The $5 Trillion Shift To AI-Driven Buying

Imagine never opening twenty tabs to compare flights, hotels or warranties again. Instead, an AI agent that understands your budget, tastes and timing quietly negotiates on your behalf, bundles the best options and asks for approval only when it truly matters. That’s the leap from automation to delegation—and it’s poised to reshape retail at global scale.

TL;DR:

• Delegation beyond automation and the end of shopping friction
• Real-world orchestration example of a full-family move
• Three buying models: agent-to-site, agent-to-agent, brokered
• How ads and SEO lose power to agent optimisation

With Google NoteBook LM agents assistance we dive into the mechanics and implications of agentic commerce, where autonomous systems anticipate needs, transact across competing platforms and coordinate complex outcomes. 

You’ll hear how three models—agent-to-site, agent-to-agent and brokered interactions—power everything from simple bookings to full-life projects like moving house. 

Underneath it all sit the protocols that make agents interoperable: Model Context Protocol for shared meaning, Agent-to-Agent for coordination and AP2 for secure, auditable payments across card networks, wallets and even crypto.

As agents optimise for outcomes rather than aesthetics, the economics of discovery flip. Ads and SEO tricks lose sway, while structured data, reliability and machine-verifiable value rise. 

We outline six upgrades retailers must make to stay visible to agents: machine-readable discovery, personalisation that proves itself, A2A-ready commerce platforms, autonomous-safe payments and fraud, connected stores with real-time inventory and maps, and logistics that are fast and predictable. 

New revenue streams emerge through multi-brand bundles with revenue sharing, success fees on negotiated value and contextual sponsorships where brands become the default inside devices and planning flows.

Trust is the cornerstone. We map five pillars—agent verification, human-in-the-loop controls, transparent reasoning, rigorous data security and responsible governance—and examine the risks of systemic cascades, unclear liability and cross-border data rules. 

The takeaway is clear: commerce is shifting from clicks to intent, and the question isn’t how to find customers, but how to become agent-discoverable and earn an algorithm’s loyalty. If that future excites or worries you, subscribe, share this episode and leave a review with the one task you’d delegate to an AI first.

Link to full article: Agentic commerce: How agents are ushering in a new era | McKinsey

Chapters

• 0:00 From Automation To Delegation
• 0:33 Market Size And Consumer Benefit
• 1:44 Frictionless Journeys And A Moving-House Case
• 3:44 Three Models Of Agentic Buying
• 6:13 Protocols: MCP, A2A, And AP2
• 8:06 Who Controls The Rails
• 8:30 Ads Upended And Six Retail Overhauls
• 10:31 New Revenue Models For Platforms
• 11:46 Trust Framework And Human Oversight
• 13:42 Systemic Risk, Liability, And Sovereignty
• 15:07 From Clicks To Intent And Final Challenge

Transcript

SPEAKER_00: 0:00

Welcome to the deep dive. Today we are getting into something big, uh really big. Agentic commerce. It's this idea that's, well, already starting to change how global retail works. It's not just, you know, fancy e-commerce. It's this future where AI agents don't just give you ideas, they actually act for you. Anticipating needs, finding deals, negotiating, buying stuff all on their own.

SPEAKER_01: 0:26

Aaron Powell Right. Acting autonomously.

SPEAKER_00: 0:27

Trevor Burrus, Jr.: Okay, let's unpack this. Because this sounds like a really seismic shift. We're moving past simple automation, aren't we? It's more like delegation.

SPEAKER_01: 0:35

Exactly. The AI anticipates, it navigates the market, it negotiates, and then it just does it. All based on what the person wants, but crucially, independently.

SPEAKER_00: 0:42

Aaron Powell, so the consumer, the user, you're basically outsourcing the uh the mental work of shopping.

SPEAKER_01: 0:47

Aaron Powell You are. And that outsourcing, it's creating a massive new market. I mean, the numbers are kind of staggering when you look at the research. Oh, yeah. McKinsey, for instance, they're projecting the global B2C retail opportunity here could hit somewhere between, get this,$3 trillion and$5 trillion.

SPEAKER_00: 1:03

Aaron Powell by 2030.

SPEAKER_01: 1:04

Yeah. By 2030.

SPEAKER_00: 1:06

Five trillion dollars. That's not just shifting money around. That sounds like creating whole new ways for commerce to happen just by cutting out the hassle.

SPEAKER_01: 1:14

Aaron Powell That's the idea. And just look at the U.S. B2C market alone. They think up to$1 trillion in revenue could come just from this kind of orchestrated agent activity.

SPEAKER_00: 1:23

Aaron Powell A trillion dollars just in the U.S.

SPEAKER_01: 1:26

The whole point, the core mission is to get this integrated, intent-driven flow going, something that gives you a really frictionless result.

SPEAKER_00: 1:34

Trevor Burrus, Jr.: Showless.

SPEAKER_01: 1:36

The AI doesn't sit around waiting for you to type into a search bar. It uses these complex reasoning models to figure out what you need before you maybe even realize it fully. Then it checks out dozens of options, negotiates, buys it. All that digital pain we feel now, gone.

SPEAKER_00: 1:51

Aaron Powell That friction point is so key. Because right now, shopping online, it's it's work, isn't it? Oh, absolutely. It's all over the place. Different logins, juggling different services. We end up being project managers for just like buying groceries or planning a trip.

SPEAKER_01: 2:06

Yeah, and to really get how big this changes, think about that anecdote from the source material. Imagine you have to move your whole family across the country.

SPEAKER_00: 2:16

Nightmare fuel.

SPEAKER_01: 2:17

Right. I mean, talk about fragmented services. You've got movers, real estate agents, finding schools, sorting utilities, car stuff. A million things.

SPEAKER_00: 2:27

It's just this overwhelming mess that eats your weekends for months.

SPEAKER_01: 2:31

So in this agentic commerce world, one autonomous AI agent could theoretically handle that whole project.

SPEAKER_00: 2:38

The whole thing.

SPEAKER_01: 2:39

The whole thing. It doesn't just book a moving truck. It synthesizes your life. It gets your budget, knows your kids are into, I don't know, competitive chess, understands the dog special food, your commute preference.

SPEAKER_00: 2:51

So it's taking all that qualitative stuff, not just like square footage or price.

SPEAKER_01: 2:54

Exactly right. Then it looks for houses near good schools and chess clubs. It coordinates selling your old furniture online, figures out a fair price, negotiates it, handles the negotiation for the new house, maybe even helps align up financing, coordinates deliveries, service setups at the new place, down to the minute.

SPEAKER_00: 3:11

Okay.

SPEAKER_01: 3:12

The user, you you'd probably confirm the big final decision, but the AI did like 99% of the strategic legwork and the actual doing.

SPEAKER_00: 3:20

That's the lead, isn't it? It's not just automating checkout, it's automating life's strategy. Which explains why people seem ready for this. The sources mention what?

SPEAKER_01: 3:34

Yeah, people are actively looking to delegate this stuff.

SPEAKER_00: 3:37

We want less hassle.

SPEAKER_01: 3:38

The whole customer journey just flips. It might start with the agent prompting you based on your habits, or maybe a life event it notices. Then it goes into this cycle. Find products, maybe negotiate a price, pull together options from different places, buy it, track it, handle any support afterwards.

SPEAKER_00: 3:55

And the user is just out of that loop for all the tiny decisions.

SPEAKER_01: 3:58

Right.

SPEAKER_00: 3:58

The agent is your strategist, your personal shopper who gets your style, your logistics coordinator, your customer service rep, all rolled into one.

SPEAKER_01: 4:06

It basically shifts the hard work of optimizing from you to the AI. Aaron Powell Okay.

SPEAKER_00: 4:11

So the what the benefit for us seems really clear. Massive convenience.

SPEAKER_01: 4:15

Yeah.

SPEAKER_00: 4:16

But the how? I mean, the plumbing underneath all this.

SPEAKER_01: 4:18

Yeah.

SPEAKER_00: 4:19

That seems like the real magic trick. How do all these different agents mine the shops, the delivery companies? How do they talk? How do they negotiate across different websites that are often competing? Here's where it gets really interesting.

SPEAKER_01: 4:33

It absolutely does. Because you need a whole new layer of what's called communication infrastructure, protocols. If agents can't talk, they're just fancy chatbots, right?

SPEAKER_00: 4:41

Right. Limited.

SPEAKER_01: 4:42

So there are basically three main ways a purchase might happen in this agentic world, depending on how complex it is.

SPEAKER_00: 4:49

Okay, walk us through them. Start simple.

SPEAKER_01: 4:50

Simplest is agent to site. Your personal agent goes out and talks directly to, say, one merchant's website or database. It could be like a really smart API call or maybe even advanced scraping.

SPEAKER_00: 5:03

Aaron Powell So like a travel bot scanning hotel sites for the perfect room based on my profile.

SPEAKER_01: 5:07

Aaron Powell Exactly. You don't have to check Marriott, then Hilton, then Hyatt. The agent does it, finds a match, maybe even books it.

SPEAKER_00: 5:14

Efficient, yeah. But maybe feels like an evolution of what we have now with comparison sites.

SPEAKER_01: 5:18

Aaron Powell It is to a degree. The real leap comes with the second model agent to agent, A2A. Okay. This is where your agent talks and transacts autonomously with other agents. So your shopping agent might chat directly with a retailer's own internal AI agent to negotiate a dynamic price. Maybe bundle some items, shoes, and socks from different parts of the store. They're basically haggling digitally. Right.

SPEAKER_00: 5:43

That's where the power really shifts, isn't it? It's not just is there a sale on, it's can our AIs cut a deal right now?

SPEAKER_01: 5:49

Precisely. Then the third model is brokered agent to site.

SPEAKER_00: 5:53

Brokered.

SPEAKER_01: 5:54

Yeah, here you have some kind of intermediary system helping manage complex interactions across multiple agents and platforms. Think of a restaurant booking agent. It might contact a platform like an open table type of service, which acts as the broker.

SPEAKER_00: 6:07

Oh, okay.

SPEAKER_01: 6:08

That broker checks your loyalty status across several restaurants, finds the best fit with the discount, maybe, secures the booking, and then deals with the specific restaurant system. It adds value in the middle.

SPEAKER_00: 6:18

Got it. So for all this talking, negotiating, coordinating, you need common rules, a shared language. What are the key tech protocols making this work?

SPEAKER_01: 6:28

Well, the sources point to three really critical enablers. First, something called the model context protocol or MCP.

SPEAKER_00: 6:35

MCP.

SPEAKER_01: 6:36

Yeah. It's basically about standardizing how these big AI models, the LLMs, understand user intent and share all the important stuff, the context, the goal, the data, across different AIs and tools. It's like a universal translator, so they're all on the same page.

SPEAKER_00: 6:49

So if I tell my agent, I need tough camping gear for hiking way up high, MCP makes sure the shop's AI understands tough and way up high, the same way mine does, even if they use different underlying AI models.

SPEAKER_01: 7:01

Exactly that. Keeps the meaning consistent. Second is the agent-to-agent protocol, A2A again, but the protocol layer. This lets different autonomous agents coordinate really complex actions without needing a human to step in all the time. True interoperability.

SPEAKER_00: 7:14

Makes sense.

SPEAKER_01: 7:14

And finally, maybe the most crucial bit for actual commerce. The agent payments protocol, AP2.

SPEAKER_00: 7:20

AP2. Okay, that sounds like the key to actually letting them buy things.

SPEAKER_01: 7:23

It really is. Google and others are working on this. AP2 is designed to be a secure, auditable way for agents to make payments autonomously. And importantly, it's meant to be payment agnostic.

SPEAKER_00: 7:35

Meaning doesn't care if it's Visa, PayPal, crypto.

SPEAKER_01: 7:38

In theory, yes. It handles the authorization and transactions securely without needing you to type in your card number every time. It gives the agent the authority to spend your money, but reliably and trackably.

SPEAKER_00: 7:50

Okay, but uh you mentioned Google developing AP2. Does that raise flags about, you know, control? Is this going to be truly open or are we heading for a world where maybe Google or Apple controls the rails for how agents pay?

SPEAKER_01: 8:04

That's a really important question. The stated aim is usually open standards because frankly, a closed system limits how much business can happen. But yeah, whoever controls the core protocols has enormous influence, right? Even if AP2 is technically payment agnostic, if most transactions flow through it, the body governing it sees a lot of data and could potentially set rules or requirements, it's definitely something to watch.

SPEAKER_00: 8:25

Absolutely one for sus listeners to keep an eye on. Okay, let's switch gears a bit. If these agents are doing all the searching, the comparing, the negotiating, the buying, what happens to the way e-commerce works now? What about ads? That whole industry built on getting our eyeballs on things.

SPEAKER_01: 8:44

Yeah, that gets turned upside down.

SPEAKER_00: 8:46

The source said it's flipping the script on how we engage online.

SPEAKER_01: 8:49

It really is flipping the script because the whole way products get discovered changes completely. For years, companies obsessed over SEO, getting clicks, making eye-catching ads, optimizing the human journey online. Right. But agents, they don't click ads because they look nice. They're optimizing for the best outcome for their user based on the instructions they have. If an ad doesn't help achieve that outcome, the agent just ignores it.

SPEAKER_00: 9:13

So brands can't just buy their way to the top anymore. They have to actually convince the algorithm they're the best choice, earn the agents trust.

SPEAKER_01: 9:20

That's the core of it. And it means businesses need a massive internal overhaul just to play in this new field. The sources talk about six key areas they have to renovate.

SPEAKER_00: 9:30

Six areas, okay.

SPEAKER_01: 9:31

They have to totally redesign customer engagement and product discovery so agents can easily find and understand their offerings. They need way better clientelling and loyalty systems because agents will demand proof of hyperpersonalization. Their core commerce platforms need upgrading to handle all this agent-to-agent chat.

SPEAKER_00: 9:49

That's a lot already. What else?

SPEAKER_01: 9:51

They need to rework payments and fraud detection for automated systems. They have to connect their physical in-store point of service, think digital store maps, real-time inventory so agents can manage tasks that cross online and offline. And finally, fulfillment and returns need to be super efficient and automated because agents will constantly look for the fastest, cheapest, easiest logistics. It's a root and branch renovation.

SPEAKER_00: 10:15

Wow. Okay, so that's a huge investment for businesses. The payoff must be accessing these new revenue streams you mentioned, the ones replacing potentially declining ad money.

SPEAKER_01: 10:23

Exactly. The money comes from different places. It shifts from selling eyeballs to selling coordination and service. First big one is multi-brand bundling and revenue sharing. Companies collaborate. An agent puts together a complex trip, that honeymoon package maybe. Flights from one place, hotels from another, tours from a third. The agent platform, the orchestrator, gets a fee for making that seamless bundle happen.

SPEAKER_00: 10:48

So the big platforms, maybe Amazon, maybe Google, they shift from just showing ads to being these high-level orchestrators, taking a cut of the total package value.

SPEAKER_01: 10:57

That's one major model, yes. Second is real-time negotiation fees.

SPEAKER_00: 11:01

How does that work?

SPEAKER_01: 11:02

If your agent successfully negotiates extra loyalty points or gets a better warranty or finds a hidden discount, the platform or the agent service itself might charge a small success fee. A percentage of the value they created for you. Exactly. And third, something called contextual sponsorships in connected devices.

SPEAKER_00: 11:23

Contextual sponsorships.

SPEAKER_01: 11:25

Yeah. So brands pay to be integrated deeply into the AI's planning process. Not an ad you see, but maybe being the default option, like the source example. Tesla might pay, so it's charging stations or preferred music service is the default when the car's AI plans a trip. You're sponsoring the context, not flashing an ad.

SPEAKER_00: 11:44

That's subtle.

SPEAKER_01: 11:45

Yeah.

SPEAKER_00: 11:45

But potentially very powerful. Okay. This efficiency, this automation. It sounds amazing on one level, but it feels like it must come with some pretty big risks, too, right? When you hand over that much control to code, trust becomes everything.

SPEAKER_01: 12:00

Absolutely central.

SPEAKER_00: 12:01

The source talks about the trust equation changing. Trust gets filtered through layers of data, automation. So what does this all mean? How do we make sure this is trustworthy?

SPEAKER_01: 12:10

Trust is the absolute foundation. If people don't trust it, it doesn't take off. The whole system has to be built around. Well, the sources identify five key dimensions of trust for agent e commerce. It's about making sure we hand over control responsibly.

SPEAKER_00: 12:23

Okay, let's break down those five dimensions.

SPEAKER_01: 12:25

First, basic but critical. Know your agent or KYC, especially for big transactions, you need to verify the agent's identity, its authority, who it's acting for.

SPEAKER_00: 12:36

Right, like banks do now, but for AIs.

SPEAKER_01: 12:38

Sort of, yes. Second, always put humans at the center, ensure there are controls, human overrides for really critical decisions, and build ethics, empathy even, into the agent's core programming.

SPEAKER_00: 12:51

So the agent might research five houses, but I click the button to actually transfer the deposit.

SPEAKER_01: 12:56

Precisely. Human in the loop for key moments. Third, embrace transparency. The agent can't just be a black box. It needs to explain why it picked that product or that price. Show its work, basically. Show the alternatives it considered.

SPEAKER_00: 13:08

That builds confidence. Makes sense.

SPEAKER_01: 13:10

Fourth, secure everyone's data. This is huge. End-to-end encryption, strict compliance with rules like GDPR, ISO standards, especially because these agents are sharing data potentially across borders.

SPEAKER_00: 13:20

Data security is paramount.

SPEAKER_01: 13:22

And the fifth, you said it was the toughest.

SPEAKER_00: 13:24

Yeah, arguably. Governing responsibly. This means having clear rules about who is accountable when an agent messes up.

SPEAKER_01: 13:30

Ah, the blame game. Right. And ensuring the whole system complies with regulations and all the different places it operates. Which leads us straight into the potential dangers for you, the user.

SPEAKER_00: 13:41

Aaron Powell Okay, what are the big three risks we need to be watching out for as this rolls out?

SPEAKER_01: 13:46

First, there's a massive danger of systemic risk. You could call it the snowball effect.

SPEAKER_00: 13:51

Okay.

SPEAKER_01: 13:51

You have all these autonomous agents making big decisions, often linked together across different systems. One little error somewhere, say a faulty inventory number on one retailer site, could cascade. The agent sees bad data, cancels an order that triggers a shipping agent to reroute something, which messes up inventory somewhere else. It could snowball into major disruptions across linked platforms.

SPEAKER_00: 14:14

Wow, like algorithmic chain reactions causing chaos.

SPEAKER_01: 14:17

It's that interconnectedness that creates fragility. Second big risk, accountability. It's still a huge legal gray area.

SPEAKER_00: 14:23

Right. Who pays if the agent screws up?

SPEAKER_01: 14:26

Exactly. If an agent makes a bad investment for you, falls for a scam, buys that expensive, non-refundable thing you explicitly told it not to, who's liable? The company that made the agent, the shop that took the money, the platform it ran on. The law, like the EU's AI Act, is starting to grapple with this, especially for high-risk AI. But clarity is still, well, evolving.

SPEAKER_00: 14:47

The tech is way ahead of the lawyers.

SPEAKER_01: 14:49

Often the case. And third, a real geopolitical headache. Data ownership and sovereignty.

SPEAKER_00: 14:55

How so?

SPEAKER_01: 14:56

AI is global, right? But data laws are often national or regional. So if your agent, running on a US company's platform, needs to access data about you that's protected under EU rules, maybe via a server in Asia, figuring out compliance becomes incredibly complicated. Data sovereignty is a major non-tech hurdle.

SPEAKER_00: 15:13

Complex stuff. Okay. This has been a really fascinating deep dive. You've seen agenda commerce is set to fundamentally rewire the links between us as consumers, the brands we buy from, and the platforms in the middle. New tech, new rules, new ways to make money. The whole digital shopping world seems to be shifting from being based on clicks to being based on uh intent managed by AI.

SPEAKER_01: 15:35

Aaron Powell That's a good way to put it. We're moving to a world where you, the buyer, are essentially modeled and represented by an AI that's supposed to act purely in your best interest. And that changes everything about what brand identity even means. It becomes less about emotional ads and maybe more about, well, algorithmic efficiency and proven trustworthiness.

SPEAKER_00: 15:54

Aaron Powell That feels like the absolute core shift and leads perfectly into our final thought for you, the listener, to chew on as you think about this emerging world. Yeah. In a future where AI agents are doing the choosing, where humans maybe just give the final okay. The critical question for everyone, shoppers and sellers, isn't just how do I find the customer anymore? It's becoming how do I make my product or service agent discoverable? And how do I earn the loyalty and trust of the agent, not just the person that represents?