AI-assisted issue linking inside the ID.prove Service Portal
Support tickets used to be matched against GitHub issues by hand. Now an AI assistant finds the related issues and the product owner links them in one click.
Client: Rausoft GmbH
Product: ID.prove SaaS Service Portal
In production since August 2026
Anthropic Claude API · GitHub
The problem
Rausoft’s support team records client-reported bugs and feature requests in the ID.prove Service Portal. Before a ticket becomes development work, someone has to check whether a GitHub issue for it already exists. That meant manual searching across hundreds of issues, in two languages, for every ticket. Duplicates slipped through and screening was slow.
The solution
We embedded an AI matching step directly into the ticket screening view of the Service Portal, so the check happens where the decision is made rather than in a separate tool.
How it works
1
Ticket arrives. The portal stores the original German report plus an English translation. Both, together with attachment metadata, are sent to the matching model.
2
AI matches. Using the GitHub API, the portal pulls the current open and recently closed issues from the ID.prove repository. The model compares the ticket against this corpus and returns the closest candidates, each with a match score and a one-line reason for the match.
3
Reviewer decides. The candidates appear on the ticket. The product owner opens any of them on GitHub, links the right one, or falls back to entering an issue number manually.

Ranked candidates with match score and rationale, shown directly on the ticket.
Human in the loop
The AI suggests, a person confirms. Scores and reasons are shown so a 54% match is judged differently from a 92% one, and every analysis is written to the ticket history with a timestamp. The accept action itself warns when a near-identical open issue exists, so duplicates are caught at the exact moment they would otherwise be created.

Link and accept in one step.

Duplicate warning at the accept step.
Closing the loop with GitHub
Linking is a single confirmation. Through the GitHub API the portal attaches the issue reference to the ticket, posts a comment on the GitHub issue pointing back to the Service Portal ticket, and marks the ticket as accepted. Developers see the customer context on the issue they are already working on, and no duplicate issue is created.

Ticket after linking, carrying the GitHub issue reference.
Outcome
Ticket triage dropped from a manual search per ticket to a one-click confirmation, and duplicate development issues are caught before they are created. The capability was built in about one month, going from kick-off in July 2026 to production in August 2026, and has been in daily use by Rausoft’s support and product team since. Because it sits on the Anthropic Claude API and the GitHub API rather than a custom model, matching quality improves with each model generation and needs no retraining on Rausoft’s side.