The leaky roof problem

A warehouse with a leaking roof. You hire people to hold buckets. The floor stays dry, the problem is "solved." The leaks slowly multiply as the business grows and at some point you just accept that this is part and parcel of success. Fast forward a few years and now you can deploy AI robots to hold the buckets, which means you need fewer people. Incredible. Why would you ever fix the roof?

The honest answer is that fixing the roof is harder, slower, and requires convincing people who aren't standing in the warehouse to care about a problem they can't see. That's the real challenge here. Not the technology, not the data, not even the reporting. It's the organisational will to treat support contact as a signal worth acting on rather than a cost to be managed.

I'm obviously simplifying, and I have a lot of respect for the work support teams do when they're grinding through persistent, high-volume issues day after day. But it's a point worth making. Why do so many companies look at their contact volume and not see it as a signal? An opportunity to actually improve the thing that's causing the friction in the first place?

TLDR

• AI deflection is useful, but it can quietly paper over the real problems driving contact
• Most customers don't want to contact you. They just want the thing to work
• The data you need to make the case already exists in your helpdesk
• CS leaders need to connect contact volume to cost and story before walking into a cross-functional meeting
• Success looks like a company that runs internal campaigns to fix the things support has been flagging for years

The band-aid problem

With AI improving as fast as it is, the temptation to reach for a deflection tool is completely understandable. It opens up a genuine self-service channel and it can take real pressure off your team. I'm not arguing against it.

But it can act as a band-aid to wider problems that deserve an actual fix. Those problems tend to look like gaps in the product that customers keep falling into, business process decisions that create unnecessary manual work, self-service that doesn't actually let people help themselves, or marketing that sets an expectation the product doesn't quite deliver on. None of that lives in the support queue. But it all shows up there, reliably, at volume.

There are also contact types that are genuinely outside your control. Warranty returns for faulty hardware, shipping delays through a 3PL that operates on its own schedule. Some things you can't fix. But they still come at a real cost, and even those are worth quantifying.

Here's something worth sitting with. Most customers don't actually want to contact you. Maybe that's a generational thing, the preference for figuring it out yourself rather than picking up the phone. But broadly speaking, a customer reaching out is a customer who couldn't get what they needed any other way. That's not a support problem. That's a product, process, or communication problem that landed in your queue. The question is whether anyone upstream is paying attention.

The trickier problem is convincing leadership to care. Any product manager worth their salary will tell you that chasing down support contact drivers isn't exactly top of the roadmap. And they're not entirely wrong, there are always competing priorities. But I do think there's room for both, and the burden usually falls on support leaders to build the case rather than waiting for someone else to notice.

Where to start as a support leader

The goal here isn't to walk into a cross-functional meeting and declare that every contact type must be eliminated. That's not a realistic ask and it won't land. The goal is to show up with one clear problem, a confident story, and enough data that someone in the room feels compelled to do something about it.

Here's how I'd approach it.

First, make sure your helpdesk is actually capturing what you need. This might mean custom fields that agents or customers fill in, or leaning on AI to automatically tag and categorise. If you're in Zendesk, Copilot's Auto Assist and the Custom Intents inside the Intent feature are both worth looking at here. Before you build any reporting off the back of this, do a QA pass on the categorised tickets. Whether a human or an AI labelled them, you need to be confident the data is accurate before you put it in front of a leadership team.

Once your data is clean and you have a cost figure to work with, build your reporting in Zendesk Explore. I'm a fan of time-based views that show both the current volume or proportion of a problem and how it's looked over the previous few periods. It's a simple way of showing that something isn't going away on its own.

Then pick one issue. The highest volume item that has a clear owner somewhere else in the business, product, marketing, operations, it doesn't matter which. Find a few verbatim ticket comments or run a short user interview to put a human story next to the numbers. Bring it to a cross-functional meeting framed as an observation, not a complaint.

The best companies I've worked with treat this as an ongoing practice, not a one-off exercise. They run internal campaigns around persistent issues, cross-functional projects with a clear brief to fix the things that are creating the most friction for customers. It sounds ambitious but it starts small. One problem, one team, one win. That win is what gets you the next conversation.

The bigger picture

Support has spent a long time justifying its existence through cost metrics. Handling time, cost per contact, deflection rate. Those numbers matter and I'm not dismissing them. But they're a floor, not a ceiling.

The teams that shift this conversation are the ones who can demonstrate that support isn't just absorbing problems, it's identifying them. That's a different kind of value, and in my experience it tends to get leadership attention in a way that a faster average handling time never quite manages.

The roof is still leaking. You can keep hiring people to hold buckets, or you can be the person who finally makes the case to fix it. One of those conversations is harder. It's also the more interesting one.