Clever But Short On Wisdom
The Wisdom Gap in AI-Driven Customer Interactions
I'm pretty sure I got an AI cold call the other day.
The fact that I need to say "pretty sure" should give you an idea about the power of the technology. The latest OpenAI voice synthesis models are completely lifelike and indistinguishable from a human voice.
So why did I think it was AI? I can't quite put my finger on it, but it just had a smell that made me think I was talking to a computer. The feeling was like when you are talking to a customer service agent in an offshore call center who has had Californian accent training and has instructions to introduce themselves as "Joe". The interaction is close but you can always tell that “Joe” isn't from “California”.
There are thousands of people and hundreds of millions of dollars put towards narrowing the uncanny valley for AI call agents so I'm sure that in 24 months the AI interaction will be even better.
But what’s the point?
As soon as I realized that I was interacting with a neural network rather than a human I immediately hung up. My immediate thought: the company making the call clearly didn't value their service (and its ability to solve my problem) enough for them to have a human make the phone call. By having an AI call me the company is implicitly saying that I'm just a number. And who wants to be a number?
Well...when you think about it...we all want to be a number.
But not for everything.
When I buy a flight for a business trip I just want to be a number. I'm trying to get the lowest cost flight with the least amount of pain and suffering. Same with buying groceries. Or filling up my car at the gas station. These are all “I’m just a number and perfectly ok with it” interactions.
The one thing that these interactions have in common is that they are with companies who sell commodities.
This is the problem with the thinking around all of these sales / CS agents or talking head avatars out there. These systems would work well for low margin businesses where there is a lot of price pressure and interacting with customers is something both the company and the customer doesn’t want to do.
When you take a first pass at the TAM for these AI products the markets look very large. Companies can spend an average of $5.60 per customer service call, and there are 200+ billion calls a year. So it’s pretty easy to see how large a company could get if it sells something that can reduce the $5.60 number by one or two zeros.
This would be a great business IF you weren’t selling a commodity. However the underlying technology to handle automated calls is quickly becoming a commodity. Selling the cheapest thing to people who are also trying to sell the cheapest things is a difficult place to be in since everyone in the transaction is under pressure to drive costs down to survive.
Selling automation to banks is a good example of how this plays out. When you need cash both you and the bank probably don’t want or need to actually talk to each other. Given this, one of the best comps for these CS agent companies would probably be an ATM manufacturer like Diebold Nixdorf . Diebold Nixdorf controls about 35% of the ATM market and has a market cap of ~2b and should do about ~4b in revenue. DDB is coming out of chapter 11 so…yeah they are not doing great. Another example of an ATM company is NCR with revenue of ~$4b and a ~$2b market cap. Also not amazing for many reasons but one big one is selling a product with little differentiation into a market where people are trying to always find the cheapest alternative.
You are probably thinking ‘people are looking for the cheapest way to transact in all markets,’ but people don’t do this. The only place people are price sensitive is in markets where customers are just a number to the company and both the company and customers are completely fine with it.
People will pay more when they are not a number. LVMH is a good example of this. An interaction with a LVMH AI sales or CS agent would fatally cheapen the brand. In case you were wondering, this is what happened to Starbucks. They made their customers into numbers and the business went flat.
For this reason, I can’t help but think that a lot of the ‘AI customer service’ or ‘AI sales agent’ stuff going on right now is very clever but short on wisdom. It's clever because on paper it makes a lot of sense. Calls with humans are expensive and are often handled by people who don’t care in call centers. But short on wisdom because AI interactions will only work for a subset of businesses that will put a lot of pressure on the AI provider until there is no margin left.