Built on ourselves · Internal
An agentic loop from inbox to code.
LeapOps turns customer email into tickets a person approves, links follow-ups instead of duplicating them, and turns bugs into engineering stories that cite real lines of code.
Email → Approved → Ticket
Enrichment proposes. Approve writes.
The model never creates a record on its own. Approve writes the ticket. Reject leaves nothing. Edit creates a new revision.
Generation is cheap. Verification is the constraint. AI can draft all of it in seconds; the hard part is knowing what to trust. So LeapOps is built around review, not output.
Internal · Support and delivery operations
In daily use
AI drafts. A human approves. Nothing else writes.
GiantLeap Systems · our own support and delivery work
The problem
Support and delivery work starts as email. Resolutions live in threads people archive or leave. Urgency and category depend on who read the mail. The same bug returns as a new ticket, and engineering gets a blank issue instead of a story it can act on.
What we built
An agentic loop from inbox to code. Email, pasted text or a saved message is classified for urgency, product, fault origin and category, and staged as a proposal. Before anything is created it checks thread continuity, near-duplicate text and semantic similarity to past tickets. A person approves, rejects or edits. Bugs become stories with evidence split into email, code and inference, linked to real lines from an index refreshed on every push. Plain-language questions over the ticket history are answered through defined tools, not free-form queries.
Match before creating. A follow-up updates the open ticket instead of opening a twin.
Memory that lasts. Resolutions and their history stay in the system, not in someone’s archived inbox.
Stories engineering can act on. Not a blank issue: a structured story with its evidence and the code it points to.
Outcomes, on our own work
Reviewed tickets
Every email becomes a proposal. Only a human approval creates a record.
One ticket per issue
Follow-ups attach to the open ticket, with the reason shown to the reviewer.
Cites the code
A ready story with evidence and links to the exact lines in the repositories.
Why it holds
Honest readiness is a feature. The ticket loop is in daily use; code-grounded stories are live in the backend with the interface rolling out; GitHub sync is on the roadmap. No throughput figures claimed.
Stack
Email classification · Thread and semantic matching · Local embeddings · Code index on every push · Defined query tools