The short answerNot yet, for most of us. AI bots are starting to call contact centers on behalf of customers to ask for discounts, check balances, negotiate bills and cancel subscriptions. Tom thinks more than half of your call volume could be some kind of AI bot in the next 12 to 18 months, and the routing, metrics, security and legal questions are still open.

What this episode is about

This was a live solo episode on machine customers: what happens when AI bots start calling your contact center for your customers. I shared my honest, unscripted thoughts, and I don't claim to have the answers yet. If you hear people say they do, question it.

Gartner predicted that 20% of inbound contacts would come from machine customers by 2026. It's 2026 and we're not there yet, but bots are already negotiating bills and canceling subscriptions. Good bots will introduce themselves. The dangerous ones won't, so how do you unmask them when they come into your IVR?

Then routing. Bots don't care if they wait. Humans do. Should AI talk to AI while humans talk to humans? It also breaks the math: Erlang C calculators, abandon rate and service level all assume callers who hang up, and bots don't.

Last, the legal mess (who's liable when a customer's bot agrees to something the customer didn't want?) and which industries get hit first: retail, internet and energy, and SaaS subscriptions.

Key takeaways

In Tom's words

“it's happening right in front of our eyes right now and I don't know how prepared all of us are for it”Tom Laird · 01:33
“bots don't care if they wait. Human beings do.”Tom Laird · 07:05
“AI is going to have to start to be talking to AI because the volume will just be too big for human beings”Tom Laird · 07:35
“But I think the human beings should be still talking to human beings. I think that's the divide we're seeing”Tom Laird · 07:43
“this is the thing that's keeping me up at night right now. How do we validate?”Tom Laird · 11:32
Full disclosure: I founded OttoQA, so take this part for what it is. We built it in our own contact center. It scores calls against your own QA forms, there are no seat licenses, and if you want to see it on real calls it's a 30-minute demo. Book a demo.

Transcript excerpt

These excerpts come from an automatic transcript and are lightly edited for readability (filler and obvious transcription errors removed). For exact wording, listen at the timestamp.

[07:05]Think about this: bots don't care if they wait. Human beings do. If all you have is humans, do you route them differently if you can find out who they are? Do you let the humans come first?

[07:35]We are at the world where AI is going to have to start to be talking to AI because the volume will just be too big for human beings. But I think the human beings should be still talking to human beings. I think that's the divide we're seeing.

[08:24]The Erlangs and the Erlang C calculators assume people hang up. They assume these lengths of time, but bots don't. Abandon rate can drop while the queue explodes. Does a bot count toward a service level? I don't know.

[11:32]This is the thing that's keeping me up at night right now. How do we validate? How do we authenticate?

Related reading