Machines Are Customers Too

Thank you so much. Alright. I'm here to bring it home. We have spent decades obsessing about human customer experience. We've designed for emotion, we've designed for delight, we've designed for moments that matter, and for a really long time, that was enough.

But I'm here to tell you this afternoon, the customer is changing. Your next customer may never read your carefully crafted copy. Your next customer will not care how your brand makes it feel. Your next customer may never even visit those lovely websites and apps that you've created no whether no matter whether they're for laptop, tablet, or mobile because your next customer may very well be an algorithm.

It might be a delegated agent that's been sent out into the world to find things on behalf of its human and and come back and report back or even just buy them. It might be a self driving car that can book its own services and buy its own tires. It might be an AI procurement platform that has $10,000,000 to spend and it doesn't need humans to close contracts.

And they're not gonna care how your story makes them feel. They will only choose you, your business, and your products if you can make sense to them in the code. So what does it take to be chosen by a machine? What does customer experience even look like when your customer is no longer human? Well, I love to take things for a scenario spin to sort of ground us in a little bit of futuristic reality.

So I'd like to introduce you to my as yet fictional agent, Tyler. Let's see how Tyler chooses.

Good morning. I've been optimizing your life. Shall we review my recent decisions?

Yes, please, Tyler. Let's start with yesterday's grocery order.

Ah, yes. Your fridge noticed you were running low on milk, eggs, and spinach. Based on your preference for organic products and the local store's current sale, I ordered replacements. Delivery is scheduled for 6PM tonight.

Okay. That's great. But, why don't you stick with our usual store?

The usual store had higher prices on organic spinach. By switching, I saved you 12%.

Okay. That's pretty logical. That's fine. But what about the loyalty points that I lost by not shopping at my usual store?

Your loyalty points redemption value is 5%. Decisions still resulted in a net benefit.

Okay. Very efficient. Thank you, Tyler. Let's turn to something a little more personal. I understand that you've been tracking my well-being.

Indeed. Your wearable detected elevated stress levels last week and flagged that you hadn't exercised in three days. To help, I booked a yoga class for tomorrow at your favorite studio.

Okay. So you booked it with my money.

Your wellness goals prioritize stress reduction. This aligns perfectly. Plus, your membership covers the class, so there's no additional cost.

Okay. But what if I don't wanna go?

You are free to cancel. But based on your history, I predict an 83% likelihood that you'll thank me after attending.

Alright, Tyler. So there we have it. An example of, you know, a possible futuristic interaction between me and my delegated agent getting my life optimized and getting things done. It's annoying, a little efficient, you know, trespasses boundaries, nudges me in the direction of good behaviors, and all of that is lovely. But how much trust are we willing to put into this?

Thank you, Tyler.

You're welcome, Katja.

Alright. I don't want you to be thinking about this from the perspective of how do I design interactions between me and agents. We saw a fantastic example of that this morning. What I would like you to do is think about that interaction and think about it from the point of view of the grocery store, from the point of view of the yoga studio, from the point of view of the loyalty program.

How ready are those businesses to interact and transact with Tyler? This is how we try to get people to be aware of our products and get us in the consideration set and get people to be emotionally attached to what it is we're trying to sell them. This is Visa shopping around their travel card and they're using this wonderful aspirational imagery of being in a balloon, going on safari, northern lights, lovely dinners, massages in the rainforest.

Now, for me as a human, I look at it and I aspire to be these people. So, the brand tells me if you choose this brand, this is how you can feel about you. You can be these people. Let's flip that over though and see what would happen if I sent Tyler out into the world and I said, Tyler, go find me the best credit card for my upcoming trip.

I want something with low foreign exchange rate transaction at the fees. I don't actually, I don't want fees at all. I want long interest free periods. I want a loyalty program, and I want it to be associated with, you know, my airline of choice. And also, I care about sustainability, so I want you to interrogate this organization's value chain all the way down to find out whether or not they're going to match up up to my personal set of sustainability values. Off you go, Tyler.

Tyler's not gonna care that I can go in balloons. Tyler's not gonna care that I can get massages in rainforest. The only thing that will get this in the consideration set for Tyler is if it makes sense in the code. Looking at that, you might think an organization like Visa, if that's how they're speaking to their customers, they're probably, you know, in a bit of trouble as the Tyler's of the world turn up.

However, April 30, Visa launched Intelligent Commerce, which is an AI agent that you can send out into the world on your behalf with instructions like, find me the best headphones under $250 and when you find them, buy them. This provides your agent with a tokenized credit card of its own, which has parameters around what it can use it for and what it can spend. So, while you looked at me talking to Tyler earlier and thought, maybe this is a bit of a futuristic thing that's going on, this is in market today.

There are a number of players who've got things like market today. We see all sorts of organizations bidding big on this. Mastercard has agent pay, does the same thing as the Visa version. Although, they obviously say this is better. We've seen OpenAI partnering with Stripe. I think it was about two or three weeks ago, they launched buy and chat GPT where now you don't actually even go to the website to look at the product or understand anything about it.

Everything happens within the machine customer landscape. Perplexity has a tie up with Shopify. You can buy in pretty much any Shopify store. Although, when I was at Singapore Fintech Festival last week, Mastercard made a point of saying that the Perplexity Shopify credit card exchange was insecure. So, everybody started to rag on everybody else's offerings.

And we see Amazon big players and Walmart also big players in this space launching their own in platform agents that can work on your behalf. So this is Sparky from Walmart, and I just don't think that anybody learned the lessons of Clippy, like, ever. This is Sparky. Do I wanna punch Sparky?

Yeah. Maybe. So what Sparky does is Sparky is an in platform agent that is there to assist you, and these are where I think we're gonna see the very first versions of this type of Tyler delegated agent. Sparky will read reviews for you. Sparky will tell you which ones, you know, are the things that people like buying and all of those things.

You can tell Sparky, I'm gonna have a party. Can you put a bunch of stuff in my shopping cart that's gonna sort that out for me? So this is where we are going to see this and it is definitely, inside ecosystems at the moment. Now Sparky doesn't necessarily work for you. Sparky works for Walmart And this is where it's starting to get really interesting because the battle lines are getting drawn.

So I think it was about two weeks ago, Amazon sued Perplexity over its Comet browser and said, Comet is behaving like a human being inside the Amazon ecosystem and therefore, it is in breach of its terms of service because it's not a human being. Therefore, we're going to sue you with our big boy money. Perplexity went, oh, we're so small and tiny.

Please don't. Please don't, horrible giant. And so that's kind of the narrative that's going on there. And you might think, oh, well, maybe Amazon has a point. Maybe they're coming in there and messing around in their ecosystem. Amazon is shutting all the agents out of its ecosystem. It is making it very difficult for any agent tech to get inside the Amazon playing ground.

The reason for this is because Amazon has Rufus. So Rufus is Amazon's version of Sparky. Rufus will look through and find things for you, tell you about reviews, make sure that you understand what you're getting and you can ask questions about the product. However, Amazon has also very recently launched Help Me Decide, where you can just throw your hands up in the air, throw it over to Rufus and say, Help me decide.

I just can't decide. And then Rufus will decide for you. And you might think that you're getting a objective perspective, but Rufus prioritizes brands that pay to be prioritized. So, what's happening here isn't necessarily Amazon going, get out of my system you pesky, you know, agents and trying to protect itself in that way. What it's trying to do is protect its billion dollar ad and promotional revenue stream.

Because when these agents are starting to buy out of the Amazon ecosystem from not within the Amazon ecosystem, they can choose products that have not paid to be put at the top of the list. So this is a it's an interesting battleground watching, how this one is going to play out. This is the retail consumer version. Right? It's pretty pretty straightforward.

You have the agent, you tell it what you want, it goes does the thing for you. It comes back, either buys it, etcetera. Very straightforward. I work in corporate and investment banking, so the clients that I do client experience for, governments of countries, large multinationals, you know, other banks. And I think that there's some really interesting b to b scenarios that are playing out in the same space.

And while I grant you, this is unlikely to take place in a conversational interface in this way, it's a little bit more entertaining to do it that way, so that's how we're gonna do it today. Let's take a spin in a b to b version of this.

You have been identified as a potential vendor for supply chain automation across a distributed logistics fleet operating in the MENA region.

Okay. Well, who do you represent?

I represent node seven forty one, an autonomous procurement agent responsible for optimizing fulfillment for 36 drone networks and their associated logistics systems. I operate with a purchasing authority exceeding 10,000,000 USD this quarter.

Okay. Well, what are you evaluating me on?

Three data points. One, maintenance API reliability must exceed 99.98%. Two, emissions per shipment below 20 grams of carbon dioxide equivalent in line with our sustainability benchmark. Three, predictable latency with under 2% deviation. Ready to ingest your schema for evaluation.

Okay. I'm curious. Who in the audience participates in RFPs? Who loves it? Oh, zero hand and blur noises. Okay. That's new. I haven't had the blur noise before. Yet, nobody likes this stuff. Procurement sucks. It's the seven circles of hell, Both for the people who are putting the RFPs out in there to the market and then have to evaluate the responses and the people who have to respond to them answering a million questions ticking every single box.

Now, while you also might think that this is a bit of a far distant future that I'm pitching for you here with Node seven forty one, Walmart has been running AI procurement since 2022 and it closes nearly 70% of its contracts with its more than 2,000 vendors without a human intervening.

It's fortunate that three quarters of the vendors actually prefer to negotiate with the AI. But this is in market today. This is in market today and it is something that I think we will see becoming very prevalent just within the next twelve months. And I think there's some really interesting things that are kind of come out of this especially in procurement supply chain space where we see, you know, humans on one side wanting to have a conversation with vendors about how are you gonna support my three year digital transformation program. And then on the other side, we've got the AI that's going through every single aspect of the response, ticking every single box and making sure that you're not lying about any of your ESG credentials. So there's some really interesting dynamics coming through, and there's also an amazing pilot program going on in Germany at the moment with a milk distribution factory called Sassenmilsk.

I'm not German. But it is actually got a predictive maintenance system. So an AI predictive maintenance system that says that rotor is off balance and it's gonna wear out. So let's replace that before it actually does break. It's connected that AI to its ERP system so that the predictive maintenance system can order parts autonomously with no humans required in this.

So, these are the types of b to b interactions that we're thinking. Imagine if you're a widget seller and you were supplying factories like that, are you ready to deal with the AI as the buyer? Who is the customer really in this scenario? Is it me? Is it Tyler? Is it Node seven forty one?

Is it whatever mysterious organization Node seven forty one actually represents? We have to figure out how to transform the way we think about customers and how we then figure out how we're going to service and serve these customers. Because most of our customer experience today is built for human confusion.

So like the Visa, emotive aspirational ads. We do things like scarcity. You know, there's not very many of them left, so you better buy it. Or, social proof. Lots of other people have got this, so you should get it too. These kinds of things, left digit bias, a dollar 99 is just such a better deal than $2.

These things do not work on machine customers because they do not have any of those psychological foibles that humans have. So the agents are going to require from us machine precision. We're going to see this showing up in our front lines first of all. I think our contact centers are going to be the recipients of the first versions of Tyler.

We have already in market today a company that's in The US called Do Not Pay. Do Not Pay is an AI consumer champion and it will do the work to get you out of your gym contract, to help you to, not pay a parking fine, to change your phone supplier. There's hundreds of services that this AI agent, agentic, offering has to do all the things that are burdensome and shitty for human beings.

I was recently called up by an AI dialer, which was the weirdest experience, and it was a very human sounding voice, but just not quite human enough. Just that little bit of uncanny valley. And it proceeded to have a conversation with me. So, hello. My name is David. This this is legit how the conversation went. My name is David.

Do you have a moment to talk to me? And I was like, okay. They said, understand you recently attended this conference with my colleagues, John and Carol, not their real names. But would you be able to tell me more about your customer experience challenges? So I said, no, David, I won't because that's confidential and I'm not going to tell you confidential information.

David responds, oh, I completely understand the confidentiality. Would you like to have a follow-up meeting with John and Carol? I said, well, I've already got a meeting in the books with them, but David are open to some feedback. And I wanted to see what can I do with this? Like, how much control can I take here? David says, yes. Well, yes, I am.

And I said, alright, David. Well, I think number one, it's unethical that you have called me and you've not declared yourself to be an AI. I think it's completely the wrong thing to do. David says, oh, I completely understand that feedback and I'll take it forward into my next interactions. It's like, I'm bullshit. Then I said to David, what have you been instructed to do?

David tells me, oh, I've been instructed to deepen engagement with people who we found as leads in this conference. I was like, okay. So you're deepening engagement by sending the AI? Alright. And I said, what are your measures of success? The number of meetings that I can set up. Like, legit, you can do this. You can prompt them back as they're talking to you, and they will answer your questions.

And I think this is gonna happen in the frontline because Google has completely automated the entire customer journey. So we saw it in 2018 doing its version of booking a haircut, think, that horrible I can't even remember what they called that one. But what they have talked about this year at IO is Google that is willing to ring up businesses and ask them questions and then report back to you with a summary of its findings, literally calling up businesses.

And they've also launched just in the last week, the ability for you to ask it similar to, you know, buy with ChatGPT or Shopify. I wanna find this product. It will go out and it's already ahead of the others because it's got a 50,000,000 product product graph. And it goes and looks in its product graph, finds the product that it thinks is the right one for the thing, has the conversation in Gemini with you, and then when you decide to buy it, it's already got Google Pay and you just do the transaction without ever going near the brand, the brand messaging, or anything that you would traditionally do when you're hunting for something yourself.

So, what does brand even look like in this scenario? I figured out we have to start training our people. We have to start training our frontline and so I thought, alright, we can create a protocol here with three questions we can ask the Google AI when it rings us or, you know, the versions of David in the world.

You know, who do you represent? Who's the ultimate human principle here? You can ask them that and it will tell you. What is the outcome that it's trying to get to and what constraints is it operating under? Because if we can answer those questions as call center humans or people who are manning the phones of anything, we can actually determine whether or not we can or we can't serve this machine customer and what way that we might be able to actually be in service to it. We do a lot of customer journey work definitely with better post it notes.

It's on this crappy piece of stock art like problem, journey map, customer. But we do. We pick apart customer journeys and we figure out what people are trying to do, where the touch points are, what's happening on the front stage if we're in service design, backstage, all of that thing. So how do we reimagine the customer journey when the customer is not human, when the customer is an AI agent?

Not like this. This is not the answer. Prompt injection is really unethical. So if this is where your mind was going, maybe not. Maybe, like, have a little revisit of our ethics classes from earlier this earlier today. What I didn't want to do was drop all this on you and then send you home with nothing to figure out how am I going to deal with this in my business.

So I've created what I've called a machine customer experience strategy map to help get you started. Four pillars in it. One is around signal clarity. Let's make our businesses legible to machine customers, There's a lot that we can do in that in our existing skills. So, the fundamentals of information architecture, content strategy, accessibility, accessibility codes, designs and codes for sometimes a non visual customer.

That's exactly what a machine customer is. It is a non visual customer. So, there's a lot that we can take from those existing skill sets that we have and apply them here to make our businesses legible to machines. If we have to create a situation where we can say we trust Tyler to interact with us and Tyler can trust us as a business to interact with, we need to find a way to establish that in a machine readable way.

There are some ways we can do that. So reputation through reliability, telegraphing what our uptimes are, our service level agreements, all of those kinds of perhaps credentials around ESG if that is something that is important, which it is important. So I'm hoping everybody will take that away and put it in, and create those sorts of, reliability measures that a machine customer can go, yes. I can trust this business.

We need to make sure that our offerings are actually aligned with the machine customer priorities. So, the instructions we give our LLMs are not private. We use our LLMs in all sorts of ways that are deeply private, but what we're asking them and instructing them is not private.

So I've kind of wrestled with this one as to, you know, is it a good way to look at customer experience to try and get people to opt in to letting us know how you instructed your agent to do something kinda like cookies, so that we can serve you better or is it just a complete invasion of privacy? I'm not entirely sure where I am with that one, but making sure that the machine customer can understand the value proposition without a human explaining it to it is really crucial.

And having an engagement architecture for machine interaction patterns, machine customers operate in a very different way to a human. They do not have fingers and eyeballs. They don't point, click, and move things around. I don't know if any of you have tried Chat two p t in agent mode or the Atlas browser or Comet browser. It is deeply painful watching them try and work a website.

Like, it's it's like watching a usability test with somebody who's never ever seen a computer. So, that's how bad it is. Maybe it's better since I ran that experiment a few months ago, but who knows? But we need to make sure that we're actually designing for machine customer interactions. I think we'll see probably four levels of maturity here.

Foundation where we've got the basic machine readability, that basic API access because the first question I often get asked is, Katya, isn't this just an API? And it's not. It's the that's API is like the plumbing for it. That's what you need to put in place to be a utility. But unless you can orchestrate and figure out the customer behavior across the top of that, you're not going to be successful.

There's trust to figure out. How do you onboard? How do you service when something goes wrong? What does loyalty look like? An API is not going to help you do any of that. That's customer work. Once we've got that happening, we can get into differentiating ourselves, actual trust mechanisms, know, agent specific customizations. And then when we start to get more mature, I think we'll see ecosystems springing up, actual marketplaces for agents where this kind of commerce is taking place, where agents sell to agents.

And fourth level of maturity is where we get into learning systems, so actually optimizing. So say Tyler learns from interacting with my business and I learn my systems learn from Tyler, and we can start to do dynamic adaption on the fly to create a better outcome for both. I think it's a a real opportunity right now to do this well. I want to also take you for a little bit of a scenario spin with Tyler and show you a version of what I think could happen if we don't do this right.

Alert. Valley Fresh Market's organic spinach listing contains false data.

Okay. Well, what did you do?

They advertised certified organic spinach at $4.99. I cross checked their claim against the official organic database, farm records, and certification registries. Zero matches. The farm lost organic status fourteen months ago.

Okay. So now that what was the next action?

Transaction halted. Violation reported. Valley Fresh Market blacklisted. Their trust score dropped from 94 to 61. That's below my threshold.

Okay. Well, blacklisted. That seems really harsh.

Data integrity is binary. It's either verified or not. I shared this finding with the agent network. Within six minutes, 312 other agents flagged them. Their machine customer traffic will drop 68% by tonight.

Okay. Understood. What do you reckon this is gonna cost them?

They can restore access.

Oh, hang on. No. That's not what I wanted to ask it. I wanted to ask it, how can they get a second chance?

They can restore access by correcting their data, implementing verification systems, improving ninety days of accuracy. Until then, I've sourced your spinach from Greenleaf Coop, certified organic, blockchain verified supply chain, 99.7% data accuracy.

Alright then. So now how do they actually get back in? This is a very horrible outcome for them.

Valley Fresh Market's cost for this data error, one hundred eighty day exclusion from machine customers. Estimated loss, 127,000 in agent driven sales. Your verified organic spinach arrives at 6PM.

Well, you know, at least I'm gonna get my spinach. But you can see here that Tyler has got all of the possible energy that it needs to interrogate the entire value proposition to figure out whether or not what you're saying is true about your products is actually true. And what I think is really interesting about this is this is a moment where we can actually encode our values into the entire system of commerce. Right?

This is what I stand for. This and like more stuff than that that doesn't come in, you know, a picture that's nice and on Unsplash. I also stand for data ethics and data privacy and responsible use of emerging technology for great societal outcomes. And I want that in these systems. If we instruct our agents, you must buy something from a vendor that can prove that it matches your values, whatever those values are, that it actually can verify its ESG claims, that it can do all of the things that it's saying that it can do, the agent has no option but to choose vendors who do that.

Therefore, if the vendors want to capture that market share, they have no option but to have those values encoded in their product set and be able to verify them independently. So this is an opportunity for us to create a really amazing shift in how commerce functions and works, And I think that is really worth putting some energy and effort into.

And if you think this is not coming for you, think again. The red ocean here, this is our traditional e commerce, b to b, b to c, c to b, c to c. E commerce globally last year was a $6,000,000,000,000 market. So, imagine how much money we have explored this to the benefit of in the last thirty years of e commerce.

When we add things into the mix here, cars, smart fridges, smart homes, and agents, Tyler, Node seven forty one, we end up with three times the business model space for us to start exploring. Gartner predicts that by 2030, this is going to be resulting in a $30,000,000,000,000 opportunity. $30,000,000,000,000.

So for all of us, I wrote everything down in a book that I have thought about and know on this topic. This takes you to the Amazon link, which is if you prefer the Kindle, You'll just have to wait till Amazon finishes reviewing my bloody submission, which has been doing for days, but Kindle will be available very soon. I have the physical copies outside for those who are keen to see how I've taken this for a spin.

All of us are here and now you know probably more than 99% of people about machine customers and machine customer experience. We're here to write the rules for this new economy. I think we need to do it thoughtfully. I think we need to do it with our values. And what matters tomorrow is designed and coded today. So, I hope that all of you go home and start thinking about how you're going to get started.

I'm working out loud on this. I've got a community on LinkedIn and I'd be very happy to have conversations with anybody on my very favorite topic. But thanks for the time this afternoon. It's been a delight for me to present to you.