Auto scaling for contact centres

The maths on contact centres is brutal.

Losing half of a 4-person contact centre doesn't double your wait times. It makes them infinite. Not "much longer". Infinite.

That sounds like an exaggeration, so this post walks through the actual numbers. They explain most of the queue meltdowns you've ever sat in, and they point at a fix.

Here's the pattern in about a minute, and the rest of the post is the working behind it:

The maths

Contact centres run on the same queueing maths as supermarket checkouts and web servers. It's over a century old and it's called Erlang C.

Take a small contact centre that gets 30 calls an hour, averaging 4 minutes per call. That's 120 minutes of talk time arriving every hour: two people's worth of work.

To hit the industry standard of 80% of calls answered within 20 seconds, you need 4 people on the phones. Run the numbers with 4 and the average wait is about 21 seconds, with 85% of calls answered inside 20. A healthy queue.

Now one person calls in sick.

With 3 people the average wait jumps to 107 seconds. One person out, five times the wait.

Lose a second person and it stops being "slower". With 2 people you have two hours of work arriving every hour and exactly two hours of capacity. The queue never drains. Every blip pushes it longer and nothing ever pulls it back. The average wait time is no longer a number.

At that point the only release valve is customers giving up. And they give up after a wait so long it poisons their relationship with your brand, right before they vent about it to whoever finally answers.

The thing to take from this: wait times are not linear in staffing. A queue near full utilisation doesn't degrade, it explodes.

Why you can't staff your way out

If the fix were "hire more people", contact centres would have done it. Two things make that a trap.

The first is that customer support has some of the highest absenteeism of any office role, up to 3 times the norm (5 to 10% against roughly 3%). With the pressure the maths above puts on the people answering, it's no wonder. So the understaffed day isn't the rare case you plan around. It's a routine event.

The second is that demand spikes don't announce themselves. An outage, a recall, a billing error, a storm. Any of them can double your call volume in an hour. Staff for the worst day and you're paying people to sit idle every normal day. Staff for the normal day and the worst day is the cliff above.

Contact centre managers have been stuck between those two for decades. If you've ever cursed a hold queue, spare a thought for the person running it.

Software solved this years ago

In software this exact problem has a boring, settled answer: auto scaling. When a server fleet hits 80% capacity, another server spins up. Load drops, it spins back down. Nobody staffs a website for Black Friday by buying servers that sit dark the rest of the year.

Contact centre managers have never had that option, because their capacity is people. You can't spin up a trained human in 40 seconds.

Now you can spin up the next best thing.

Overflow: auto scaling for contact centres

VerticalAI plugs into your existing phone system and watches your queue. On a normal day nothing changes: your team answers every call, exactly as they do now.

When the wait time crosses a line you set, say two minutes, our voice agents start answering the calls your team can't get to. They handle the routine ones end to end: order status, bookings, account questions, opening hours. Anything genuinely complex still reaches your team, handed over with the caller identified and the situation already gathered, using the same warm transfer we use everywhere.

When the queue drops back under your line, the agents step back. Your team is the whole experience again.

That's the release valve the maths says you need: capacity that exists only in the minutes the queue is unstable, so no caller ever sits in an infinite queue, and no salary sits idle waiting for one.

We call it auto scaling for contact centres, and it's one of the most effective patterns we've seen for introducing voice AI to a team, because it doesn't replace anyone's job. It catches the calls that were going to be abandoned anyway.

It's also the third of three design patterns we use when bringing voice AI into contact centres, and the two still to come are the ones we're finding even more effective.

---

VerticalAI is a voice AI provider in Perth, Western Australia. We build phone agents that answer every call, handle the routine ones end to end, and warm-transfer the callers who need you. If your queue has ever melted down, we'd love to talk.

Email us at: hello@verticalai.com.au

Visit our website at: https://verticalai.com.au