Kayce is the AI agent that automates identity verification and customer onboarding. It works on your onboardings like any other analyst: it reviews the documentation, leaves the evidence behind and writes its report. Your team makes the final call.
This article walks through the five sections you fill in to get it running.
Where to configure it
Settings > AI & Agents > Agents configuration.
That screen lists your organization's agents. To create a new one, click "Create".
If you cannot find the section, your organization does not have AI agents enabled yet. Write to Support to turn them on.
The form starts with the first section only. The other four appear once you pick the agent type, because all of them depend on it.
1. AI agent type
Pick the Kayce card and give it a name.
The name is what shows up as the analyst on every onboarding Kayce works on, so make it one your team recognizes at a glance. Two agents cannot share a name.
You can create several Kayce agents with different setups. One for individuals and another for legal entities, each with its own instructions and triggers.
If you change the agent type after filling things in, you lose the instructions, the tools and the triggers you loaded. The platform asks you to confirm first.
2. Roles and groups
Roles define what Kayce can do. Groups define which resources it works on.
Both assignments are required: at least one role and at least one group. An agent with no roles has no permission to act. With no groups, it has no onboardings to work on.
Give it the same roles you would give an analyst doing this job. Kayce never does more than its roles allow. Neither the instructions nor the tools widen that limit.
3. Instructions
Instructions give Kayce context on how your organization works. They apply to this agent only. The section is optional: leave it blank and Kayce runs with its default behavior.
You have two paths:
Instructions library. The dropdown on the top right offers ready-made instructions for Kayce, in your organization's language and country. Load one and edit it.
Custom. Pick this option and write your own text from scratch.
The editor supports basic formatting and accepts up to 50,000 characters.
Loading an instruction from the library replaces whatever you already wrote, and you cannot undo it. The platform warns you first.
Be specific about what you expect from Kayce. Set the tone and the format of the report. Add examples when they help clarify.
A few examples for Kayce:
Check that the declared address matches the one on the proof of address. If it does not match, note it in the report and request additional documentation.
For legal entities, review the chain of ultimate beneficial owners. Flag any shareholder above 25% ownership.
If the customer operates in a high-risk jurisdiction, review every document more strictly and point out any inconsistency in the dates.
Two things to keep in mind:
These instructions add to the AI Instructions you set on each onboarding template. Both combine when Kayce works.
Do not write sensitive information. Any user with access to the settings can read them.
4. Tools
Tools extend what Kayce can do on the onboardings you assigned to it. The section is optional.
Each tool is a switch. Some also ask you to choose what you enable:
Write report. Kayce writes and saves the onboarding report. Comes on.
Send requirements. Kayce requests documentation from the customer using the requirement templates you select. Comes off.
Manual triggers. Kayce runs the manual triggers you select during its work. Comes off.
If you turn on a tool that asks for options, you have to pick at least one. To enable none, leave the tool off.
5. Triggers
Triggers define when Kayce runs automatically. This section is required: an agent with no triggers would never run, so the platform does not let you save it.
Click "Add trigger" and pick the event and the onboarding template:
When an onboarding is created. Kayce starts as soon as the file opens.
When an analyst is assigned. Kayce starts once it is assigned as the onboarding's analyst.
When the stage changes. It also asks for the stage. Only the active stages of the template you picked are offered.
When all requirements are completed. Kayce starts once the customer finished sending the documentation. This event takes several onboarding templates at once.
You can load several triggers for the same agent. Each one is listed with its event and its template.
Saving
Click "Create". If you try to leave with pending changes, the platform asks you to confirm before discarding them.
Turning agents on and off, and editing them
The list shows each agent's name, roles, groups and creation date.
Edit: the settings icon opens the same screen you used to create it.
Turn off: the switch on the row stops the agent. It stops running and the row dims. You turn it back on with the same switch.
On an agent that already exists, triggers save instantly: you add, edit, turn off or delete one, and the change is applied. The name, the roles, the groups, the instructions and the tools save with the "Save" button.
If you deactivate an onboarding stage that has a trigger pointing at it, the platform warns you first. It deletes that trigger, because a deactivated stage never changes and the agent would never run.
Worth knowing
Kayce shows up as the analyst on every onboarding it works on, with the same traceability as any team member.
Permissions win. If an instruction asks for something its roles do not cover, Kayce does not do it.
Kayce gathers evidence and recommends. Approving the file stays with your compliance team.