BA Star in practice.

Replacing a homegrown ML engine — without disrupting a single user.

An employee engagement platform set out to replace its homegrown AI models with a third party large language model — without any interruption for users.

The transformation involved creating a large amount of backlog items, so the team looked for a way to automate the work. Initially, they integrated their Jira with an agentic AI solution through the Atlassian MCP server to automate the new backlog creation. A large part of it was automated, but at a certain point they noticed some of the key context wasn't being taken into account: most of the items came out similar and too general, and many of the important artefacts still had to be added manually.

With BA Star, the transformation project became a fully structured backlog. Fed with the platform's architecture docs, GitHub codebase, Confluence, Jira history, and migration notes, BA Star generated over 150 grounded items. Everything was prioritised, dependencies between the legacy and new engine were flagged automatically, and a phased migration roadmap fell out of it. The team reviewed and published the whole backlog in three days.

How they use it

• Architecture docs, GitHub codebase, Confluence, Jira history, and migration notes fed in as context
• Over 150 grounded backlog items generated and prioritised
• Dependencies between the legacy and new engine flagged automatically
• Phased migration roadmap created

A complete, prioritised backlog
— from the start.

A new travel platform is building an AI-powered travel agent that designs custom routes for travellers from the criteria they choose and how they behave on the site. With the product still taking shape, the founders needed to turn a broad vision into something a development team could actually build from.

They used BA Star to generate their full product backlog from the ground up — epics, user stories, tasks, and spikes spanning the route-generation engine, the interactive map, and the tiered membership model. BA Star prioritised the backlog, flagged dependencies between features, and laid out a build order, giving the team a clear plan from day one.

How they use it

• Product vision and feature notes turned into structured epics and stories
• Route engine, interactive map, and membership tiers all scoped
• Dependencies between features flagged to set the build order
• A prioritised backlog ready to hand to the development team