A new type of customer is quietly reshaping the marketplace—and it doesn’t feel a thing.

As AI agents increasingly act as autonomous buyers on behalf of humans, businesses are being forced to rethink how they sell, serve, and connect. These digital shoppers, guided by data rather than emotion, represent a seismic shift in consumer behavior—and a major challenge for traditional marketing and customer engagement strategies.

Unlike human customers, AI agents don’t form brand loyalties or respond to scarcity tactics and influencer cues. Instead, they function purely on logic, assessing variables like price, quality, and ethical credentials based on parameters set by their human users. If an AI is instructed to find organic spinach, for example, it won’t be swayed by a flashy sale or a “limited time only” sign—it will coldly calculate the best value based on its instructions.

This evolution means businesses must become fluent in machine-readable data. Thousands of variables—ranging from supply chain transparency to sustainability ratings and even labor practices—must be presented in standardized, accessible formats so AI agents can make purchase decisions aligned with human values.

“While there is no emotion in a machine customer business transaction, emotional information can still be processed as emotional data,” noted Katja Forbes in a recent episode of Background Briefing.

Banks, too, are preparing for what’s being called “agentic commerce.” One major institution is already developing systems to accommodate AI treasury bots—autonomous financial agents that can negotiate and complete transactions without human intervention. The goal: build the digital protocols that will allow AI-to-AI financial flows to happen smoothly and securely.

Despite the rapid rise of this phenomenon, many businesses—and even the media—have yet to grasp the full implications. “Being ready to receive AI-driven requests is not just a technical issue,” one industry insider said. “It’s about rethinking the entire customer experience for a world where the customer might be a machine.”

As AI agents become more widespread, the businesses that thrive will be those prepared to speak the language of logic and data—flawlessly, transparently, and in real time.

About the speaker:

Katja Forbes
Head of Client Experience, Insights and Development
Standard Chartered Bank

Katja Forbes is an award-winning digital visionary and global authority in customer experience, data science, and emerging technology. Recognised among the Global Top 25 AI leaders in CX and a “Woman of Influence,” she currently leads a gold award-winning team at Standard Chartered Bank. Her role involves steering inclusive and climate-positive financial solutions. Forbes was previously a Vice President on the Global Board of the Interaction Design Association and DEI Director of the Data Visualization Society. She advocates for human-centred innovation, shaping the future of client experience and preparing businesses for the rise of machine customers. She is also writing a book on this fascinating subject.

Frequently Asked Question

1. What is a “machine customer” and how does it differ from a human customer?

A “machine customer” refers to an AI agent that acts on behalf of a human to organise and optimise their life, performing tasks like shopping, booking appointments, and managing administrative duties. The key difference from a human customer lies in the absence of emotional connection or susceptibility to psychological marketing tactics. Unlike humans, machine customers are not influenced by scarcity, social proof, or brand loyalty based on emotion. They operate purely on logical parameters and data, seeking the best price, quality, and other pre-defined criteria set by their human user.

2. How are businesses preparing for transactions with AI agents?

Businesses need to develop digital capabilities to serve autonomous buyers. This involves creating machine-readable protocols to understand product information, establishing direct data connections for payments, or integrating with payment processors that facilitate transactions between the AI agent, the business, and the human’s payment method. The focus is on enabling a seamless, digital interaction where the AI agent can autonomously request and pay for products or services based on pre-set parameters.

3. What kind of information will businesses need to provide to machine customers?

Businesses will need to provide a vast array of data variables about their products or services, going far beyond what is typically presented to human customers. This includes detailed information about the supply chain, environmental sustainability practices, ethical considerations (e.g., modern slavery), and other ESG (Environmental, Social, and Governance) factors. The AI agent, acting on behalf of its human, will be able to evaluate thousands of such variables to make highly informed purchasing decisions that align with the human’s values and sustainability ambitions.

4. Can emotional preferences or subjective human values be communicated to a machine customer?

While machine customers don’t experience emotions, emotional information can be processed as emotional data. Humans can express subjective preferences, such as wanting a holiday that is “comfortable” and “relaxed” or a product that aligns with personal values. Businesses will need to signal emotional data variables about their offerings (e.g., a hotel having an “incredible spa that provides parental relaxation”) for AI agents to match these emotional needs. The AI agent acts as a translator, inferring needs from descriptions and matching them with relevant emotional data provided by businesses.

5. How will the interaction between humans and AI agents evolve over time?

Currently, strong and explicit instructions are often required from humans when interacting with AI agents due to the nascent stage of the technology and potential for “hallucinations” or errors. However, as AI agents become more sophisticated, learn from every interaction, and are trained on larger datasets about human behaviour, they will become more competent and capable of inferring human needs without such explicit instruction. This will lead to a more intuitive and less friction-filled user experience.

6. What is “agentic commerce” and its significance for the banking sector?

“Agentic commerce” refers to the entire ecosystem of commerce and economies that emerge when AI agents and autonomous buyers are involved in transactions. For banks, this presents a significant opportunity. Banks like Standard Chartered are positioning themselves as facilitators of these new financial flows. Instead of traditional human-led interactions (e.g., a client calling to request a currency transfer), banks need to be ready for AI treasury bots to autonomously transact with them, providing protocols for automatic and autonomous connections to facilitate cross-border movements of funds and other financial operations.

7. Beyond consumer applications, how is AI impacting business-to-business (B2B) transactions?

AI is rapidly transforming the B2B space, particularly in areas like autonomous procurement. Companies are already using AI-driven procurement systems that can negotiate with vendors without human intervention. For instance, a large US retailer closes nearly 70% of its vendor contracts using smart contracts and AI, with three-quarters of these vendors preferring to negotiate with the AI. This highlights a shift towards more automated, data-driven B2B interactions, where businesses must be prepared to receive and process transaction requests directly from AI systems.

8. What crucial aspect of AI-driven commerce is often overlooked by journalists and businesses?

Journalists and many businesses frequently focus solely on the human instructing the AI to do something. However, the often-missed aspect is the readiness of the business on the other side to receive and facilitate the transaction request from the AI agent. Businesses need to proactively prepare their digital infrastructure, data presentation, and operational processes to effectively serve these non-human customers. This involves understanding the unique requirements of machine customers, such as their data-driven decision-making and lack of emotional responses, to ensure they can capture the “awareness” and ultimately secure transactions with these autonomous entities.

5W1H summary

Transcript of the interview

Okay. It is a little bit of a brain breaker. I recognise that. So, let’s ground ourselves in a for example to understand it.

Sure.

All right. Imagine a very not-distant future where you have an AI agent that organises and optimises your life. It will do your shopping for you. It will book your appointments for you. It will do all of those administrative tasks that make your life much, much easier. Now, I’m not wanting to talk about what experience you have as Mark with his AI agent. That’s somebody else’s thing to play with. The thing that I’m interested in talking about is when your agent tries to book a haircut for you, tries to buy something for you to put in your smart fridge, how does the business that your AI agent is trying to transact with get ready for that transaction? How do they provide customer experience for a customer that is not human?

And so presumably if it was a human being, they would order something on a website. There’s lots of e-commerce type platforms and then that order is fulfilled by somebody who then stocks up the smart fridge. How would it be in the future?

In the future, and it’s a very near-term future because we have offerings in market today that allow on our credit cards Visa has an offering here, Mastercard has an offering here that allows an agent to pay for things on your behalf, that’s in market today. So it’s not far. So what it would be is the AI knocking on the door of the business digitally saying I want to buy this. These are the parameters that I’m looking for in terms of what type of product and what is acceptable to me and these are the permissions that I have in terms of transaction. Can you please serve me? Now, it’s up to the business to figure out how to digitally serve that autonomous buyer, that AI agent. Is there a technical way that they do it? Do we create a machine readable protocol for them to understand what the product even is? Do we create a data connection so that the AI agent can pay directly? Do we have a connection between a payment processor separately that the agent and the business and the payment processor create a process that allows that agent to pay on behalf of its human. So this is the things that we’re trying to figure out in the customer experience evolution into this machine customer landscape.

So far though, that doesn’t sound all that different from a human customer knocking digitally. I go to a website, here’s my credit card details, can I please have the product? Is there something fundamentally different about the machine transaction that we don’t have with humans?

Yes. Because there is no emotional connection with the brand of the product that is being sold. So as we try to get awareness of our products, we’re trying to capture human attention. We have optimised for human confusion really. So scarcity, oh my god, there’s not very many of these left. You should really buy that. That gets a human engaged and moving towards a state to I must buy the thing or social proof. Lots of people have got this and like this so you should get it and like it too. Now those kinds of psychological tricks don’t work for a machine customer. They will not work on the AI agent who is working on your behalf. Your agent when it’s shopping for organic spinach will go and find the product that is the best price in terms of the parameters that you Mark have set for it. It will go and get it at the place where it can see it is going to get the best quality if that is a parameter that you set for it.

But it will not care whether or not that organic spinach is on sale from $2 to $1.99 if its other parameters aren’t met. So none of that emotive pull will work on a machine customer. And that is the key difference between trying to get awareness and selling to a human and trying to capture the awareness and sell into a customer.

Cool. So the next question then must be if you’re the one selling the spinach, what must be in your digital storefront to get the AI to stop at your door?

So this is where it gets really interesting because while a human being will probably evaluate a couple of data points about a product, is it spinach? Is it organic? Is it at the price point that I want? AI agent operating on behalf of its human in that machine customer mindset will be able to evaluate thousands of data variables about that spinach. It will be able to evaluate all the way down the supply chain of where that spinach came from. How environmentally sustainable are the practices of the spinach grower? Do they have modern slavery? What are the other things that form part of that ESG portfolio of considerations that we could have our agents interrogate to make sure that it matches our human values, whatever those are, and also perhaps our sustainability ambitions as a human being operating in the world. Now, Mark doesn’t necessarily have time to interrogate all of that, but Mark’s agent does. So, what we need to present as the spinach seller are all of those possible data variables about that product so that the agent can make the best possible decision in the tolerance of the parameters that have been set for it. But in the end, it’s still a human being that consumes the spinach or the milk or whatever it is that the machine is buying on your behalf.

In other words, I’m still going to have that milk on my breakfast table. And if the bot has bought something that meets all the parameters, but the brand is important to the human consumer. Do you foresee that actually the machine will simply buy the brand that I have made those subjective decisions about previously? That actually it’s still a human who starts that purchasing process with all of those subjective considerations.

This is where it gets even more interesting. You can tell I’m quite enthusiastic about this.

You’re writing a book on the subject.

I am writing a book on the subject and it’s utterly fascinating to me. So, while there is no emotion in a machine customer business transaction, emotional information can still be processed as emotional data. Psychologists and psychiatrists have been collecting and processing emotional data for decades. Sentiment, the how people perceive an experience I suppose. I can see a future state where the human is able to express a bunch of emotional things like let’s imagine we’re going on a holiday and we say to our agent I’m going on holiday I’m taking the kids so I want to be somewhere that’s comfortable and I need to be relaxed go and find me something that meets those parameters and I want it to be under $350 a night. There’s a whole lot of inferences that these AI agents and machine customers are going to be able to make from that description what does comfortable mean what does relaxed mean? And in order for that to be able to match the emotional needs of that instruction, the people who are on the other side of the transaction selling those holidays need to signal emotional data variables about their product offering that this place has an incredible spa that provides parental relaxation, that this place has got kids facilities that are comfortable and you won’t have to move around to different places. So there’s a matching that has to go on and the agent in the middle becoming the translator as it carries out those tasks.

And is this going to be done automatically or are we still going to now that we still have to become very good at instructing the AI? We no longer need to be good necessarily to find the best deal for that holiday destination. But if we’re not asking the AI the right questions or setting the right parameters and that sounds like a lot of work actually.

It starts that way. I think as this field is quite nascent. So at the moment a lot of the agent stuff that is out in the market is quite hype-centric and so we are creating things that aren’t super great at what they need to be able to do yet because they’re all still learning and we’re still learning and it’s a tricky landscape to be operating in.

Well, you’re at the cutting edge by definition. There’s a crystal ball gazing.

Naturally. And so I think what we’re going to be seeing is probably quite strong instruction that needs to happen. And the friction territory here, I think, is that so many of these AI consumer AI offerings appear so capable and so human in the interaction that we as humans can’t help but naturally ascribe capability to it and believe that it’s really good at what it says it’s good at doing. Yet all of us know that hallucinations happen, that it gets things wrong, that we have to check. And so at this stage there’s going to be some friction there and that very explicit instruction of what you want and what you need from it is going to be required. However, as they get better, as they learn more, as they understand how humans operate, as we train them on bigger and bigger sets of data about how we do things. They will become more competent and capable of inferring our needs from what we have interacted with them. So it will take every single interaction that it has and learn from it in order to do a better job. Now the consumer version of this is pretty easy to understand I think. Yes you have an agent it does things that you don’t want to do pretty easy. But the other side of this I think is in the B2B space because I work in corporate and investment banking at Standard Chartered and so the clients that I do client experience for governments of countries, large multinationals, other banks. So for me exploring the B2B side of this is also very interesting and worth a conversation.

Finally then what questions should journalists ask? What is it that they often miss when you listen to the questions that you could help advance their thinking.

They see only one side of the transaction which is the human being or the business instructing the AI to do something. What I see missed over and over again is as the business on the other side receiving the request for transaction from that AI being ready to answer that, being ready to facilitate that transaction. We’re going to see it in the B2B space where autonomous procurement is going to become everyday. There are organisations that have autonomous procurement, AI-driven procurement that provide the ability for vendors to negotiate with their AI. So there’s a large retailer in the US who currently closes nearly 70% of their vendor contracts without a human in the loop using smart contracts and three quarters of those vendors prefer to negotiate with the AI. So that organisation is on the front foot of that space where I see journalists not asking questions and businesses not being ready. They are ready to serve a customer that is not human.

You mentioned that you’re in corporate and investment banking. How does the bank view this? Because in a sense you’re actually in the middle of this conversation between the consumer and the business, isn’t it?

Yes. So the current catch-all for this, because we know that language on this is changing often, is agentic commerce. So the commerce that happens when we have AI agents and autonomous buyers in the transaction and the economies that create are created from that. We have three times the business model landscape to explore here than we did with the traditional e-commerce. So there is a huge opportunity pool. The way that the bank is looking at agentic commerce is seeing how can we position ourselves in this commerce as a facilitator of the flows as the cross-border network bank that we are so strong at being. And the ways that we can do that is say let’s imagine we have a client who has an AI treasury bot that optimises its liquidity that looks at I have liquidity in China and I have bills to pay in South Africa and because I am autonomous and I have the parameters that I’m allowed to operate in I’m going to move my R&B over to Rand to make sure that my business in South Africa can pay its bills. Now that’s not somebody calling up the bank and our relationship managers saying hey can you move the R&B to Rand for me please by this date because I need to do it which is this human sort of transaction. So the bank needs to be ready for that AI agent from the treasury platform to transact with us and for us to go we can help you move that flow. We can give you the protocols to connect to us to do it automatically to do that autonomously and be present for our clients in the way that they need us to be there at the times that they need us to do it. We see a lot of opportunity pools in this space and as we’re exploring agentic commerce more and more of them are becoming apparent and so we are looking to be a differentiator in that market.