A customer asks an AI agent to order groceries, buy a phone from a marketplace seller, or book a vacation rental. In each case, the agent can have permission to act while the proposed purchase still carries fraud risk.
Take a grocery order. The agent is verified, and the marketplace has confirmed that it has permission to make the purchase. But the order includes a delivery address the customer has never used before.
Does that make the order suspicious?
The customer may regularly spend time at that location, even if they've never ordered groceries there. Or the unfamiliar destination could be part of a broader pattern of suspicious activity.
Verifying the AI agent and confirming the customer's authorization leaves the marketplace with another important question: Does what the agent is trying to do make sense for this customer?
As customers delegate more purchases and bookings to AI agents, marketplaces need ways to assess the risk of those actions before completing them, without requiring customers to take over every transaction.
That’s what we’re bringing with Intent Intelligence.
Incognia Intent Intelligence helps marketplaces assess whether digital actions proposed by AI agents make sense for the customers they represent, including requests from verified agents acting with customer authorization.
It supports purchases and bookings requested by a customer’s AI agent or a marketplace’s own AI assistant. Depending on the action, the assessment can consider the customer’s connection to a delivery destination, a seller’s links to known fraud, or outcomes from previous bookings.
When an agent requests a purchase, Intent Intelligence evaluates the proposed action against the customer's established activity, the events leading up to checkout, and available risk intelligence about the seller and destination.
It also draws on Incognia's physical-world intelligence to assess whether a delivery or pickup location is consistent with places the customer has an established connection to.
This helps marketplaces prevent fraudulent purchases before orders are accepted, while allowing lower-risk requests to proceed without asking customers to take over checkout.
AI agents may operate from cloud infrastructure, separate from the customer's own device and physical location.
Incognia uses intelligence from the customer's recognized devices and established location history to bring physical-world context into the assessment of an agent's proposed action.
A key part of this approach is Incognia's Physical Social Graph, which combines apartment-level location intelligence with persistent device recognition to uncover evidence indicative of real-world connections.
For marketplaces, these connections can help establish whether a requested delivery destination matches a customer's Trusted Locations, even if they've never placed an order to that address before.
Consider the grocery order from our earlier example.
The requested delivery address might be new to the customer's order history, but Incognia's location intelligence may recognize it as a place where the customer regularly spends time.
The marketplace doesn't need to know the nature of the location. Repeated real-world activity can provide evidence that the customer has an established connection to it.
Incognia's apartment-level precision also helps distinguish activity within a specific apartment from unrelated activity elsewhere in the same building. Combined with historical patterns and persistent device intelligence, this provides more precise evidence about the customer's connection to a destination.
When an AI agent requests a purchase, Intent Intelligence can help answer questions such as:
Does the customer have an established connection to the delivery destination? A new delivery address may already be a Trusted Location based on the customer's historical physical activity.
What happened before checkout? Recent account recovery, device changes, or the addition of a new delivery address may provide important context when assessing the request.
Is there relevant risk associated with the seller or destination? Available device relationships and connections to known fraudulent activity may reveal concerns that agent verification alone cannot address.
By combining these signals, Intent Intelligence helps marketplaces evaluate the proposed purchase using the customer's established behavior and available risk evidence.
A customer’s verified AI agent requests a grocery delivery.
Before accepting the order, the marketplace can use Intent Intelligence to assess the proposed purchase in real time.
1. The agent requests the purchase
The marketplace calls Incognia's API to evaluate the proposed order and available context before accepting it.
2. Intent Intelligence assesses the request
Incognia evaluates the customer's connection to the delivery destination, established activity, and relevant events leading up to checkout. It also considers available seller risk intelligence.
3. The marketplace uses the assessment
Incognia returns risk assessments and supporting intelligence.
The marketplace applies its existing risk and authorization policies to determine whether to proceed with the purchase, request additional customer confirmation, or investigate the activity.
4. The marketplace brings the customer into the process if confirmation is needed
If additional approval is required, the marketplace can ask the customer to review and confirm the specific purchase through its own channels, such as its app.
Incognia's human-in-the-loop verification supports this process by assessing the customer's device, account relationship, device integrity, location context, and network risk alongside their approval.
The marketplace then considers the customer's confirmation alongside the transaction's risk assessment before deciding whether to accept the order.
The value of Intent Intelligence becomes clearer when considering how different signals can affect a marketplace's assessment of an authorized purchase.
An AI agent requests a grocery delivery to an address the customer hasn't used before.
However, Incognia's location intelligence identifies repeated physical activity at that specific location, indicating an established connection to the delivery destination.
The order also fits the customer's usual purchasing activity.
Combined with other risk checks, these findings support allowing the purchase to proceed without unnecessarily interrupting the customer.
Now consider an agent requesting delivery to a newly added address with no established connection in the customer's available location history.
The request follows a recent account recovery and differs significantly from the customer's usual ordering activity.
Although the agent is verified and authorized, the combination of these findings gives the marketplace reason to investigate before accepting the order.
An unfamiliar destination alone doesn't establish fraud. Customers regularly order gifts, travel, or have items delivered to new locations. Likewise, a Trusted Location doesn't guarantee that a purchase is legitimate.
The assessment depends on the combined evidence.
Intent Intelligence can also help marketplaces evaluate risk associated with the seller involved in a purchase.
For example, a customer authorizes an AI agent to purchase a phone from a marketplace seller.
Incognia's available risk intelligence connects the seller's repeated activity to a specific apartment associated with confirmed fraud.
That finding gives the marketplace a reason to investigate further before accepting the purchase, even though the customer's agent is verified and authorized.
Incognia's apartment-level location precision helps distinguish the seller's connection to that location from unrelated activity elsewhere in the building.
The marketplace can evaluate this evidence alongside other available risk signals to determine whether additional action is warranted.
Intent Intelligence can also draw on the outcomes of completed transactions to inform future assessments.
Consider a customer whose AI agent books a vacation rental at a property they've never visited before.
The booking is unfamiliar, but the platform evaluates the available evidence and determines that the transaction can proceed.
During the booked dates, the customer's recognized device is observed at the property. The platform later reports that the stay was completed without a dispute.
Together, these observations provide additional evidence that the booking resulted in an actual stay.
When the customer's agent requests another vacation rental, Intent Intelligence can consider the history of that completed booking and its outcome.
The new destination may still be unfamiliar, but the customer's established record of successfully completed agent-assisted bookings provides additional context for evaluating the request.
This can help marketplaces assess future activity with greater confidence and potentially reduce unnecessary interruptions for legitimate customers.
The usefulness of these assessments depends on the activity and outcome information available to Incognia and the marketplace.
One of the main benefits of AI agents is their ability to complete purchases and bookings without requiring customers to manage every step.
Requiring customers to manually review every order reduces that convenience.
Intent Intelligence helps marketplaces determine which proposed actions warrant additional scrutiny based on the customer's established activity, physical-world connections, and available seller risk evidence.
A new delivery address might have a strong connection to the customer's physical-world history, supporting a purchase that would otherwise appear unfamiliar. Meanwhile, suspicious account changes or seller activity associated with known fraud can give the marketplace reason to investigate.
As marketplaces gain more information about completed transactions, that history can also contribute to future assessments.
The marketplace retains responsibility for enforcing permissions, obtaining customer approval when necessary, and deciding whether a purchase or booking should proceed.
As customers increasingly delegate shopping and booking tasks to AI agents, marketplaces need ways to assess whether those actions make sense for the people the agents represent.
Learn more about how Incognia Intent Intelligence helps marketplaces assess AI-driven purchases and bookings before they're completed →