The app reads orders and shipments. The agent proposes a governed action for a planner to approve. The workflow fires when on_time_probability drops and runs its action through the same controls.
Long-running agents that handle complex operational tasks
Supply agentDone with SH-40118
SH-40118 will probably miss Nov 14. on_time_probability fell to 0.64 after a port delay.
Read the shipment, its order, and the facility
Checked the contract: a 5% credit if it's late
Priced air freight on the last leg
Expedite shipmentNeeds approvalApproved
Cost$3,900
ArrivesNov 14
Margin after15.2%
ApproveSee other optionsOver $2,500, so it goes to the director
Priya, supply chain director
Booked. The account team has the new route and date.
Applications that surface key decisions and actions to your team
Order riskSupply chain
QueueShipmentsMargin
Orders at risk38
Revenue at risk$1.2M
On-time rate96.1%
Order
On time
Status
What the agent did
SO-88214Northwind Foods
0.64
Needs approval
Drafted an expedite
SO-88190Alder Medical
0.71
Agent working
Shared a new date
SO-88177Kestrel Retail
0.82
Waiting on pricing
Sent to pricing
SO-88165Bramble Outdoor
0.93
Resolved
Rebooked a dock slot
SO-88150Cobalt Labs
0.97
On track
No action needed
Workflows that automate critical business processes
Whenon_time_probabilityis under0.70
theSupply agentdrafts options
Over$2,500needsPSSupply chain director
RunRebook dock slotand tellAccount team
Run 214 finished in 6 min
All of it runs on one ontology
ShipmentSH-40118
carrierMaersk
etaNov 16
OrderSO-88214
customerNorthwind Foods
promisedNov 14
FacilityDayton DC
open_dock_slots0
regionMidwest
on_time_probabilityML model
etaNov 16
promisedNov 14
open_dock_slots0
0.64Below 0.70
Expedite shipmentSent for approval
Runs when on_time_probability is below 0.70
The supply agent is allowed to run it
PSOver $2,500, so Priya approves it
AI agents
Agents for complex operational work
Long-running, multi-party. Our agents specialize in long-horizon tasks that require safe orchestration across multiple parties.
Powered up with incredible context. Ontology gives agents an environment for understanding and executing operations.
Governed by your ontology. Actions are gated by ontology state, enforcing critical business logic for high-stakes transactions.
Naturally self-improving. Agent work automatically contributes to ontology memory, improving future tasks against what worked.
Supply chain agentSupply chain · works inside logistics permissions
1,240
Watched today
38
At risk
31
Recovered
7
Escalated
9:14Drafted an expedite for SH-40118on_time_probability fell to 0.64 after a port delay. Sent to a planner.
9:02Rebooked SH-40102 into an earlier dock slotWithin policy, so no approval was needed.
8:47Told the account team about SO-88190Shared the new date and the reason for the customer.
8:30Checked 1,240 open shipmentsMorning run against carrier and weather feeds.
Applications
Applications that empower your team to make the right call
Applications read and write the same ontology the agents use, so a change made in one place is the same change everywhere. A planner sees what the agent did, why it did it, and what's waiting on them.
1,240 open shipments checked this morning against carrier and weather feeds.
Orders at risk
38
Down from 61 last month
Revenue at risk
$1.2M
Across 38 orders
On-time rate
96.1%
Up 2.4 pts
Customer
Promised
On time
Margin
What the agent did
Status
Northwind Foods
SO-88214
Nov 14
0.64
18.4%
Drafted an expedite, $3,900
Needs approval
Alder Medical
SO-88190
Nov 12
0.71
22.1%
Shared a new date with the account team
Agent working
Kestrel Retail
SO-88177
Nov 15
0.82
9.8%
Margin below floor, sent to pricing
Waiting on pricing
Bramble Outdoor
SO-88165
Nov 11
0.93
16.0%
Rebooked an earlier dock slot
Resolved
Cobalt Labs
SO-88150
Nov 10
0.97
24.5%
No action needed
On track
Expedite for SO-88214
Drafted by the supply agent, 9:14 AM
A port delay pushed SH-40118 two days past the promised date. Air on the last leg costs $3,900 and keeps the date. Margin stays at 15.2%. The customer's contract carries a 5% credit for late delivery.
Workflows chain agents, applications, and people into one process. Each step names who owns it and what it changed, so you can explain any run after the fact.
Actions and governance
Every change goes through the controls you set
Actions are how the application layer changes the real world. Each one is defined in the ontology, so the same rules hold for every agent, screen, and workflow that uses it.
Defined in the ontology
Each action states what it changes, who can run it, and when it needs approval.
Approvals
Consequential actions wait for a person with the authority to approve them.
Audit trail
Every action records who or what ran it, why, and the result.
Write-back
Changes land in your systems of record, so teams keep working where they already work.
Bring your own app layer
Use our agents, applications, and workflows, or build your own on the ontology. Everything leverages the same governed actions, permissions, and audit trail.
Built once, reused everywhere
Each new application starts from the object types, expressions, actions, permissions, and integrations already in place. The tenth deployment takes a fraction of the work of the first, and every one makes the ontology richer.