AI for Infrequent Business Needs

In my experience, stakeholders who frequently express and request their needs tend to get better at framing them in way the organization can prioritize. Without abundant delivery capacity the threshold for infrequent stakeholders increases.

It is hard to evaluate and compare the potential business impact of two needs where one has a clear business case, feasibility study and dependencies mapped and the other is more of a napkin sketch.

If the work to find a constructive way to frame the need and validate technical feasibility isn't done it will most likely be left to rot in an ideas pile. This doesn't mean the need was of lesser value than the better framed one, it never got to the state where such an evaluation could be done.

If we could take loose ideas, suggestions or conversations about things we'd like to change in our services and have near unlimited capacity to frame them as state of the art Jira-tickets (verified by the requester) more needs would get past that initial hurdle.

Imagine we then had just as much capacity to work out what in which team's code would need to change, implement a draft pull request containing the Jira-ticket reference and a summary of the changes made.

Enter LLMs!

We can utilize AI Agents to create better change requests with clear motivation and acceptance criterias. Then we can have them plan and build an initial draft solution.

The teams maintaining the services changed quickly gets to the stage where they can evaluate if value created justifies the cost of maintaining the proposed change, or what it would take to make it so.

This will let more ideas from more stakeholders get the chance to fail or succeseed early at a lower cost and a lower lead time.

I believe this would help many organizations to quickly move forward with a greater focus on business impact.