Q: How does the agent know what REST API to use?
A: The agent is provided with information in the form of tools — specifically, a custom API connector is pre-created and associated with the claims REST API. The base URL is configured, and the agent is given that connector as a tool, so it knows to use that specific claims REST API when executing the workflow.
Q: What are the best practices for preparing our SPICE dataset in order to ensure flows automate accurately and reliably?
A: There is currently no native integration between SPICE datasets and Quick Automate. You could still connect through a REST API, but native integration is still being worked on.
Q: What do the validation rules look like when returned from the service?
A: The validation rules are straightforward — they are returned as a JSON object. For example, a rule might say “check whether the policy status is valid or invalid.” The rules are defined in an MCP server, and you can plug in any set of rules you want. The MCP server can return a JSON format or any other format you define.
Q: Is there a way to get a copy of the life insurance automation that was shown in the demo?
A: Yes. If you have an AWS account team, they can run the workshop for you. A self-service option is also available: you can download a CloudFormation template, deploy the infrastructure in your own account, and run the exact workflow shown in the demo.
Q: Team members want to leverage the automation function but are non-technical and not using S3 or databases. What functions should they feel confident using, and should they pause on trying automation?
A: Non-technical users do not need to be highly technical to use Quick Automate. Recommended options include: (1) having a technical admin create connectors for S3 or other systems and share them — users can then leverage those connectors without needing to set them up; (2) a soon-to-launch feature will allow users to simply upload a file (e.g., an Excel file) directly into an automation without needing a connector. Many web automations can also be done without connecting to a database at all — the UI agent can navigate websites, extract data, summarize, and share results. Users should not pause; they can get started with the tools available today.
Q: Is there a way to limit information available through an integration based on user identity?
A: Yes, through automation groups. You can create connectors scoped to specific teams (e.g., finance or HR) and link those connectors only to the relevant automation group. Builders in that group will only have access to the connectors assigned to them, limiting what data and resources they can access. Additionally, a “user identity run” feature is coming soon — this will allow automations to run using the credentials of the person executing the automation (three-legged auth), automatically limiting data access to what that individual user is authorized to see.