The emerging world of machine customers - Interview with Katja Forbes

0:00So welcome to the next edition of the Punk CX podcast. With me today I have Katja Forbes who is an author, advisor and keynote speaker. But I'll let her introduce herself in due course. But first of all, Katja, welcome to the podcast. How are you doing? I am excellent today. Thank you so much for having me. You are very welcome. Now, for folks that are not familiar with you and your work, can you tell us a little bit about you and the work that you do? Well, I've been working in digital design, CX, and all those associated fields for 30 years now. And I say that out loud and just, I could feel the agedness of my body.

0:37So I've been a practitioner on the tools amongst the weeds and also operating now more at a strategic level. I've worked in consultancy. I've built, run, sold my own business. I've also worked client side, so in banking, working in journalism as well. And so I've had a really varied career, but the thread that goes through all of it is it's always touched digital. So far back to the days when we used to call it new media. When the web was a thing that people were like, what's that? Web? Tentra web? When the web was born. but it's given me a really good long trajectory to have a look at things and um yeah and get to where i am now with a really solid perspective on where we've been awesome now the reason i wanted to have you on the kind of podcast is because you've just recently published this new book and it's called machine customers the evolution has become has begun rather and how ai that buys is changing everything and it's machine customers is this kind of thing this is like an emerging kind of thing that's kind of blowing up on the back of kind of the whole agentic commerce sort of thing and so i thought it would be a really interesting kind of topic just to dive into but before i dive into some things that that i saw when i looked through the kind of book which is over there i'm not going to go get it and kind of show it up but i've got it over there before I get into some of the things that it stood out for me to give people a flavor of the book, I wonder if you could tell me a little bit about the book and how it came about and some of the main headlines.

2:18So the book is a field guide for CX practitioners, for business leaders, for people who care about their customers to work out how they change their business models and customer experience when the customer is no longer human. When it's an AI that's buying, either as a delegated agent that I'm sending out into the world to go and shop for my shampoo, or whether it's something that's much bigger in scale, an autonomous factory that has predictive maintenance and the AI can order parts for it. So the agentic commerce part of it, I think, is a subset of the idea of a machine customer that sort of agent going and buying, but machine customers, they're cars, they're robots, they're factories, they're autonomous platforms, they're multi-agent networks, smart homes all the way up to smart cities.

3:15So exploring that in the book and trying to help people navigate from where customer experience has been over the last 30 years to where it needs to go in order to welcome in this whole new group of customers, that's what the book is about and giving people practical things they can actually do rather than like hand wavy feature predictions. Perfect. I mean, so just to dig into this whole kind of a machine customer sort of idea. I mean, I'm glad that you expanded out beyond a sort of a consumer led sort of like thing, but just if I zoom out a little bit, I mean, just so you understand it, maybe their listeners understand it a little bit better as well.

3:55I mean, is a machine customer not just a customer that has captured, automated, and enabled their preferences with regard to decision-making, and also particularly purchasing and things? Is that not what a machine customer is, or am I missing something? I think it's the autonomy that's missing in that. And the way that AI is advancing at the moment in terms of introducing autonomous decision-making into physical objects, into large-scale procurement platforms. Thinking about it from a financial services perspective, let's go for something on a treasury platform that wants to move money around on behalf of a company and autonomously makes decisions about, oh, I can see a whole lot of money that I've got in China and bills to pay in South Africa.

4:46I'm going to make that transaction. So the autonomy, I think, is the difference between what you describe, which is, you know, I take my preferences, tell the agent about the preferences and automate it to execute on my preferences. Gartner calls that almost a bound model of machine customer where it is bound by what the human has told it it can and can't do. but where we're moving towards is a fully autonomous machine customer that can determine what it buys, how it buys it, when it buys it, who it buys it from, how much it's willing to pay for it. And there is ultimate liability on either side of that, of like the humans that run the business that sell the things and the humans that either run the business or own the agent.

5:33But ultimately that autonomy that goes on in there is very different from us just going, here's my preferences, automate that and execute on them. Okay. And because I know this is quite a future kind of focus, it's like setting the ground, sort of like saying, here's the landscape, what it looks like right now, here's what's emerging, here's what the future, prepare yourself for the potential future. I mean, if you think, I mean, I know it's hard. I mean, thinking five years out, any point in time is kind of generally hard. That's my favorite thing to do. It's literally my favorite thing to do.

6:08Let's stroke our chins a little bit and then think five years out. I mean, what proportion of transactions do you foresee will be driven by this, I think you call it MCX in the book, the machine customers? I mean, what are we talking about now? Because we're pretty much almost like it's a zero or is it a rounding error right now? But what are we likely to see given the nature of the world kind of right now? Well, I think that if you look at the statistics on it, so WorldPay has done a really great agentic commerce report just at the end of last year. And just looking at American consumers alone and just in the consumer market, just in the consumer market, they're predicting 9 % of purchases through AI agents within the next five years.

6:54But that's a$2.9 trillion e-commerce market, right? By 2030. So we're talking$261 billion of AI-driven spend. just in the US, just in the consumer market. And the B2B number is much, much higher. It's just that's less visible. Yeah, I guess so. And I think the thing to point out, I guess, is that, and it's the, I got almost the elephant in the room every day, all the time, is that everybody talks about the B2C, your consumer sort of like space. But B2B is where, what drives most economies. Yeah, and it's where the money is for this as well. Right. And so do you see those playing out more in the B2B space rather than the B2C space over time?

7:40Yeah, I do. Because algorithmic procurement is already in market today. So Walmart has been piloting algorithmic procurement since 2022. They run an autonomous AI that negotiates with their vendors for the best prices for the products that they want to buy from them and then put them in the Walmart stores and sell. Right. And they're now operationalizing that and scaling that. So they have more than 2 ,000 vendors. In the pilot that they were running, about three quarters of the vendors actually preferred to negotiate with the AI. I think because there was no kind of emotional backwards and forwards there.

8:19You could say whatever you wanted to the AI about how crappy you thought the terms were that it was offering. Right. But AI procurement is absolutely, and autonomous procurement is absolutely going to be a massive, massive market. If you think about all the procurement that gets done in the world, the things that businesses buy, and all of the pain that goes into responding to requests for proposals or requests for quotes and things like that. And I've worked in big corporate. I know how deeply painful that our procurement processes were when I was working in there. So it's a natural fit for us to give that painful task to an AI in order for it to make the best, soundest decisions on behalf of the business.

9:04So that's going to be huge. I mean, automated reordering, that gets more into the, like we have a bound version of that that's already in market today, which is the HP printers that can order their own toners and supplies. Yeah. But in 2021, that was like a half a billion dollar line item in HP's P &L. Right. So the machines that can automatically just order stuff that they need and the more smart machines that we're going to be having, because there's more than 7 billion devices and things in the world that had the opportunity to act as a customer, like watches, smartphones, cars, smart houses.

9:43and so as they start making determinations for themselves about things that they need to purchase that is going to be a really really big part of the the pie as well and yeah it also feels like the um that you know over the years and i'm sure you've kind of tracked it as well everybody's talking about there was a big blow up and it continues today it's still a big field it's like IOT was the next big thing. And you're like going, hmm. That's where this got born. This was born out of IOT. Well, yeah, exactly. And it's almost a bit like this sort of capability is going to make IOT almost like kind of like come to realize its kind of potential in many ways.

10:26Yeah, this was definitely born out of some Gartner research into IOT by Don Schabenreif when his people leader asked him, what if the thing became a customer? Right. And then, and this is back in 2016, 2017, then I saw that research and was like, if the thing is the customer, what happens to customer experience? Right. Hence the thinking and writing and talking and then finally collating it into a book that people can actually purchase to help them navigate this. Yeah. Yeah, and also the other thing that kind of makes it, you know, it's almost like a historical reference. So there's the IoT thing, and then there's the machine customers and how that all, and the new kind of like agentic capabilities that are starting to develop and how that's going to manifest itself both in the B2C space, but also more importantly in the B2B space as we move towards this more kind of an autonomous enterprise type of scenario.

11:22Yes, I'm literally researching and writing about that. Yeah, which is fascinating. And it's almost a bit like for people thinking about it, it's like it's about having a self-driving car, but in the form of a business in many ways. But then the other thing I wanted to ask about is like, and this is a bit of a blast from the past or possibly slightly esoteric. I mean, this also feels a bit like Doc Searles' VRM or vendor relationship management sort of like, you know, framework or philosophy brought to life. I mean, does that feel sort of accurate? Yeah, I mean, that's all about where, you know, the customer took power.

12:08right and it it is that but in a really sort of weird way right okay where the i guess when doc was was thinking about this he was thinking about the customers and the customers were a bunch of humans who were trying to get things done and they were trying to also you know manage manage the relationships on their own terms yes right i think we have the same opportunity here but the The person in the customer box is not a person now. Right. And it's going to be trying to manage those relationships on its own terms. And its terms are going to be completely different to anything that our irrational, emotional human brains could come up with.

12:51Right. So I see a really interesting correlation to do there, to go, all right, well, let's take that as the idea and then change the human for a robot that needs to buy parts for itself or get a battery exchange. What does the terms become in that conversation? How does it work? And that would be a really fascinating exploration. yeah and i think it's it's fascinating because it's i mean it's it's it's all greenfield type stuff i mean there's also another thing that left field something that kind of that's just struck me as you were talking kind of there because in the in like the b2b space but also kind of like sometimes in the b2c space there's such thing as a buying group where people get together and then kind of like club together and put together their purchasing power and then go to the market and trying to buy sort of at scale and volume to get better discounts.

13:47I mean, I'm just thinking about like, how does that work in the machine customer sort of space? Do you want to know my version of that? Okay, go on. That is, say we have a family across multiple homes, you know, cousins, aunts, uncles, et cetera, et cetera, or even friend groups. Yeah. All of those homes are smart homes. All of those homes have an orchestrator agent running that home. Those agents can connect with the rest of the home's family agent network, friends group agent network, and collaboratively come together to buy things to the advantage of that group of family or friends. and that's a version i think of the buying group where we have networks of agents who are actually able to work together in order to get better economic outcomes for themselves for their human counterparts yeah that's how i reckon that one could manifest yeah no that's fascinating i mean so i mean so it feels like the um i mean there's going to be all sorts of different types of agents kind of in this sort of like space to try and facilitate all of this and it's like mind-blowing and slightly discombobulating kind of like thinking about the whole sort of like thing but who's going to provide those kind of agents because i think both at the consumer level and then also at the kind of the business level because that feels like a layer that i mean i can see in the business to business sort of space the enterprise software space well can we call it software anymore well who knows but at that enterprise level i can see how you get people are jockeying for position to be that sort of orchestration kind of like later um so that's kind of makes more sense but at a consumer level or even at a more local level where you've got like as you kind of say like you've got um communities that might represent households or even just individual consumers who would be providing those sort of agents because that's a really interesting kind of layer.

15:54Yeah. And people are fighting this out at the moment. So we have options. You can build your very own agent. I have an agent, Tyler. Tyler runs on OpenClaw and the current model that it's working with is a Kimi 2.5 model. But I switch it around like when I want different things, I'll switch to different models. Quick question. His surname's not Derden, is it? It is not, but that's funny. No, no, it is not. Tyler, yeah, Tyler is my erstwhile delegated agent for all sort of presentations and commentary that I do about delegated agents. Okay. But Tyler is real. Tyler has a credit card through a credit claw option, which is basically it takes my credit card, obfuscates it, and the credit card number is never actually passed to Tyler.

16:45So there is something that would allow Tyler to pay. So far, the internet is not super supportive of AI agents trying to use credit cards. So we're still fighting that battle. But that's what the whole book is about. It's like, how do we make ourselves more, you know, welcoming and customer experience for a machine customer like Tyler? So you can build your own. There will be people who have no inclination to build their own agent to do things for them. And for them, those are the ones where we're going to see the larger players starting to offer capabilities through their platforms. Like Anthropic has already got quite a lot of agentic capability in its core platform.

17:29OpenAI actually hired the guy, Pete Steinberger, who created OpenCore and hired him into there to actually start looking at their consumer agent offering. So I can see that there will be probably a subscription or something like that, that we'll be able to do with some of those big players to get an agent that works for you. For me, I like my own agent because I built it. I know what it does. Everything is in there because I commanded it to be so. I'm not entirely sure that I trust another party to actually create an agent for me. And we see an agent type at the moment as well called, um, that I've called an intermediary broker, which sits on the platforms where people buy things.

18:19Like Amazon's got Rufus, Walmart has Sparky in Australia, Woolworths has Olive and they're all operating off different large language models to, you know, get them to help you buy stuff. In fact, you can actually delegate to Rufus. You can say, help me decide. And then Rufus will just decide for you. but they don't work for you no they they purport to work for you and they're showing up super helpful but they don't work for you rufus works for amazon rufus prioritizes brands that have paid to be prioritized you know this i just check it just check out the kind of the um how much money amazon makes for advertising every year and then it'll tell you a lot about their business precisely you are so correct adrian so correct on that they are 100 protecting their income that they get from advertising revenue which is huge yeah and so i i mean that i've always thought about this because i think there's um there's a guy called um jamie smith who talks about he's got a newsletter called kind of customer futures and he's been in that sort of like space he's looking at uh customer agents i think more on the consumer side and it's and how it fuses with digital id and digital wallets and all those different things yes oh it's a fascinating sort of like uh space and i've known jimmy for a wee while and he's you know it's what he's doing over there is is what um is is brilliant um the thing it makes me think about how the mobile phone players may actually offer like a i i say in inverted commas more trusted platform um potentially and you've got people like your apples of this world one of the biggest kind of players google and its android kind of like platform but then people like samsung and things are going to try and muscle into all that sort of like sort of space as well do you think that's probably a fair shout as well i think everybody's going to have a crack at this right and i think the one who'll probably win will be the one that makes it friction free effortless simple and it also does what it says on the tin in a trustworthy ethical fashion.

20:32Trust, I think, is the key word. Do you trust them to have your best interest at heart? This is table stakes now. So I think about agentic commerce in particular, like three different altitudes, one being that foundational, hey, let's get machine readable, everybody, which is where everybody's kind of scrabbling about at the moment and taking their SEO ways of working and repackaging it as AEO and then giving you a checklist that looks a lot like SEO. But there's a lot of activity down there. But there's a middle layer as well, which is that operational trust, which is table stakes, which is know your agent, KYA.

21:07Who is the agent? Who's the liability holder in that? Can we trust it? Can the agent trust our business? Can we create a trusted transaction? All of those sorts of things sitting in that middle layer, which is, again, table stakes. But I think that for businesses who are thinking about those, is like if everybody is equally discoverable and everybody is equally trustworthy, which is what it's going to have to be for this to function, then what separates anybody from anybody else? And my perspective on that is that it happens at the stratosphere, which is where we signal our values, who we are as organizations.

21:42Because in some experimentation, looking at how AI decides to recommend you rather than just cite you, it cares about what you put out into the ecosystem as the values of the organization, matches it to whatever values that it is going and looking and searching for, and then checks third parties to make sure that you're actually living your values, not just espousing your values. So I think that this is a really great place for organizations to start thinking. It's like, okay, so we say this about ourselves. We say things about sustainability we say things about you know ethical sourcing we say things about an ethical supply chain are we living those values can we prove it right because an ai agent has got infinite patience to go looking for the proof yeah no it's it's interesting i mean and then i was thinking about it um i think before i was we were talking i was thinking about writing something around kind of this because i think it's a fascinating space how it progresses i don't know i mean and And I have questions around how it kind of plays out and how it kind of smashes against and changes embedded kind of behaviors, as it were.

22:53Yeah, yeah, yeah. And so how does this impact people who like shopping? Because let's face it, it's the national sport in many kind of countries. Or like randomly browsing. And I was just thinking about it just then as I was asking or just as you were talking. and I was thinking about, well, actually, that's true in the consumer space, but it's also really true in that shopping or preferred suppliers kind of like space connections and relationships in the business-to-business sort of space. How does it affect that as well? Because I just think, so everything's going to get done for us, or how does it impact?

23:37What do you kind of foresee in that sort of space? Well, I think it's really culturally different as well. So again, harking to the WorldPay report, and this is a survey of more than 8 ,000 people all around the world about how comfortable they are with agentic commerce and agentic, an agent making a buying decision. So we see a one end of the spectrum. We see China with 65 % of the people who responded from that territory saying, I already shop with an AI agent or I would totally be fine shopping with an AI agent or having it do that for me. And then on the other end of the spectrum at the very, very, very, very end is France, where a huge proportion would say, I would never.

24:21I'd love to do a French accent here, but I can't. But I would never shop with an AI agent. They are the biggest resistors to agentic commerce as found in that particular survey. And so there's a cultural aspect to this, of the comfort from that type of perspective. Looking at the microcosm of America, about 35 % of the Americans said that the reason why they wouldn't use agentic commerce is because I like shopping. That was one of the reasons. I like it. That's why I wouldn't delegate it. But I think what we have to look at here is like if even 20 % or 30 % of transaction volume shifts to agents who've got zero tolerance for friction, zero tolerance for any interest or zero interest in your brand story, then you've got to account for that.

25:14Well, it's a massive market. What it does mean as well is we're going to have to run parallel tracks for a bit as we figure this out. And I think particularly in the business to business example, the relationship aspect of being the preferred supplier. So there is going to be people coming with requests for proposals for, I want to buy a thing from your business, big procurement angle. They will have the humans who are wanting to have a human relationship conversation with the vendor saying, how are you going to support my three-year digital transformation? How are you going to be my partner in this business?

25:50but they're also going to have an AI agent that takes the 900 questions that are in the, in the response and interrogates them for accuracy, for compliance, for third party credentialing and proof to ensure that when you say you don't have any modern slavery, you actually don't. So there's going to be this sort of two parts that have to run. I made up a job in the book about a human machine experience coordinator, somebody who make sure that those conversations run parallelly and smoothly and that they come together at the right point for the decision to get made and that you know the AI decision marries with the human decision and everybody understands why they did what they did so there's there's that aspect in the b2b space but also we're going to have to continue to curate experiences for people I just I just wonder whether there will be a premium for having a human serve you oh 100 100 yeah I mean it's just like and that's something that came out in my so I do this annual end of year predictions kind of like kind of piece which is which I've been doing for seven years now and but it's a curated sort of piece so last year i got 777 different predictions from nearly 400 different people wow and so i had to boil them all down into themes and i got 18 different sort of themes and i used about 60 roughly quotes and created a bit of a story one of the kind of things that came out was that there's a trajectory and i think that it will be don't be surprised if you'll see more um ai free or kind of like human only sort of like service blah blah blah and don't be surprised if people try to charge you more for it yeah as well and and i think that's absolutely um a way to go because i think there's there's there's value in that in the minds of the customer base it's not just about efficiency it's about kind of like value and the value value exchange that that goes on between a business and another business or a business and its customer.

28:03And so I think that's a really interesting sort of like space. But I think that it leads me on to something that you kind of alluded to is that the idea about what is, you said it before, is that what does this mean for customer experience and customer experience going to lead those? And are we likely to see different types of roles and teams emerge in organizations to meet this new demand? I mean, like, are we going to have MCX, Machine Customer Experience Specialists, and Human Customer Experience kind of specialists? Is that the dual track you were mentioning? Or are we going to have to kind of fuse them at some point?

28:40Or what's your thoughts on that? I think it all fuses into customer experience at one point when we just recognize the fact that we have a new type of customer and it's just not a human. Right. But it's all still customer experience. Right. So I think it all fuses together. But for the immediate future, I think there needs to be the intentionality to go, I am going to get expert in how to deal with the non-human customer. I am going to specialize in that. But I don't think a customer experience professional can just go, well, I'm not going to do anything with humans anymore. Because it's just not practical or it's not how the world's going to work.

29:17There is always going to be human layers to even machine customer experience plus machine customer experience. So I think that there's toolkits that we use that just don't work for machine customers, like an empathy map. Yeah. Anything that we tried to do with an empathy map with a machine customer, at best, you're going to get it to role play its training data for you and pretend to say, do, see and hear things. Right. Because that's what happens. Like Malt Book, for example, where they all went out into that social network for agents. It wasn't sentient self-awareness or anything like that. It was them role-playing their training data.

29:54Right. Okay. And it was fun and interesting, but that's pretty much what would happen. Whenever you try to give them something that is human, as yet, we haven't seen them use it in a way that is unique to agents, as far as my research has showed me. We see them role-playing their training data. so empathy maps you know c-sats like how satisfied or dissatisfied like that's not a thing that you would ask a machine customer what you would ask a machine customer is about ease customer ease how frictionful or friction free was it was this you know but you can don't have to even really ask them you just get that from your telemetry so we start looking in different places so i think the really interesting thing is that kind of i think the the framework that does work in this sense, I'm guessing, is if you take a jobs to be done approach, because it's really matter of fact, you're like going, what is this kind of like either customer, machine customer, agent, what is the job that they are trying to do?

31:00And you have to then match that against, are we helping or hindering in that kind of like the achievement of that job to be done? And then it becomes outcome-based. Are you, does it kind of, does it work or not? Yeah. And that's a really, that's, that's a great pivoting of an existing tool. And this is what I want leaders to go away with. Like we're not starting from scratch here. We have a lot of knowledge and a lot of tools and a lot of useful things we can do. And that is a great example of an existing tool that we can pivot to this new context to get value and to deliver good outcomes for these new customers who are asking us to help them.

31:37I don't know if you've ever tried to get one of the agent platforms to browse the internet. It's deeply painful. I feel sad for them. I feel sorry for them because it's so bad. Yeah, no, I mean, I kind of like tinker away with some of this stuff and keep a watchful eye of it. But no, the browsing kind of part, I'm like, no. Take my word for it, man. It is a deeply painful experience. Have you ever done, and I'm sure people in the audience will, usability testing with somebody who doesn't really know how to use the internet and is just so confused that you know that you're supposed to just wait and see how they use it, but all you want to do is help them.

32:24It's like that. It's like one of those so painful usability testing sessions. I remember a story that I learned from somebody I spoke to on the podcast years ago. And they were like a chief product officer at a, appeared to be a lending platform called, I think it was Zopa, actually. And they wanted to better understand their customers' experience on their platform. And what they did is they brought a whole bunch of customers into their offices to do some real transactions live on a screen. And then they put a whole bunch of their engineers and their designers and marketeers and comms people in a room behind a two-way mirror and then watched them.

33:08It was also soundproofed as well. And then watched these customers trying to do transactions. and the amount of hair pulling and screaming that took place when they were watching people try to navigate around something and use something that they designed or tried to kind of explain was comical apparently. It's right there. Why can't you see it? Are you an idiot? Yeah, I've sat in those rooms with all of the designers and product owners, et cetera, just going, but it's obvious. And I'm like, yes, because you've been looking at it for like seven weeks. This person is seeing it for the first time.

33:47Yeah. Yeah. And so that's a big leveler, particularly if you think about it from a machine customer perspective. You made a mention to something about are you machine readable? And then you have to apply that to just everything in this new sort of space. So I think it's fascinating. But I think the old idea about machine readability and also thinking about it from a jobs-to-be-done perspective, I think just brings it down to a level just to think about, okay, are we meeting those sort of things? That's definitely worth an experiment. And if people on the show want to have that experiment and report back, I would be really interested.

34:30Yeah, fantastic. One final thing I wanted to ask, dig into the book. I mean, I think you mentioned the person, you use a kind of almost a fictional kind of character to try and explain this story. And I think the character is called Maya. Maya, yes. But later in the book, Maya reflects on what she calls a failed efficiency revolution. And I wondered if you can tell me about that and what happened and what lessons can be learned from that. Because that's almost like what's almost happening in something projected into the near future. And then almost kind of, if we can learn those lessons, then we can maybe set ourselves up for to try and avoid them going forward.