back, and reconciliation.

Request · Supply agent
Expedite shipment
Air on the last leg, $3,900
ApproveApproved

Memory

Policies, interviews, and documents, each linked to the objects they explain.

Northwind FreightCustomerdelivery terms4.2 Late delivery credit
Dana RuizReno dock planner

Dock 3 only takes six trucks before noon. Anything booked after that waits for the afternoon crew.

Supply agentRun 7,312read SH-40118on-time0.64propose expedite
explainscontext

Permissions

Roles decide who can see and change every object, property, action, and document. Agents act with the permissions of the person they work for.

Logistics planner
  • see margin
  • propose
  • run expedite
Supply agent
  • see margin
  • propose
  • run expedite
Account manager
  • see margin
  • propose
  • run expedite
gates
  • CRM
  • ERP
  • Supply chain planning
  • Email and documents
  • Transport system
  • Interviews, reviews, approvals
mapped inwrites back

AI FDE Builds every part with your team, and keeps it running.

  1. Gathers context about the enterprise
  2. Builds and maintains the ontology
  3. Deploys agents, applications, and workflows
  4. Monitors the ontology for new optimizations and extensions
  1. 1

    Objects and relationships

    The canonical objects that make up your business and how they connect to each other with typed relationships.

    signsplacesincludesships asleaves fromis stocked atCustomerContractOrderProductShipmentFacility
    ObjectOrder14 properties
    PropertyTypeSourceAs of
    promised_dateDateERPNov 1, 9:12
    valueMoneyERPNov 1, 9:12
    statusOpen, Allocated, Shipped, DeliveredCanopy actionsNov 1, 9:40
    delivery_termsTextContract, section 6Mar 3
  2. 2

    Expressions and ML models

    Enrich your objects with functions that make first class the logic of your business. From expressions to powerful ML models, you can encode the levers that actually drive your results.

    ExpressionOrder.marginPercent
    (value − product.unit_cost × quantity − shipment.freight_cost)
    ÷ value
    Used bySupply agent, Order risk app, Pricing approvals
    SO-8821418.4%
    Modelon_time_probabilityVersion 4
    Bound to
    Shipment
    Writes
    Shipment.on_time_probability
    Inputs
    Carrier history, route, weather, dock slots at the facility
    Checked
    Backtested on the last four quarters before release
    SH-401180.64 · At risk
  3. 3

    Governed actions

    The changes people and agents are allowed to make. Each action has its inputs, its checks, who approves it, and the system it writes back to. After it runs, the ontology confirms the change happened.

    ActionExpedite shipmentGoverned write
    ShipmentSH-40118
    ChangeAir on the last leg
    Cost$3,900
    Who can run it
    Logistics planners; the supply agent can propose it
    Approval
    Supply chain director over $2,500
    Writes to
    Transport management system
    Reconciles
    Reads the transport system back to confirm the change happened, and flags it if not
    Logged
    Who ran it, why, and the values before and after
    Waiting on the supply chain director
  4. 4

    Memory

    Policies, interviews, documents, and agent traces, linked to the objects they explain. Agents cite it, and AI FDE uses it to see what should move into the model next.

    Customer delivery termsPDF

    Orders delivered more than two days after the promised date earn a credit of 5% of the order value.

    Linked toContractOrderExpedite shipment
    Interview with a Reno plannerInterview

    Dock 3 can't take more than six trucks before noon. Anything booked into a late-morning slot usually slips a day.

    Linked toFacilityShipment
  5. 5

    Permissions

    Every object, property, action, and document has roles that can see it or change it. Agents act with the permissions of the person they work for. Because permissions live in the ontology, every application and workflow enforces them the same way.

    PermissionsRoles on this sliceDown to the property
    Account managerLogistics plannerSupply agent
    Order.marginViewHiddenView
    Contract.pricingViewHiddenHidden
    Shipment.carrierViewEditView
    Employee.pay_rateHiddenHiddenHidden
    Expedite shipmentNoRunPropose
    Agents get roles like any other user. Pay rates stay with HR.

AI FDE builds and maintains your ontology

It extends the ontology as your business changes. Every change comes with its evidence, and your team approves what ships.

  1. Builds it from your systems and interviews

    AI FDE maps your systems onto object types, interviews the people who do the work, and drafts the first expressions and actions.

  2. Extends it as the business changes

    New sources, new object types, new rules. AI FDE proposes each change, and your team reviews it like any other.

  3. Validates models against history

    Before a model's output lands on an object, AI FDE backtests it on outcomes you already know.

  4. Proposes new actions

    When agents keep escalating the same case, AI FDE drafts the action, its approvals, and its permissions for your sign-off.

  5. Keeps memory linked

    New policies, interviews, and traces get linked to the objects they explain, so nothing important sits in a folder.

More on AI FDE

Safely deploy powerful AI

Agents, applications, and workflows inherit the ontology’s actions, permissions, and audit trail.

Governed actions

Agents and people change your systems only through actions you've defined, with approvals where they matter and a check that each change happened.

Permissions enforced

Every read and every write is checked against roles, down to a single property.

A full audit trail

Every run records what was read, what was proposed, who approved, and what changed.

Real work, end to end

Because those controls hold, agents can work whole queues, apps ship in days, and workflows span teams.