Host: Alan Kohler
Guest: Wyatt Roy
I’m joined now by Wyatt Roy who used to be the member for Longman in Queensland and I think he was the youngest parliamentarian ever or at least certainly for a very long time. He lost his seat in the last election and ended up running the Australian end of a business called Afiniti, which was started in 2006 and it’s an artificial intelligence business that uses AI algorithms to match customers with call centre operators. They increase the efficiency and effectiveness of call centres by using data to match the customers with the correct agent or call centre operator.
Very interesting business, making heaps of money, huge fast-growing business and they’ve just started an Australian end last year run by Wyatt Roy. I thought it’d be a good idea to talk to Wyatt, firstly about Afiniti and what they’re doing, but also more generally about AI and how investors should approach the subject of AI and how they should think about not just investing in AI but investing generally in a world of AI. Here’s Wyatt Roy, the General Manager in Sydney for Afiniti.
Wyatt, firstly let’s hear a bit about what Afiniti does and firstly, how it began and what was the original idea of it?
Afiniti is this incredible artificial intelligence company based in the United States. There’s a lot of things that I find really interesting about this company but for me the most fascinating thing is actually our founder. His name is Zia Chishti, he’s 46, it’s his third multi-billion dollar enterprise that he’s built from scratch. His first company was Invisalign, the 3D printer replacement for braces, which he built out when he was at Stanford and I think listed that for more than $1 billion dollars in his late 20s and probably has a market cap above $20 billion today.
He had a number of different companies, one being a private equity fund which effectively had a BPO, contact centre business, and many years ago, about 2006, he realised there was an opportunity to pair two people within a contact centre based on behaviour.
And the two people being the customer and the call centre operator, so you get the right person answering the phone to the particular customer?
Exactly right. Anyone who’s called up a large contact centre for a big telco or bank or insurance company would know it’s not always the most enjoyable experience but the idea follows that if you and I get along I’m going to have a better experience, but from a commercial point of view it also means that I’m more likely to sell you something, retain you as a customer, cross-sell, up-sell, improve collections.
It’s interesting, Wyatt, that in 2006 when Zia started the business he was obviously rich by then, but he wrote the code for it.
He did, literally at the kitchen table.
At the kitchen table, wrote the code. Do you have any understanding of what that code basically did? I mean, obviously not technically, but what was the intention and the process that he wrote?
It’s a pretty simple insight but obviously has a very powerful impact. You’re right, I’m not an engineer or a data scientist so I can’t go to the ins and outs, but if you think about it on the customer side, there’s a number of things that these companies will know about those customers, whether it’s age or gender or demographic data, how often they call, their background with that company, so on and so forth. And on the agent side, if you can imagine an agent in a contact centre, they’re always on the phone. For the last 10, 100, 1,000 calls, they’d been on the phone, they’re not necessarily on the phone with the same person you would hope and there’s lots of inferences that could be made about the different interactions that were happening.
So when a particular agent is talking to a particular person about a particular thing, are they successful or not? And that’s usually some sort of revenue metric – did I make a sale or not, did I retain a customer or not? And essentially at a really basic level, what you want to do is repeat success and avoid failure. When I’m talking to a particular person, if I’m successful we should try and do that. The really interesting part about our IP is, one, how do you do that and get the improvement, but also, how do you do that without having unintended consequences on the operations of very large contact centres.
How do you not affect handle time, how do you not affect a call being answered, how do you not affect agent utilisation in that environment? That’s a very complicated problem to solve and that’s really where our IP sits.
One question that arose as I read about what Afiniti does is, do you use data other than what’s owned by the company, that’s employed at Afiniti? I mean, for example, do you bring in things like Facebook data, Instagram and so on about the customers or not?
Protecting data is something that we thought very carefully about. We actually installed Afiniti on-premise behind the firewall of our clients. We’ve intentionally designed this so that no data ever leaves these clients. So you could imagine that has a commercial negative impact for us because where I learn about a client at a bank, I can’t take that down the road and use at a telco. But when we are deployed with these clients we normally would start with the data that they currently hold because it’s a very large amount of data, but we also have conversations with them and where they’re happy to, where they’re comfortable, we would do a one-way batch feed of all sorts of publicly available data.
Things like Facebook linking, Google, things that can be hash-able to a phone number so that we can identify them. That’s just a one-way batch feed into that appliance. In some clients we do that, in some we don’t. It’s a conversation we have with them and you make an assessment on the benefit that that would have. Often it’s data that would be used for marketing purposes or those sorts of things.
How does it work? The machine instantly matches a customer with the right agent in the call centre?
Yeah, and a really basic element I like to think about is the wedding planners for contact centres. If you kind of imagine, today if I call somewhere where Afiniti’s not there, it’d be like arriving at the wedding and sitting down at the first table and the next person sits down and so on and so forth. Whereas, Alan, if you and I got together and we said, ‘Well actually, why don’t we plan that table? Why don’t we say if Tom sits next to Jenny because they like each other and they like these things and they’ll get along?’ Effectively what we’re doing that in, is in a microsecond and rather than getting the first available agent you’re going to be talking to somebody who the machine has learnt and thought about how you can be better connected.
How did you get involved in this? I mean, you were the world’s youngest parliamentarian for a while I think and then suddenly you’re popping up in Afiniti, what happened?
Well I lost an election, so that definitely made sure I wasn’t in the parliament. But I was really lucky, through some mutual business connections I was introduced to someone who had just joined the company socially and they said that we’d like you to run Australia which initially I had other things that I wanted to do but I got to know some people in the company. I met Zia, who is this incredibly impressive person and spent some time with him in different parts of the world which was a great experience and he really was the reason why I joined the company about 12 months ago as sort of employee number one in Australia.
And is this now your career, in AI?
I’m incredibly happy here, I’m loving the private sector. This is a remarkable company with a lot of people and I definitely see myself here for some time yet. I think for me, one, to grow and build this business in Australia, to see the incredible growth that we have – we have about 1,000 employees globally now in 18 countries and it’s a very exciting time for the company. I think certainly for the foreseeable future I’ll be here and just happy to be a part of an exciting journey.
Obviously, some of the customers globally are quite big and partnerships is a very early on deal with McKinsey and then there’s a partnership with Avaya. I think Sky, the pay TV business in the UK is one of the big customers. What about in Australia, have you picked up some big customers here yet?
We have and it’s very early days. You can imagine effectively going to the B2B market in a completely new market does take some time. We’ve just gone live with a very large Australian insurance company, currently deploying at two other very large Australian companies and making great progress and there’s a long tail after that. I’m really happy with the presence that we’ve made into the Australian market. One of the other things we did which was interesting was the company’s in a strong financial position, not actively sort of looking for investments, but just before Christmas we raised about $20 million dollars from Australian investors for the global company and that’s a great list of significant identities in the Australian business world.
On the advisory board we’ve also added Richard Freudenstein who’s the former CEO of Foxtel and Michael Stock who used to run the investment bank at Credit Suisse, so we’ve really brought a strong Australian flavour into this business in the very early days. I think we’ve had a great start in a new market.
I note from the website that the customers can choose to be charged either by per seat subscriptions or extra revenue or extra profit share basis, that is to say that there’s a partnership or a sharing of the success. What do you find that most customers want to do?
Overwhelmingly, most customers will pay for us on this pay for performance model. This is something that is I think very unique to Afiniti, but also generally in the AI space. We can precisely measure the economic impact that we have and we do that by turning that matching algorithm, so that algorithm matching customer and agent together, on again and off again in very short cycles. It’s all statistics, but normally we’d be on 80% of the time, off 20% of the time – on for 20 minutes, off for 5, on for 20, off for 5…
We do that in perpetuity and that allows us to measure that benefit. When we’re off it doesn’t matter what’s happening, agents could change, the economy could change, the weather could change, office could change – all of those things are reflected in the off sample, creating a perpetual baseline. And the delta between the on and off, that’s us, and we just take a slice of the revenue that we generate. Our business model in most cases is, we carry all the investment costs, we carry all the risk and then we simply revenue share what we create. You could imagine if I’m a CEO looking at this, we improve revenues by about 3-6%, which is pretty significant when there are globally trillions of dollars going through contact centres and some of these very large businesses have literally billions of dollars going through the contact centre.
It’s a kind of a free decision, I get a 3-6% improvement that goes straight to my bottom line and Afiniti just takes a slice of that precisely measurable benefit that we generate. It’s a pretty powerful business model.
What’s the slice?
It’s a negotiation – when we negotiate around term and scale – and it depends what we’re optimising. It’s a relatively easy conversation because this is revenue that you can’t get from somewhere else. We’re not hoping to get more people on the phone, we’re not building shops in the street. Someone’s on the phone, we’re improving those conversion rates effectively which you just can’t get from somewhere else. It’s a relatively easy conversation to have with people but we negotiate it based on a number of variables.
Are we talking more like 10% or 50%?
Somewhere between the two. I’m not about to say what it is, but somewhere between the two.
Okay. Let’s talk about AI generally, I mean one of the things that Afiniti says is that we are not trying to replace call centre operators, we’re making them more effective. And I can see that there’s obviously a lot of truth to that but I did read an article yesterday to the effect that someone was saying there’s going to be tens of thousands of jobs lost in call centres – this was in the UK, this piece – to AI. Afiniti may not be doing it at the moment, but in general wouldn’t you say that AI is going to mean that human beings don’t end up populating call centres?
I wouldn’t get to that point in a great hurry and the reason that I would make that observation is, it’s true, contact centres are sort of a large cost base for a lot of business. Essentially what those businesses globally are trying to do is remove that cost base and the really obvious place to do that is on service calls. If I’m calling up and trying to change my email address or my home address or trying to get something fixed, that’s a service related call and we’re seeing a big shift to chat to bots, to some forms of AI in that space.
But if you think about the Afiniti proposition which is, we’re connecting two human beings, that’s for really high value interactions where there’s a really good reason to have two human beings talking to each other, I’m trying to sell you something, trying to retain you as a customer. What we see in these organisations is, yes, it’s true, call numbers are coming down. Yes, it’s true in certain sectors you’re seeing a significant increase in chat and bots and other forms of AI, but in the sales and retention space, you’re seeing as much demand or increasing demand.
And interestingly, the contact centre industry globally, because of things like push to call – so anyone that’s got a smartphone now, which is the vast bulk of the human population on the planet, if I’m looking at a website or Googling something I can very quickly call someone through that push to call feature and that’s also increasing demand. There is these competing elements to this and I think Afiniti is a great example in the AI space which these stories kind of get lost because of the doom and gloom outlook that often gets out there. This is actually using AI to increase demand to human beings to interact. So you can take from that that’s actually an increase in demand for labour and I think it’s a great story.
I suppose it’s similar to all of this kind of thing where mechanical repetitive tasks and functions are likely to be taken over by machines and AI, whereas the more judgemental things such as what you’re talking about with human beings engaging in sales conversations will stay with humans for quite a while.
Yeah, I think that’s right, Alan. I kind of break it up into two things in my mind, where there’s really easily repeatable tasks AI and machine learning will do a great job. AI is effectively finding patterns in large amounts of data and it’ll do a great job at that and make our workforce more efficient in that sense and increase productivity. The interesting stories I found about AI is when you can use those tools to improve how a human otherwise does their job. It’s something that improves their productivity, it improves their ability to do their job. In many cases there are a lot of things where humans need to be involved and AI is a great empowering element to that and I think the latter is often kind of lost in this broader debate.
Obviously, our newsletter is mostly about investing and our subscribers are all investors. What sort of advice would you give to investors about how to think about AI and to invest not necessarily directly in it but also to adjust their thinking about the future with AI?
I think it’s a really good question and I think in this environment where it’s fair to say AI has a lot of noise out there around it, there’s a lot of hype, there’s a lot of uncertainty. What I would suggest to investors is when people are talking about artificial intelligence, it is a tool to solve a problem and AI is incredibly powerful when it’s being pointed at a singular problem for a long time trying to solve that.
So it looks for those kind of case studies and I think what’s even more important is you need to be able to measure the benefit of that artificial intelligence company or tool or solution and if you can’t measure it, you can’t define that impact, it’s very likely that there’s probably a lot of noise or hype about that and you often see certain companies start up and say, I’ve got some sort of solution, it’s going to cost you hundreds of millions of dollars and we’ll never really be able to see the result but it’s going to change the world. I think if those sorts of red flags are happening I think it’s an appropriate time to put more questions to people.
But where you find those AI companies, those AI solutions that are solving a clearly definable problem with a clearly definable benefit and you can measure that benefit, I think those are the kind of opportunities that are really interesting.
Are there any AI type companies, apart from Afiniti, that you’re excited about, that you think are the real thing?
I mean, generally in terms of trends rather than individual companies I am really interested in terms of what will happen in driverless cars or driverless vehicles and I think there’s a long term horizon on some of these things and often that’s more a regulatory issue than a technology issue. But I think seeing where that goes and how that impacts our society. The other area that I’m really interested in is in the medical space. If you think of, for example, effectively vision learning for these computers or for these algorithms, spotting cancer or skin cancer, these sorts of things, when you’re applying AI in that space I think that has a huge economic impact, but a big positive impact on humanity as well and in many cases is a good example when you apply those tools with a doctor or clinician you can have a huge impact.
I think in that space, again there might be quite long horizons on some of this, but a really interesting space to be looking at.
Just on driverless cars, I saw a chart the other day showing the amount of kilometres that Waymo, the Google subsidiary, has driven compared to everyone else and it’s vastly different. They are way ahead, it would seem, Google is in driverless cars.
Yeah, making a solid long-term investment, I would say. The interesting debate of course that will come out of this – and we saw this play out this week actually with Tesla – inevitably at some point driverless cars will be involved in accidents, we’ve seen very few of examples of this to start with, but if you actually are impartial about this and you say, ‘Well, human beings cause what I assume would be probably millions of deaths on the road across the world every year. The likelihood that robots or driverless cars will do that obviously will be far, far less.’ So the challenge, because there’s emotion involved in that, very difficult for regulators to tackle some of these issues but ultimately long term I would say this technology will save a lot of lives, improve productivity and I think seeing how that plays out over the coming years will be really interesting.
I guess it’s safer to match customers with call centre operators.
That’s right. I don’t know if call centres is quite as sexy but it’s definitely a very interesting business to be involved in.
Great to talk to you, Wyatt, thanks.
Thanks so much, Alan, cheers.