Agentic commerce: How agents are ushering in a new era

Chapter 3 | Business model evolution in the era of agentic commerce

In 1942, the economist Joseph Schumpeter coined the term “creative destruction” to describe the power of innovation and new technologies to destroy existing economic structures—jobs, companies, even entire industries—while simultaneously laying the groundwork for new structures to emerge. The emergence of agentic commerce represents precisely this kind of moment for business leaders.

Adapting to this paradigm shift will require more than a user interface refresh. Instead, it means grappling with a fundamental restructuring of how products are discovered, how buying decisions are made, and how customer relationships are formed and maintained. After all, in an agentic world, your customer may no longer be a human with a browser—it is just as likely to be an autonomous agent, acting on that customer’s behalf.

We are currently in a phase of rapid experimentation—the future shape of the customer journey remains uncertain. It is safe to say, however, that the widespread adoption of AI-enabled conversational interfaces—ChatGPT now has more than 800 million weekly users, and Google’s AI overviews powered by Gemini now reach more than 1.5 billion users per month—means commerce will increasingly flow through AI channels, impacting billions of consumers and tapping into a significant share of global disposable income. In this hyperconnected age, dramatic shifts can occur rapidly; as agentic-interaction models are tested in the market, consumers will begin to express their preferences through their behavior. As they do, businesses will be forced to rethink how they engage with customers, optimize operations, and deliver value—in many cases, fundamentally reinventing traditional business models. It won’t be the first time a new technology has forced these kinds of changes. Historically, each wave of technological disruption has reshaped where value is captured, and the AI/API disruption will be no exception.

How will businesses respond to agentic commerce? At a minimum, product directories will need to be optimized for agent readability. Many companies will pilot new agent-first experiences. And others will reimagine their models entirely, owning the orchestration and becoming indispensable in an AI-agent-driven ecosystem. But one thing is certain: Remaining static is not a viable option.

Innovate or renovate? Six key domains businesses should consider

To thrive in the agentic era, retail businesses will have to address six key domains of their business models: 1) customer engagement and product discovery; 2) “clienteling” and loyalty; 3) core commerce platforms; 4) payments and fraud detection; 5) in-store point of service; and 6) fulfillment and returns. The challenge for businesses will be to think strategically about each of these areas—discerning where to innovate (building entirely new strategies and structures) and where to renovate (upgrading existing systems and processes to perform in an AI-native environment).

In three of these domains, innovation will be paramount. To drive engagement and discovery, companies need to develop agents capable of understanding customer intent and proactively suggesting products, services, bundles, or other alternatives. This will require embedding semantic and behavioral metadata into product catalogs, while also developing agent-authenticated interfaces for autonomous discovery and verifying intent and identity in real time to ensure that transactions are both secure and efficient. Driving clienteling and loyalty, meanwhile, will demand new experiences that are hyperpersonalized, with offers triggered by inferred intent. This will require building persistent customer-context layers that are accessible by agents and exposing loyalty services and eligibility engines via APIs.

Renovation, by contrast, will be essential in areas where legacy systems hinder AI integration. Core commerce platforms, for example, should be revamped to enable agents to execute structured transactions with minimal human input, incorporating AI capabilities like dynamic pricing or inventory-aware recommendations. In-store point-of-service systems must be retooled to synchronize digital and physical journeys by sharing context with store associates, accessing digitized store maps and inventory, and integrating spatial computing for in-store navigation. Finally, fulfillment-and-return systems will need agents able to automate fulfillment decisions, negotiate return logic, and orchestrate postpurchase actions—which will require agent-ready fulfillment orchestration APIs and integration with multicarrier and last-mile brokers through modular connectors.

Collectively, these domains form the backbone of an agent-ready commerce stack, and a merchant’s ability to thrive in an AI-native future will likely hinge on its ability to adapt across all six.

Disrupting your own business

The bottom line is that businesses must be willing to disrupt their own processes and models to stay ahead. This means rethinking traditional e-commerce strategies and revenue streams to embrace AI solutions that can improve product discovery, customer service, and postpurchase experiences. Of course, there will be nuances. Many consumers, after all, enjoy the experience of shopping, while others might opt for as much automation as possible. By integrating agentic AI early and investing in the right infrastructure—whether through AI-powered recommendation engines, chatbots for real-time assistance, or smart shopping assistants—businesses can gain a competitive edge by offering an increasingly frictionless, intuitive shopping journey for all their customers.

As discussed in the previous chapter, none of this will be possible without a new technical infrastructure structured around APIs that enable communication among different software systems. To facilitate smooth operations and seamless user experiences, businesses should focus on building an efficient, intuitive API infrastructure tailored to agentic needs, ensuring that AI agents can navigate the platform securely, authenticate users correctly, and manage secure transactions. And it will be crucial to build the capabilities to differentiate between a benign agent and a malicious bot. (For more on managing agentic risk and trust, see chapter 4 below.)

As AI agents increasingly influence consumer purchasing decisions, businesses must evolve to ensure their products and services are easily discoverable—not just by people, but by the agentic systems acting on their behalf. In fact, designing the “agent experience” could soon become as important a consideration as the customer experience. OpenAI took a major step in this direction with the recent announcement that it was releasing an AI-native web browser with agentic products, such as Operator, integrated into the browsing experience. This shift transcends traditional SEO, which stands to become less relevant in an agentic world. Instead, companies will need to understand and align with the data structures, preferences, and decision-making logic of AI agents—while still preserving the emotional, brand-driven experiences that build trust and loyalty with people.

This dual approach has the potential to unlock new revenue streams. That’s because in agentic commerce, the path to purchase begins when intent first ignites. AI agents, acting on behalf of consumers, can detect and interpret early intent through contextual signals: a calendar invite for an upcoming move, a message about a new baby, or a search for homes in a new city. By engaging at this upstream moment, businesses can position themselves not just as vendors but as integral solution providers in the customer’s planning process. This early access allows businesses to capture intent before a consumer even visits a product page or compares different options, effectively bypassing the traditional discovery funnel. In the case of our cross-country move, an AI agent might proactively assemble a comprehensive shopping plan: sourcing and comparing furniture options, identifying local service providers, calculating shipping costs, and optimizing timing. For those planning a vacation, an agent could preselect destination options based on calendar gaps, loyalty programs, and travel preferences.

These capabilities shift brand presence from being chosen at the point of comparison to being present at the point of need—a process with the potential to scale to thousands of micro-interactions across customers, with personalized precision and little additional human labor. The result is more efficient customer acquisition, higher conversion rates, and greater lifetime value.

Similarly, rather than relying on generic promotions, AI agents can negotiate tailored offers for customers—say, bundling furniture purchases from various retailers during a move, customized to fit budget, style, and delivery preferences. Businesses with subscription models could benefit, given the ability of agents to manage replenishable recurring purchases, whether coffee, skin care, or pet food. Real-time dynamic pricing is another potential frontier, with AI agents assessing competitors, customer intent, and inventory to propose optimal prices that maximize conversion and protect margins.

Beyond these front-end innovations, agentic AI streamlines back-end operations, from inventory planning to service automation, reducing costs and enhancing scalability. In essence, AI agents are not just a new channel; they are a catalyst for reimagining value creation and capture throughout the customer journey.

Managing risk from ad revenue decline

Of course, agentic commerce also threatens to bring about a decline in traditional revenue streams, particularly from advertising. Retail media networks, which rely on ad-based models, could face challenges as consumers shift toward agent-driven experiences that bypass traditional ad channels. Businesses will need to diversify their revenue sources and consider innovative ways to monetize through AI, such as charging for data insights or offering premium services to consumers using agents. In this evolving landscape, data from agent usage will be pivotal in guiding monetization strategies. New approaches to branding and partnerships will be essential, with potential models including affiliate-type arrangements, discovery premiums on marketing, premium placements, and subscription fees. The space is still nascent, but the following are a few examples of innovative monetization:

• Multibrand bundling and revenue sharing. AI agents coordinate purchases across multiple brands, bundling them into seamless experiences. For instance, when a user shops for a honeymoon package, the agent negotiates and compiles flights, hotels, excursions, and dining from different providers. Each provider receives a share of the total package fee, while the coordinating AI platform takes a service or coordination fee.
• Real-time negotiation fees. Agents can negotiate in real time on behalf of users, such as securing hotel upgrades or loyalty point redemptions. Platforms enabling these negotiations can charge a success fee or margin per transaction. For example, an airline might pay a fee when an agent successfully negotiates a seat upgrade.
• Premium skills and subscription models. Frontier labs or third-party developers can offer vertical AI agents, such as fashion stylists or upscale trip planners, through subscription plans or tiered access to advanced features and integrations, generating recurring revenue.
• Data insights and analytics sales. Brands can pay for anonymized, agent-filtered consumer behavior analytics to provide data on product consideration or rejection, price sensitivities, and competitor comparisons, helping brands refine marketing and product strategies.
• Conversational marketplaces. AI agents could evolve into full conversational marketplaces, where purchase decisions occur through dialogue. Marketplace owners monetize via brand listing fees, sales commissions, and payment-processing or protection fees.
• Interagent protocol fees. When AI agents from different platforms interact, monetization can happen through protocol-level fees for interoperability or commission sharing when joint value is created.
• Contextual sponsorships in connected devices. Brands can sponsor contextual experiences through integrations with wearables, autos, or smart homes. For example, Tesla might pay to be the default electric-vehicle provider for AI-planned trips, or Spotify could sponsor music during AI-curated events.
• Sponsored, context-aware suggestions. While direct advertising can compromise consumer trust, subtle “sponsored smart suggestions” aligned with user intent can be monetized, provided the agent maintains impartiality to preserve user trust.

First-movers have the unique opportunity to set and define these pricing models, capitalizing on the current monetization gaps where agents are often free and providers have yet to develop a clear revenue strategy.

Finally, as companies rethink their business models, they should proceed with the understanding that success with AI also necessitates a fundamental rewiring of organizations and the empowerment of employees. Fortunately, while leaders often cite employee readiness as a barrier to AI adoption, McKinsey research shows that employees are more prepared than their C-suite leaders believe.

Chapter 4 | Navigating trust and risk: Why responsible AI is key to agentic commerce adoption at scale

In this evolving landscape of agentic commerce, trust transcends mere consumer sentiment. It becomes foundational infrastructure—and it will likely come under significant pressure.

The trust equation

When a person walks into a store, the trust equation is straightforward: Do I trust this brand, this merchant, this product? However, when an AI agent shops on your behalf, trust becomes abstract, filtered through layers of data, automation, and institutional frameworks. This shift prompts a profound question: Who do we trust when we aren’t the ones making the choices?

For many consumers, the answer might simply be, “no one.” In countries like Germany and Japan, for example, consumers still prefer traditional payment methods—such as invoices or buy now, pay later (BNPL) options—over credit cards, which tend to drive e-commerce globally. Instead, account-to-account transfers dominate in Germany, accounting for 26 percent of e-commerce payments, alongside a notable 20 percent via BNPL. This shows a clear preference for methods that emphasize control and transparency. In such environments, agentic transactions might indeed be a leap of faith too far. If consumers are already hesitant to share their banking details with a static website or even a familiar payment platform, how likely are they to entrust an AI bot with not only payment credentials but also the decision-making process that leads to a purchase?

Getting AI trust right is a shared responsibility between the organizations deploying AI and the platform providers, governments, international organizations, and standards bodies aiming to ensure that AI is safe and reliable. In this dynamic environment, academic researchers, open-source communities, and developers also play a big role in building AI that is more trustworthy, transparent, and explainable. CEOs and CTOs can do their parts by getting their data houses in order, empowering teams to innovate safely, and monitoring all AI deployments for signs of bias or misinformation.

Roger Roberts, McKinsey partner and global leader of the Digital Trust service line

Trust, in other words, is deeply contextual. What feels intuitive in Seoul might be unthinkable in São Paulo. Adoption doesn’t necessarily follow innovation; it follows comfort, norms, and credibility. For agents to earn trust, they should be built around people, not just processes. That means technologists must meet users where they are—not just with legal disclaimers, but with clear, ongoing dialogue. Trust, after all, is not a one-time agreement; it grows through interaction (Exhibit 5). Users should be able to ask: How is my data being used? What does this choice mean for me? They should also be able to define the boundaries of that trust. If agents are truly tools, users need intuitive ways to express and define what they’re comfortable with, preferences that the system can understand and respect. Consent, in this context, cannot be a checkbox; it must be a living, flexible agreement—one that is deepened and shaped by the user through agent–user interactions.

The risk landscape of agentic commerce

If trust serves as the foundation of agentic commerce, then risk acts as the architecture’s stress test. Trust encourages participation, while risk delineates the boundaries of that engagement. As agents gain autonomy—operating across systems, borders, and industries—new risks emerge that traditional compliance frameworks and tech governance are ill-prepared to address. Organizations should consider the following three key areas of risk:

Systemic risk: The snowball effect. Autonomous agents are more than just interfaces; they are decision-makers. Decision-making at scale introduces systemic risk, where a single faulty prompt can trigger a cascade of unintended consequences—an incorrectly booked flight, overordered inventory, a purchase that occurs without consent. When agents are interconnected across multiple systems, minor errors can have exponential impact, and resilience becomes a crucial design principle. Do agents fail gracefully? Can they backtrack? How can businesses repair reputational damage caused by a nonhuman error?

Accountability: Navigating a legal gray zone. When an AI agent makes a poor decision, determining accountability is complex. Who is to blame for that faulty transaction? The platform that developed the model? The brand that deployed the agent? The user who approved it? Currently, there is no global consensus on responsibility. The European Union’s AI Act provides some clarity for high-risk systems, but enforcement is still evolving. In the United States, fragmented regulations leave companies navigating a liability vacuum. Until clearer frameworks emerge, overdisclosure and caution may be the safest approach—though this can stifle innovation, especially for start-ups lacking extensive legal resources.

The ambiguity of accountability is not just theoretical; it has commercial, legal, and reputational implications. For example, if an AI agent books a trip that gets canceled, who is responsible? The travel site? The agent? A malfunctioning third-party plug-in? Mistakes by agents can lead to brand liability, regulatory scrutiny, and systemic risk. Companies integrating agents face the challenge of aligning decision-making autonomy with explainability. Consumers need to understand not only what an agent did but also why. Explainability is likely to become a consumer right, and auditable logs may soon be a regulatory requirement. Companies are preparing by incorporating permissions layers, identity mapping, and tiered trust levels—known as the “TRiSM” stack (trust, risk, and security management). But technical solutions alone are insufficient; the deeper issue is legitimacy.

Data ownership: A geopolitical challenge. Because AI agents operate on data, the issue of data sovereignty becomes increasingly geopolitical. Countries such as India and France are drawing firm lines with data localization efforts; Europe is seeing calls for “AI sovereignty.” This raises critical questions: If an agent processes EU citizen data via a US-based API, is it compliant? If it is trained on global data but acting locally, is it lawful? While AI councils, cross-border frameworks, and watchdog coalitions are emerging, global standardization is lacking. Until such standards exist, the risk is not just technical; it’s political. Regulations like Europe’s General Data Protection Regulation and the EU AI Act signify a shift in data management and control, with regions focusing on digital sovereignty or innovation and risk management. OpenAI’s “for countries” model, offering localized infrastructure and regulatory flexibility, highlights the challenge of scaling agentic platforms without breaching national data protection, tax, and consumer protection laws. What’s more, when consumers use agents powered by globally trained models, the question of legal protection arises, and the answer remains unclear. For agentic commerce to fulfill its promise, trust must be integral to its architecture. This will require transparent decision trails, fail-safes, reversibility, and regional customization of behavior and ethics to ensure agents align with user values. The combination of novelty and perceived risk often amplifies user concerns, making it essential to invest in clear communication, transparency, and the ability to safely test and explore AI systems.

Risk is not the opposite of opportunity

Finally, agentic commerce also introduces a novel kind of risk: the risk of the unknown. As agents learn to reason, improvise, and self-chain actions, emergent behaviors become possible. What is helpful today could be harmful tomorrow. Even minor shifts in how agents interpret intent could lead to subtle manipulation, adversarial exploits, or unintended amplification of bias. The question is no longer, “Is the agent safe today?” but rather, “Is the system prepared for what the agent could become?”

Every innovation, of course, brings exposure to new risks. Delivering on the promise of agentic commerce will not be about eliminating risk; it will mean learning to manage it faster than it evolves. For businesses, this means building not just for scale but for containment. For regulators, it means shifting from reaction to anticipation. And for consumers, it means demanding the same level of accountability from the systems acting on their behalf as they once demanded from people.

Chapter 5 | The challenges and opportunities of agentic commerce

The potential of agentic commerce is immense, with the power to fundamentally transform how businesses and consumers interact. However, this transformation is not without its hurdles.

The transition from static e-commerce to a more dynamic agentic commerce system unlocks transformative potential, redefining customer experiences and operational agility. To thrive in this era of autonomous digital commerce, businesses must act now, as AI is already changing the way we interact. Investing in agent-ready infrastructure—encompassing APIs, data interoperability, trust frameworks, and governance—is essential to staying competitive in a rapidly evolving landscape.

Lari Hämäläinen, senior partner and global gen AI and agentic services lead at QuantumBlack, AI by McKinsey

Progress toward capturing the projected $3 trillion to $5 trillion economic opportunity will depend heavily on how effectively these challenges are addressed. Companies must navigate the intricate task of integrating AI into existing systems, uphold rigorous standards of data privacy, and adapt to fast-evolving consumer expectations. To fully unlock the potential of agentic AI, organizations should address a series of critical strategic and operational questions across three key domains.

Capabilities, go-to-market strategy, and brand

The following questions can help organizations form the basis for a strong competitive position as they develop their agentic commerce strategies.

First-mover advantage. How can your business swiftly gain a first-mover advantage and build a defensible moat through strategic API development and AI partnerships? Identifying the prerequisites, such as robust tech infrastructure and a modular partnership ecosystem, will be essential for adapting to shifts in consumer behavior.

Scaling AI competitiveness. What strategies will enable your business to ramp up in the next two to three years to effectively compete with AI tools? With no single dominant AI player, businesses need to approach AI partnerships with adaptability in mind.

Proprietary versus partnership. To what extent should your business develop proprietary AI solutions versus exclusively partnering with AI developers? Consider the type of partners needed to build a future-proof ecosystem that creates a unique selling proposition in an agentic AI world.

Evolving work practices. How will your operations evolve in the era of AI? Transitioning from an isolated “AI team” to embedding AI across all teams, similar to the mobile shift, will be critical for success.

Brand differentiation. How can you differentiate and strengthen your brand in a landscape where AI increasingly handles customer interactions and decision-making? Understanding what brand identity looks like in a future dominated by AI and automation is crucial.

Concierge experience. How can your businesses create a unique and personalized concierge experience in an era where AI agents are central to customer interactions? Crafting a truly distinct concierge service through AI can drive customer satisfaction and loyalty and set the brand apart in a competitive market.

Monetization strategies

Agentic commerce potentially threatens existing revenue streams. The following questions can help leaders mitigate those disruptions—while they seek to capitalize on new opportunities the technology presents.

Innovative revenue models. As AI disintermediates traditional revenue streams like ad revenue, what innovative revenue models can your business create? Leveraging AI to offer value-added services, premium offerings, or new products that customers are willing to pay for is key.

Data monetization and personalization. What role can data monetization, AI-powered personalization, or subscription-based models play in generating new streams of revenue? Agentic AI can enable your business to capture an even broader share of the wallet.

Trust and risk

In addition to opening new opportunities, agentic commerce also presents novel risks, especially regarding customer trust and relationships. The following questions can help leaders understand the potential risks they might face.

Consumer trust. How can your business earn and preserve consumer trust when delegating decisions to autonomous agents? Clear communication, transparency, explainability, and human override features can help build confidence, but local culture influences both trust and adoption.

Trust transferability. Will consumers trust agents simply because of the brands deploying them? Understanding the limits of brand halo effects, and knowing when users need proof of safety, reliability, or explainability, will be essential.

Data sovereignty. How does the movement of data across borders impact user trust in agentic systems? Consumers and local governments might demand stronger assurances of local data handling, auditability, and ethical oversight.

Systemic risk. What guardrails are needed to prevent cascading failures across interconnected agents and ecosystems? Building resilient, modular architectures and fail-safe protocols will define the maturity of agentic infrastructure.

Adoption. Will agents ever be widely adopted in markets with low digital payment penetration or institutional trust? Adapting agent behaviors to local trust norms and offering layered human control will be essential for global scalability.

Admittedly, this is a lot to consider. But leaders who embrace this moment with vision and courage will find themselves at the forefront of a new era. They will see beyond the risks of disruption to the possibilities of reinvention. They will recognize that the future of commerce is not about replacing human ingenuity but amplifying it—with systems where AI and humans work in harmony to deliver great experiences and great value. Businesses that act now—investing in flexible architectures, experimenting boldly, rethinking their models, and aligning with their customers’ desires and preferences—can not only adapt to this new reality but shape it.