Skip to main content
Perspective:
Perspective
26 Jun 2026
25 minute read

Agentic commerce: The future of B2B commerce

Imagine a commercial landscape where AI buyer agents negotiate directly with AI seller agents, where sales and procurement transactions are executed in seconds, and where the friction points that have defined business-to-business (B2B) commerce for decades disappear. This is the potential of B2B agentic commerce—the use of AI agents to conduct business on behalf of enterprises across both the buyer and seller sides—and it is closer than most organizations realize.

Apurva Pangam (AP)
Gagan Mehra
Paul do Forno

Key takeaways

  • Buyers spend nearly 30% more with suppliers that deliver great buyer experiences, but there is a growing gap between how they both perceive the sales process.
  • This gap, combined with the rise of agentic AI, spells opportunity for sellers who move early to embrace B2B agentic commerce.
  • The future of B2B commerce is agent-to-agent commerce, where collaborative multiagent systems research, negotiate, buy, and sell on behalf of enterprises.
  • This article outlines the stages, risks, and readiness requirements for planning an effective B2B commerce strategy.

The timing is opportune. AI agents have matured dramatically over the past year, arriving just as enterprises are looking for new ways to navigate margin pressure, supply volatility, and labor constraints. Some B2B enterprises are already modernizing the core systems that agentic commerce will run on, creating a strategic window to build for an agent-ready future. Each of these pressures on its own would strain a business model. Together, they render "business as usual" increasingly untenable—and make the case for a faster, more agile, more autonomous operating model. The market also recognizes the imperative:

74%

Leaders who their organization will be using agentic AI at least moderately within two years

Source: Deloitte’s 2026 State of AI in the Enterprise report

$35 billion

Estimated value of the AI agent market by the end of the decade 

Source: Deloitte’s TMT Predictions 2026

In fact, Deloitte predicts AI agent market value could rise even higher to $45 billion if agents are strategically orchestrated and risks are appropriately mitigated.1

That realization surfaces a new set of defining questions for the future of B2B commerce: What is practical today? What happens as value chains become predominantly machine-to-machine? And how do enterprises navigate the shift from business as it was, to business as it will be?

The trajectory is clear: Near-term needs and long-term potential both point toward autonomous AI agents in commerce: agentic-driven negotiation at scale, closed working-capital loops, and direct agent-to-agent transactions across the enterprise. The question is no longer whether to move, but how quickly to operationalize agent-led business.

Where B2B agentic commerce stands today

Before examining where agent-to-agent commerce is headed, it is worth grounding the conversation in where B2B enterprises stand today. Deloitte's 2026 B2B commerce research2, based on surveys of more than 1,000 US suppliers and buyers, reveals a market that is ambitious but unevenly prepared. Three patterns stand out:
A perception gap between buyers and suppliers.

Seventy-two percent of suppliers said their sales processes were mostly or highly automated. Only 47% of buyers agreed; in fact, buyers were six times more likely than suppliers to describe B2B processes as mostly manual. In other words, internal automation has not yet translated into the external buyer experience.

A widening maturity divide.

Suppliers with high digital commerce maturity exceeded annual sales goals by 110% more than low-maturity competitors and are roughly five times more likely to use AI extensively. The distance between digitally mature suppliers and the rest is already substantial.

A buyer-supplier adoption asymmetry

Buyers are adopting agentic AI faster than their suppliers. Nearly 40% of B2B buyers already use agentic AI in purchasing — evaluating products, configuring orders, reviewing contracts, and benchmarking prices. Supplier adoption is trailing but poised to grow, with 24% of suppliers using agents in the sales process and 67% reporting that they plan to in the future.

The commercial stakes of these gaps are substantial. Suppliers estimate that, on average, 13% of sales bids are lost due to negative buyer experiences—that’s revenue left on the table because of friction that automation and agentic AI for sales are well-suited to resolve.

On the upside, positive buyer experiences drive an estimated 36% revenue uplift, and buyers spend nearly 30% more with suppliers that deliver them. Experience is a margin lever, not just a satisfaction metric.

B2B AI use cases in agentic commerce

The path toward a future where commerce is more autonomous relies on a singular capability: AI agents that can reason, act, and coordinate without continuous human direction.

As agents proliferate, the imperative is to move from thinking about them as stand-alone use cases (e.g., a sophisticated chatbot) to agentifying architecture and combining multiple agents to sell more directly and impactfully to customers. Agentic AI systems can set goals, perform multistep tasks, use tools and application programming interfaces (APIs), and coordinate with people or other agents—all within guardrails set by the enterprises that deploy them.

The shift is already beginning to show up across several B2B use cases, including:

  • Agentic customer engagement and service – Interpret customer intent, compare products, coordinate purchases or bookings, and resolve service requests through policy-driven actions.
  • Buyer intent and solution discovery – Interpret buyer needs, requirements, and account context to recommend relevant products, services, and solutions.
  • Transaction and service orchestration – Coordinate quotes, approvals, orders, fulfillment, and service requests through policy-driven workflows and escalation rules.
  • Autonomous commercial negotiation and deal governance – Generate proposals, apply pricing and contract terms, and negotiate within governance guardrails while escalating exceptions for approval.
  • Intelligent procurement and supplier orchestration – Evaluate suppliers across cost, availability, compliance, and risk to automatically trigger compliant purchasing decisions.
  • Autonomous payments and financial settlement – Select optimal payment methods, execute transactions within risk parameters, and reconcile invoices and purchase records automatically.
  • Predictive fulfillment and inventory optimization – Monitor demand signals, dynamically rebalance inventory across locations, and recommend substitutions to maintain service levels.
  • AI-driven sales acceleration and pipeline management – Detect buyer intent, surface qualified opportunities, and coordinate personalized outreach to accelerate deal progression.

This requires a transformational approach to enterprise strategy, technology investments, change management, risk mitigation, and numerous other business factors. It’s not just about adopting new technology; it’s a fundamental reimagination of B2B commerce and all the changes to strategy and operations that come with it.

The trajectory toward AI agents in commerce

Over the past year, agentic AI technology has matured from compelling one-off experiments into systems that can connect to tools and data, complete complex workflows, and orchestrate with other agents—and the pace is accelerating.

In their nascent forms, agents assist humans inside existing workflows—this is where most B2B enterprises sit today. In their fullest form, agents from different organizations transact directly with one another. Business-to-consumer (B2C) is proving this concept and paving the way for B2B to follow: Leading B2C enterprises are using agents to reach customers across channels, present offers dynamically, and reshape the shopping experience—with early signals that customer behavior is starting to shift.3  

Let’s say a hospital needs to order stents. Today, procurement teams compare offers from four suppliers, weigh pricing against clinical requirements, and route decisions through approval chains.

In the near-term, an agent embedded in the hospital's e-procurement platform tracks the patterns behind those decisions—where the hospital consistently optimizes for cost, prioritizes delivery speed, accepts premium pricing for specific device characteristics, etc.—and surfaces recommendations.

In a more mature form, the hospital's procurement agent and the suppliers' sales agents engage directly to configure pricing, quoting, and negotiation in real time. A human retains final purchasing authority.

In the fully autonomous form, agents independently procure the right products at the right time for the best price. Humans move from operators to overseers, setting policies, defining guardrails, and intervening only where judgment is required.

Open protocols are beginning to define the foundation for both B2C and B2B agentic commerce. Universal Commerce Protocol (UCP), developed by Google, brings together major commerce, retail, and technology players around a shared approach to product discovery, checkout, order management, and post-purchase experiences. While its early momentum is strongest in consumer and retail commerce, UCP’s design could extend to B2B commerce workflows over time.

Alongside UCP, Agent2Agent (A2A) protocol provides an interoperability layer for agents to discover, authenticate, and collaborate across platforms, while Agent Payments Protocol (AP2) extends into trusted, authorized transactions. Together, these standards signal that the infrastructure for agent-to-agent commerce is moving from concept to open ecosystem development. That said, the protocol landscape is still forming—especially for B2B, where protocols are still in early stages and no clear leader has emerged. 

Signs that B2B agentic commerce is evolving are visible in the data: The average supplier now supports 4.7 commerce channels, up from 3.4 two years ago.4 Channels include direct online commerce, direct sales, agentic (AI agents), B2B marketplaces, e-procurement, dealer portals, CPQ, and sales-assisted.

Electronic Data Interchange (EDI), the standardized document-exchange format that has carried B2B transactions for more than 40 years, is losing ground: 92% of buyers currently using it plan to shift partially or fully to other channels. That migration points toward the more flexible, API-ready rails that agents can plug into. 

Beneath the surface of channel strategy, a deeper shift is underway: Nearly 90% of B2B suppliers are currently upgrading or preparing to upgrade their ERP systems, creating an opportunity to modernize the front-to-back integration that B2B agentic commerce will run on. The choice is whether to treat ERP modernization and agentic B2B commerce strategy as one integrated program or two competing ones.

Today, most B2B businesses struggle to understand where they are in the agentic commerce trajectory, what the next stage requires, or what progress looks like. Let’s explore how the future of B2B commerce unfolds within the enterprise.

Stages of B2B agentic commerce transformation

Achieving agent-to-agent interaction in commerce will take time due to the work and degree of change involved. There are likely to be stages of agentic maturity across the journey, with different sectors aspiring to different end states depending on readiness, risk tolerance, regulations, geography, and supply chain complexity. Recognizing these nuances, we can illustrate a common B2B agentic maturity pathway through an intelligent procurement use case, followed by the underlying sample architecture.

The stages below describe the progression of B2B agentic commerce via “workflow autonomy,” or how much of the buying and selling process agents own end to end. This is a counterpart to the consumer-journey progression outlined in Deloitte's Agentic commerce: Redefining retail economics thinking5, which maps the B2C path from assisted discovery through autonomous shopping to agent-to-agent commerce. Both progressions converge on the same end state: agentic systems from different organizations transacting directly with one another.

Agents are introduced to assist in specific tasks, with the intent of solving discrete business problems and reducing manual effort. These agents are not fully autonomous but instead sit alongside people in existing workflows.

Example: Agents assist in discrete procurement tasks (e.g., supplier discovery), surface insights, summarize information, and automate repetitive steps. Procurement professionals remain the primary workflow operators.

Multiple agents are connected to automate workflow segments that have been reimagined to capitalize on agentic capabilities. While value is increasing, the agents are primarily used internally and are not yet connected to the broader ecosystem or external people or agents.

Example: Agents coordinate multiple procurement tasks across a sourcing workflow, including launching requests for quotation (RFQs), evaluating supplier responses, scoring proposals, generating contracts, preparing purchase orders, and orchestrating sourcing internally. Humans define strategies, set criteria, manage exceptions, and intervene in unusual scenarios.

Self-optimizing agents that improve overtime are orchestrated to pursue business goals such as efficiency and profitability, with human roles and responsibilities shifting from execution to oversight. Despite high maturity, agentic systems are not yet open to external agent-to-agent interactions.

Example: Coordinated multiagent systems optimize for procurement goals, monitoring markets, initiating sourcing events, renegotiating contracts, and learning from past outcomes. Procurement professionals set objectives and constraints, oversee strategies, and intervene in only the most complex negotiations, focusing instead on ethical use, regulations, and supplier relationships.

This is the end state where agentic systems from different organizations transact directly in fully autonomous commercial activity. This will require not only enterprise maturity but also broader ecosystem maturity that incorporates standards and external integration.

Example: Procurement agents interact directly with supplier agents to negotiate pricing and contracts, execute transactions, and coordinate fulfillment in real time across organizations. Humans are elevated to oversee fully automated commercial exchanges, defining policies, guardrails, and strategic priorities.

The underlying architecture

The right technical architecture is essential to the advancement described above. Agentic architecture integrates existing systems to enable autonomous commerce activities across a given enterprise and its ecosystem. For a closer look, explore our deep dive on the principles of agentic architecture and design.6
Interaction layer

Users, processes, and platforms (e.g., ERP, CRM) interact with the multiagent system to initiate or continue actions.

Commerce layer

Controlled workflows, possibly with a human in the loop, for commercial tasks (e.g., pricing and ordering).

Intelligence layer

Where agents are created, managed, deployed and optimized, it includes large language models, a prompt registry, and an agent factory for role-specific agents, as well as internal and external data sources and analytics tools (e.g., demand planning).

Agent operations layer

Outputs and metrics are monitored to track and validate that agents are performing as intended.

Reimagination required for the road ahead

How quickly organizations move from early use cases to enterprise-wide transformation depends in part on where the enterprise sits on the maturity spectrum and what they do next.

Where they are:
Agentic AI experiments are ticking up: 85% of companies anticipate customizing agents to automate aspects of the business, according to Deloitte’s most recent State of AI in the Enterprise report.7 While experimentation is valuable and important, it’s not an end in and of itself. Deloitte’s Tech Trends 20268 found only 11% of enterprises reported using agents in production, perhaps owing to the fact that 42% of organizations are still developing an agentic strategy and 35% have no strategy at all.

Where to next:
To reach the boldest future in agentic or autonomous commerce, businesses will need to contend with the complex task of industrializing multiagent systems. This goes beyond basic agent customization and compelling use cases. It entails a reimagination of workflows and responsibilities. It also includes activities such as ensuring clean master data, aligning a multicloud environment, developing APIs for task execution, securing data flows, attention to risk mitigation, and AI governance, and navigating the perennial questions around whether to buy or build agentic solutions.

The priority is to redesign processes with agents in mind, rather than layering agents onto existing workflows.

Because agentic commerce will touch or impact every part of the enterprise, the entire organization needs to be conditioned for the fully autonomous B2B future. With changes to brand strategy, investments, talent development, and more, the scale of transformation ahead requires significant, enterprise-wide work and change. The priority is to redesign processes with agents in mind, rather than layering agents onto existing workflows. Simply automating current processes risks carrying forward the manual constraints of the legacy model.

Agent-to-agent commerce: Risks and the cost of standing still

All technology presents risks that need to be mitigated, and agentic AI raises both familiar and new risks. Technology trust is essential, but regulatory compliance; customer privacy and security; and adapting roles, responsibilities, and incentives are just as important.

As with any AI solution, accuracy, reliability, and transparency are paramount. Agentic systems also run the risk of inaccuracies (or hallucinations) that propagate from machine to machine, degrading outputs. And because enterprise and operational data are the fuel that AI agents run on, data security is a priority in terms of designing agents to limit or remove data vulnerabilities.

In the B2B agent-to-agent commerce context, the risk landscape shifts as agent capabilities mature, presenting distinct challenges at different stages of adoption:

In the near term, risk considerations center on operational integration. Enterprises should establish oversight mechanisms as agents handle commitments and agreements with greater speed and autonomy.

  • Governance: To transform quickly and confidently, agentic systems must be governed. At this early stage, legal and regulatory frameworks are still adapting to agent-executed transactions, creating the need for clear contract structures and compliance protocols.
  • Architectural exposure: Security architecture becomes critical as agents integrate into enterprise infrastructure with access to procurement systems, supplier relationships, and commercial data.
  • Identity verification and payment authorization: Agents need robust mechanisms to confirm they're transacting with legitimate counterparties and to execute payments within guardrails. Trust in agent-to-agent transactions requires new authentication and fraud prevention measures, alongside financial controls and reporting frameworks built for machine-initiated activity.

Longer term, risk considerations become more strategic. Organizations will need to define which information agents reveal through interaction patterns and negotiation strategies.

  • Sustainable expertise: Maintaining institutional expertise alongside agent capabilities becomes essential. Organizations will need internal teams that can evaluate agent decisions, intervene when context requires human judgment, and step in to operate effectively if systems fail.
  • Ecosystem architecture: Technology architecture decisions become increasingly mission critical, as agent platforms become deeply embedded in operations, making initial vendor and design choices more consequential.
  • Competition: Perhaps most significantly, ecosystem dynamics will reshape competitive advantage as agent-to-agent interactions become standard, requiring enterprises to understand how their agents perform relative to increasingly sophisticated counterparties.

These risks and mitigation steps, however, should be weighed against a potentially greater risk: falling behind. According to our State of AI report,9 about 20% of companies report having a mature governance model for autonomous agents. These companies are laying the foundation for governance of agentic commerce and multiagent systems. What about the other 80%?

The potential value of B2B agentic commerce and the pace of industry adoption mean that "wait and see" strategies carry real competitive consequences. As agent-to-agent interactions become the norm, enterprises that delay transformation may find themselves competing with diminished efficiency, slower response times, and reduced market access.

In the future, agentic commerce will simply become commerce, and the question won't be whether to adapt but how quickly and thoughtfully organizations can do so while managing the risks inherent in any significant technology transition.

Developing a B2B automation strategy: Assessing readiness

Some organizations are taking a more cautious view of technology investments, which is understandable. The degree of change required to reach the brightest future in agentic commerce can be daunting. When designing a B2B automation strategy, leaders and stakeholders can start by assessing and improving readiness across five broad categories.

Strategy is primarily the domain of the CEO, CSO, and CFO. Together, they set the agentic commerce vision, lend executive sponsorship, and establish value realization priorities.

The CDO and CIO face the challenge of enabling machine-readable, timely, enriched, and verifiable data across products, pricing, inventory, customers, vendors, and policies. As agents take on more complex commerce tasks, they will need access to detailed, nuanced data from validated and trusted sources to reason, recommend, and act effectively.

To lay the technology foundation for agentic commerce, the CIO, CTO, and CISO focus on delivering scalable, low-latency integration platforms supported by strong technology and integrator partnerships.

In preparing the enterprise for transformation and adoption, the COO, CMO, and CRO collaborate to redesign ownership, workflows, governance, measurement, skills, and change management.

Moving to new agent-first workflows and capabilities, the CCO, CMO, and CRO contend with preparing to market and sell across sales platforms.

Each readiness area commands its own competencies and questions, and it can be helpful to work with an adviser to define where your organization is today and which activities and investments are needed.

The future of B2B commerce

Agentic commerce is becoming a key agenda item for the board, and leaders across industries are moving forward. There is urgency to change among organizations of all levels of readiness. By beginning this journey today, enterprises can begin to capture the benefits of AI agents for sales and lead in the future of B2B.

Endnotes

1. Duncan Stewart, Jeff Loucks, and Paul Lee, TMT Predictions 2026, Deloitte, November 18, 2025.
2. Paul do Forno, Apurva Pangam, and Pooja Warudkar, “Accelerating sales growth through B2B digital commerce,” Deloitte, January 2026.
3. Vivek Pandya, “Adobe: AI-driven traffic surges across industries with retail experiencing biggest gains,” Adobe for Business Blog, January 12, 2026.
4. do Forno et al., “Accelerating sales growth through B2B digital commerce.”
5. Saurabh Vijayvergia, Brian McCarthy, and Roland Ehigiamusoe, “Agentic commerce: Redefining retail economics,” Deloitte, 2026.
6. Prakul Sharma et al., “The cognitive leap: How to reimagine work with AI agents,” Deloitte, December 2024.
7. Rowan et al., State of AI in the Enterprise 2026: The untapped edge.
8. Kelly Raskovich (ed.), Tech Trends 2026, Deloitte, December 10, 2025.
9. Rowan et al., State of AI in the Enterprise 2026: The untapped edge.

Get in touch

Apurva Pangam (AP)

Apurva Pangam (AP)

Principal, M.S., Human-Computer Interaction Design | B.E
Gagan Mehra

Gagan Mehra

Managing Director
+1 510 290 5780
Paul do Forno

Paul do Forno

Managing Director, BMath Joint Honours Math; Business Admin & CS | University of Waterloo

Did you find this useful?

Yes
No

Thanks for your feedback

Your feedback is important to us

To tell us what you think, please update your settings to accept analytics and performance cookies.

Need help updating cookies?

Related insights

Deloitte AI Institute - AI Insights

The Deloitte AI Institute helps organizations transform through cutting-edge AI insights and innovation by bringing together the brightest minds in AI services.
Collection

Agentic commerce: AI shopping agents guide 2025

Discover how agentic commerce and AI shopping agents are transforming retail. Learn how autonomous AI drives transactions, optimization strategies, and ROI.
Perspective