I don’t think of project backlogs as to-do lists, but as a can’t do list. A list of bugs, tasks, and features the company can’t do without a development team. This bottleneck has existed for as long as software has. So much so that an entire project management industry has been built around this very issue.
Every project backlog accumulates quickly with internal teams needing reports, partners needing data exported, customers demanding new features. Each request is small, but it adds more to the list that never shrinks because the requests keep coming faster than your team can build custom UIs around them.
There’s a better way. Today we can deploy MCP servers and CLIs built for AI agents. When you expose your product through these interfaces, you’re not just making it easy to create tools around your product, you’re making it possible for anyone with access to a chat prompt to do their work.
This may seem obvious to those who have been working with AI for a while, but I’m still seeing companies without any of these tools deployed. Once these tools are in place, teams start finding creative solutions on their own, wiring AI agents into them, loading up skills, and solving problems you never would have anticipated or prioritized.
Every feature you expose through these interfaces is a feature that never needs a UI built around it. The gap between what your product can do and what requires engineering keeps shrinking. We wrote about AI code quality recently and this has been a very effective way to further prevent technical debt.
Are there tickets in your backlog right now that can be handled by an agent?