JIRA + MCP Project Mgmt
Turning project ops
into a conversation
How I redesigned a routine task delegation workflow using AI-connected tooling — and built a system that scales from a scrappy advisory team to an enterprise operations floor.
Small teams can't afford process tax
In any lean advisory operation, the people doing the highest-leverage work are the same people managing the administrative machinery that surrounds it. Jira is indispensable for tracking work — but the UI overhead of creating tickets, assigning sprints, linking epics, and managing team assignments is real friction. Multiply that across a week, a team, a quarter, and you've lost hours that should be billable or strategic.
The question I was asking was simple: what if the entire act of assigning work was just a sentence? Not a shortcut inside Jira — a total collapse of the process, from intent to execution, in plain language.
"The most dangerous inefficiency isn't the hour you notice. It's the ten minutes, repeated daily, that you've stopped seeing."
Think first. Build the smallest possible system.
The goal wasn't to find a clever use of AI. The goal was to identify where human time and cognitive load were being spent on work a system should handle — and then remove it surgically. The right question before touching any tool is always: what does a great outcome actually look like, and what's in between me and it?
In this case, the answer was a live, connected workflow that understands the workspace without being told every time, resolves human names to system IDs automatically, handles Jira's hidden field complexity without exposing it, and gets smarter each session — not the same every time.
Every advisor wears multiple hats. Collapsing admin tasks into natural language isn't a convenience — it's a competitive advantage. The hours you save compound.
PMs managing hundreds of tickets can batch-create and reassign via a single prompt. The constraint stops being "how fast can I click" and starts being "how clearly can I think."
The Skill built at the end of this session is the real product. It captures workspace context and edge cases permanently — new team members start fully informed on day one.
Notion, HubSpot, Slack, Linear — same Skill architecture. Pull meeting notes into Jira, route CRM deals into sprints, generate status reports from open tickets automatically.
How it actually happened
This wasn't a demo built in hindsight. It was a live build — discovering the workflow, hitting its edges, fixing them in real time, and packaging the result into something permanently reusable.
Rather than assuming the connection worked, I queried the Atlassian API directly to confirm accessible resources, scopes, and project visibility. Having the right foundation matters more than moving fast — two minutes of verification saves twenty minutes of debugging.
Instead of requiring a ticket number, I designed the interaction so a plain English description was enough. The system searched, surfaced two matches, and asked for disambiguation. The human should never have to know how the system works.
Looked up three team members by full name. Two accounts existed under the same name — a common real-world data quality issue. Rather than pick arbitrarily, the system flagged the ambiguity and confirmed using the verified company email. Good systems notice what humans should resolve.
The tickets showed under the right epic visually, but each ticket's "Epic" field read "None." A real Jira nuance: the parent field nests a task in the hierarchy, but a separate custom field populates the Epic label on the ticket itself. Both are required. Knowing the difference between what looks correct and what is correct is exactly where experience pays off.
Rather than hardcode a sprint ID, I queried the live workspace for the current open sprint. A second nuance emerged: Jira's API requires the sprint ID as a plain integer — not an object. This is exactly the kind of undocumented edge case that costs an hour the first time, and zero time every time after — once it's captured in a Skill.
Every discovery from this session — the Cloud ID, custom field IDs, team member account IDs, sprint methodology, epic mappings, and the specific API edge cases — was encoded into a downloadable Skill file. Install once, and every future conversation starts fully informed. This is the difference between a one-off demo and an operational system that compounds in value over time.
The bottleneck in most operations work isn't ideas or decisions — it's the administrative distance between those decisions and the systems that track them. I build workflows that close that gap, using AI as an integration layer rather than a replacement for thinking. The goal is always the same: the system should do what the system is good at, so the human can do what only the human can do.