“Which invoice is still outstanding?”
The person asking needs the position for their own contract. General billing advice is little help. An earlier conversation may already be out of date after a payment.
We connected FRED to Stripe so the AI agent can retrieve relevant invoice information. AI interprets the question and uses purpose-built tools. The answer is grounded in the connected system.
Establish the relationship before retrieving data
The implemented workflow checks the customer and contract relationship before returning invoice data. A similar-sounding name is not enough.
Illustrative example: someone asks about an invoice for a particular contract. The agent identifies the relevant context and retrieves the associated documents. This is a fictional example of the implemented workflow.

A list must reveal its gaps
| Result | What the answer should do |
|---|---|
| An associated invoice is found | Explain the document and available status |
| Payment is still processing | Describe the intermediate state |
| Association or coverage is incomplete | State the gap and involve people where needed |
An empty or incomplete response must not become “You have no outstanding invoices.” An invoice marked paid does not, by itself, establish how funds actually arrived.

Payment data and CRM play different roles
The Stripe connection answers invoice questions. The HubSpot integration provides another part of the context: contacts, communication and cases. Together they can help FRED prepare a more relevant response.
This integration is built around particular tasks. It does not mean that every field is fully synchronised between Stripe and HubSpot.
What we developed with AI
The work covers tool integration, matching information to the case and handling missing results. AI contributed to development; FRED uses the tools during conversations.
Sources.
- SolvraONE: own development and technical source reviewOwn source · As of
- Stripe: invoice APIVendor / primary · As of

