Four people use your helpdesk. Only one of them is the customer.

A helpdesk gets configured once, by one person, usually under time pressure. Whoever that person was, the instance ends up shaped around their job. If an admin built it, it is powerful and unusable. If a manager built it, the reports are lovely and the agents are fighting the interface.

There are four audiences in there, and they want incompatible things.

The customer wants to not contact you

Nobody enjoys opening a ticket. The best outcome is that they find the answer and leave.

That means a help centre organised by what people actually ask rather than by your internal team structure, a contact form short enough to finish, and an AI agent that either answers properly or hands over without making them repeat themselves. A bot that stalls is worse than no bot, because it costs the customer time and then costs you the ticket anyway.

The agent wants fewer tabs

Agent experience is mostly about how much context arrives with the ticket. Every piece of information an agent has to go and find is time on the clock and a chance to get it wrong.

• Order, shipping and refund state in the sidebar, not in a second system.
• Macros that are current and do not contradict each other. Six good ones beat forty stale ones.
• Routing that means the ticket in front of them is one they can actually resolve.
• An AI copilot suggesting from approved content rather than improvising.

The test is simple. Watch an agent handle five tickets and count the tab switches. That number is your backlog.

The admin wants to change things without breaking them

Admins inherit other people's decisions. What they need is not more power, it is legibility: naming conventions that survive a handover, triggers that say what they do, and a written record of why a rule exists.

The failure mode here is accumulation. Cloning a macro is faster than reading forty, so people clone. Two years later nobody can safely delete anything because nobody knows what depends on it.

This is also where internal and customer-facing work diverge hardest. If IT and HR share an instance, role and visibility rules are not a preference, they are the whole design. HR tickets that IT can read is a problem you only get to solve once.

The manager wants a number they can defend

Most reporting we open technically works and is not trusted, which is the same as broken. Usually because the fields underneath it were optional, so half the tickets have no topic and the report measures volume without meaning.

Reporting is configured last, on purpose. Fix the fields, then the report has something to count. Building the dashboard first just produces a confident chart of nothing.

Where the conflicts actually land

A required field gives the manager clean data and costs the agent time on every ticket. Aggressive AI containment gives the exec a deflection number and gives the customer a worse experience if escalation is loose. Simplifying the agent interface takes tools away from the admin.

These are trade-offs, not bugs, and someone has to decide them out loud. When nobody does, the default is that whoever complains most gets their way, which is how instances end up incoherent.

What we do about it

Configure per role rather than per feature, and write down which trade-offs were chosen and why. The price includes writing it down. You own the configuration when we leave, documented, and we do not need a permanent retainer to keep it running.

If you want to know which of the four your instance is currently optimised for, the $500 audit will tell you in one to two weeks, in writing, ranked. It is credited against any project booked within 30 days.