Ticket backlog: an hour, not a day

Request: „Build the backlog for the board from the concept and planning documents — epics, features, mapping.“

1. The AI reads the source documents and the Business Brain (target picture, existing epics, board conventions)
2. It derives epics and features and slices them according to the team's conventions
3. It checks the board for duplicates, assigns parent tickets and creates the tickets through the API
4. The same pattern for one goal across around 20 products: an epic draft per product, duplicate protection and a dry run show what would be created — after GO the AI creates the epics one by one, reads each back and stops at the first error
5. Single tickets on request: read the template live from the target system, pre-fill fields from the description, ask only for missing mandatory information, show similar tickets, create after GO and read back

A backlog grows out of workshops and solo work: writing epics, slicing features, creating every ticket by hand — about a working day per backlog, in any organisation.

Scope, priority and the epic slicing are confirmed by the team; a person checks the preview, gives the GO and decides how a ticket differs from similar ones. The AI delivers the proposal and the legwork.

Any organisation with a ticket system and planning documents. The effort is in the transcription, and transcription can be delegated.

Building a backlog is transcription: read the concept, write epics, slice features, create tickets, link them. By hand, that takes around a working day. Agentically, the AI reads the documents and the Business Brain, derives epics and features according to the team's conventions, checks the board for duplicates and creates the tickets through the API. After about an hour the backlog is ready for sign-off — a good four fifths less time; the first run produced around 60 linked tickets with monthly, quarterly and backlog views. The same pattern carries the routine afterwards: one goal becomes epics for around 20 products, with dry run, GO and read-back — about an hour instead of filling the same fields 20 times; and a single ticket grows from a description in plain words, with similar tickets flagged before anything is created. The team still decides scope and priority. The figure is my own estimate, not a measurement series, but it matches the pattern of the other cases.