Ticket Routing
Send each incoming message to the right queue using answers you write. No labeled ticket history.
Send each incoming message to the right queue using answers you write. No labeled ticket history.
Business problem
Support and success teams read every new ticket to decide who owns it. The rules live in people's heads, and a new hire routes the same message three different ways.
Heimdall solution
Create a Resolve Choice question: which queue should handle this? Name the queues and when to pick each one. Billing covers charges and refunds. Support covers something that is broken. Sales covers a demo or a plan change. Send the ticket text. Resolve returns the queue, a confidence, and a probability for every queue.
Business impact
The queue is set before a person opens the ticket. Unsure messages stay visible, with the full probability split, so a lead can still override the pick.
How it works
Write the queues
Name each destination and a short description of the messages that belong there. Two queues is enough to start. You can add more later, up to 255.
Send the ticket text
Call the Resolve API with the subject and body. No historical labels and no Lake dataset.
Route, then review the unsure ones
If you set a confidence cutoff, send anything below it to a person. The API still returns the suggested queue and the probabilities.
See Heimdall in action
Tell us how tickets get to the wrong queue today
Key benefits
No training set
The answers are the ones you write
Stable queues
The same question is applied to every new ticket
Visible probabilities
See how close the other queues were
Optional cutoff
Hold low-confidence tickets for a person
One API
The help desk, form, or inbox calls the same endpoint
Fixed type
Choice stays Choice, so the response shape does not change
Related use cases
More scenarios on the use cases hub.
Urgency Scoring
Place a ticket on a scale you write, from can-wait to needs-someone-now.
Multi-Source Consolidation
Unify CSV exports, Postgres tables, and warehouse snapshots into one bronze catalog — no ETL scripts to maintain.
Unstructured Archive Ingest
Drop zip archives of PDFs, images, and documents into bronze storage for downstream structuring and ML pipelines.