01
Many workflows start the same way
Someone submits information. Then someone else has to figure out what should happen next. It might be:
- A service request
- A warranty claim
- A quote request
- An internal IT request
- A customer inquiry
- An application
- A procurement request
- A complaint
The first step is often classification.
02
Turn previous decisions into training examples
Imagine a manufacturing company receiving quote requests. Previous requests might look like this:
| Request | Assigned Team |
|---|---|
| Need replacement hydraulic seals | Parts |
| Quote for 14 custom assemblies | Custom Engineering |
| Looking for distributor pricing | Channel Sales |
Those previous routing decisions can become the training data for a StayCharted AMT model.
03
Let your organization define the destinations
StayCharted AMT does not need to know what your departments mean. It learns from examples. Your labels might be: North America Sales Technical Support Engineering Review Returns Accounts Receivable Those destinations are specific to your organization. That makes this a natural use case for customer-trained AI.
04
Use the prediction inside an application
Once trained, the model can sit behind a form, portal, ticketing system or internal application.
- Customer / Employee
- Request Form
- Business Application
- StayCharted AMT API
- Predicted Destination
- Workflow / Queue / Team
Your application uses the predicted category to route the request.
05
Automate the obvious. Review the uncertain.
A practical implementation does not have to automate every request. Your application can use prediction confidence to decide whether to:
- Route automatically
- Recommend a destination
- Send the request for manual review
This gives operations teams a controlled way to introduce AI without redesigning the entire process.