Once a Jira site grows past a few hundred people, I keep seeing the same pattern. The configuration is rarely the hard part. The volume of small requests is.
Every week someone needs a custom field, a new status, access to a project, a different project role, an app installed, or a new project. Each one is five minutes of admin time, but together they eat a large part of the admin's week. And because there is no single place to ask, the requests arrive through chat, email, ticket comments and hallway conversations.
What I've seen go wrong
- The obvious fix is "hire or appoint more admins." That helps with load, but it brings its own costs. Every admin is a fully privileged account to protect, train and offboard.
- Ownership overlaps. Requesters message whichever admin they know, so two admins answer the same ask, or each assumes the other will.
- Duplicate configuration. Two admins create two similar fields or statuses because neither knew about the other's change. The "Worst Jira Admin" thread here describes a site with 132 admins and exactly this result.
- Requests get lost. No one can find who approved what, or why, a month later.
- Jira automation only goes so far. It has plan-based monthly limits, and as far as I can tell native automation can't add a user to a project role or change group membership, so the admin still does the last step by hand.
- Privilege creep. Access granted for one incident is rarely removed, and auditors (SOC 2, ISO 27001) ask for evidence of least privilege.
There's no official admin-to-user ratio. Community answers range from "2 is the floor" to "3 to 10, depending on complexity", so the better question may be how much admin work can be taken off the admins, not how many admins to have.
An approach we've been trying
Instead of adding admins, we route every request through one intake and split approving from executing:
- One request form for common admin tasks (fields, screens, roles, apps, projects, statuses, issue types, priorities).
- Low-risk requests can be approved in one click, and the app applies the change and verifies it. Anything high-risk, or anything that triggers a warning, always goes to a human.
- Specific people can be marked as trusted for specific low-risk request types, so they skip the queue, and that trust can be removed at any time.
- Every action is written to a tamper-evident audit log you can export.
The result is that admins decide less often, not more often, and the person asking always knows who owns the answer.
What it doesn't do: it doesn't give anyone more access than Jira already allows. Changes still run with the permissions of the administrator who executes them (trusted requests run under the app's own identity), and some things, like adding a status to a workflow, still need a manual step.
Questions for you
- How many admins do you have, and how many users do they serve?
- Where do admin requests arrive today: Jira tickets, Slack/Teams, email?
- Do you review who has admin rights, and how often?
- What request types eat the most admin time?
Disclosure: I'm the developer of FlowOps for Jira (the approach above), so I'm biased. Happy to share more details if useful, but I'm mostly curious how others solve this.