The discourse around agentic commerce has been dominated by the consumer narrative: the AI shopping agent that buys coffee, books hotels, and manages household supplies on behalf of an individual human. This narrative is important - it captures the imagination and illustrates the fundamental design challenges of delegation, trust, and absent-state interaction. But it is not the whole story. The larger, faster-growing, and arguably more consequential domain of agentic commerce is not consumer-facing. It is business-to-business.
B2B agentic commerce is the domain where autonomous AI agents represent organisations - not individual consumers - in procurement, negotiation, supplier evaluation, contract management, and fulfilment. It is the domain where a procurement agent negotiates with a supplier’s sales agent, where a logistics agent coordinates with a warehouse agent, where a compliance agent verifies that every transaction in a complex supply chain meets regulatory requirements. It is the domain where the machine customer is not a personal shopping assistant but a corporate procurement system with delegated authority to commit millions in organisational spending.
This essay examines why B2B agentic commerce presents design challenges that are fundamentally different from - and in many ways more complex than - the consumer-facing scenarios that have dominated the Agentic Experience Design discourse. The multi-principal problem, the delegation chain challenge, the inter-organisational trust architecture, and the compliance imperative all demand design responses that extend the AXD framework into new territory.
01 - The B2B Acceleration
B2B commerce is accelerating toward agent-mediated transactions faster than consumer commerce, for structural reasons. B2B procurement is already heavily systematised: enterprise resource planning systems, procurement platforms, supplier management databases, and automated approval workflows have been standard for decades. The infrastructure for agent-mediated commerce already exists in B2B - what is changing is the intelligence and autonomy of the agents that operate within it.
The economic incentive is also stronger. A consumer shopping agent that saves a customer £5 on coffee beans is a convenience. A procurement agent that saves an enterprise £5 million on raw materials is a competitive advantage. The return on investment for B2B agentic commerce is orders of magnitude higher than for B2C, which is why enterprise adoption of AI procurement agents is outpacing consumer adoption of AI shopping assistants.
Gartner projects that by 2028, 15% of day-to-day work decisions will be made autonomously through agentic AI, with B2B procurement identified as a primary adoption domain. McKinsey estimates that AI-driven procurement optimisation can reduce costs by 5-10% across categories. These projections are driving rapid investment in B2B agentic systems - but the design frameworks for governing these systems are lagging far behind the technology.
02 - What B2B Agentic Commerce Actually Is
B2B agentic commerce is not simply “B2C agentic shopping but for businesses.” It is a structurally different domain with its own design challenges. In B2C agentic shopping, the relationship is relatively simple: one human delegates to one agent, which interacts with merchants on the human’s behalf. The principal is singular, the delegation is direct, and the trust relationship is bilateral.
In B2B agentic commerce, every element of this relationship is multiplied. The principal is not a single human but an organisation - a collective entity with multiple stakeholders, competing objectives, and hierarchical authority structures. The delegation is not direct but cascaded - authority flows through organisational hierarchies, with different approval thresholds and constraints at each level. The trust relationship is not bilateral but networked - multiple agents representing multiple organisations interact in complex supply chains where trust must be established, verified, and maintained across organisational boundaries.
Consider a concrete example. A manufacturing company’s procurement agent is tasked with sourcing raw materials. The agent must satisfy the engineering team’s quality specifications, the finance team’s budget constraints, the compliance team’s regulatory requirements, and the sustainability team’s environmental standards. It must evaluate multiple suppliers, each represented by their own sales agents with their own authority levels and negotiation constraints. It must negotiate terms, verify compliance, arrange logistics, and manage contracts - all within a governance framework that requires audit trails, approval workflows, and escalation protocols. This is not a shopping trip. It is an orchestration challenge.
03 - The Multi-Principal Problem
The most distinctive design challenge of B2B agentic commerce is the multi-principal problem. In B2C, the agent has one principal: the customer. The agent’s objective function is singular - serve the customer’s interests as specified in the delegation. In B2B, the agent has multiple principals with potentially conflicting objectives.
The procurement team wants the lowest price. The engineering team wants the highest quality. The compliance team wants regulatory conformity. The finance team wants predictable cash flow. The sustainability team wants ethical sourcing. Each of these principals has different objectives, different authority levels, and different risk tolerances. The procurement agent must navigate these competing demands within a coherent delegation framework - a challenge that has no parallel in B2C agentic shopping.
The AXD approach to the multi-principal problem is hierarchical outcome specification - a delegation architecture in which each principal specifies their requirements as constraints within a priority hierarchy. The agent does not try to satisfy all principals equally. Instead, it operates within a structured priority framework: compliance requirements are non-negotiable constraints, quality specifications are weighted objectives, budget limits are hard boundaries, and sustainability preferences are optimisation criteria. The hierarchy is specified by the organisation’s governance structure, not by the agent - ensuring that the agent’s behaviour reflects institutional priorities rather than algorithmic preferences.
04 - Trust Architecture Between Organisations
Trust architecture in B2B agentic commerce operates at a different level than in B2C. In B2C, trust is personal - the customer trusts (or does not trust) the agent based on their individual experience. In B2B, trust is institutional - Organisation A’s agent trusts Organisation B’s agent based on organisational reputation, contractual relationships, compliance history, and verified credentials.
Inter-organisational trust architecture requires three elements that are absent from B2C trust design. First, credential verification: Organisation A’s agent must be able to verify that Organisation B’s agent is authorised to represent Organisation B, that it has the authority to commit to specific terms, and that its declarations are backed by organisational accountability. Second, contractual integration: the trust architecture must integrate with existing contractual frameworks - master service agreements, service level agreements, payment terms, and liability provisions - so that agent-negotiated terms are legally binding and enforceable. Third, institutional memory: the trust architecture must maintain a history of inter-organisational interactions that informs future trust calibration - not just “did this agent perform well?” but “has this organisation been a reliable partner over time?”
These requirements create a trust architecture that is fundamentally more complex than B2C trust design. The AXD Institute is developing frameworks specifically for inter-organisational trust architecture - frameworks that address credential verification, contractual integration, and institutional memory as first-class design requirements rather than afterthoughts.
05 - Delegation Chains and Authority Cascades
In B2C agentic commerce, delegation is a single act: the customer grants authority to the agent. In B2B, delegation is a chain - authority cascades through organisational hierarchies, with each level adding constraints, conditions, and approval requirements.
A typical B2B delegation chain might look like this: the CFO authorises the procurement department to spend up to £10 million annually on raw materials. The procurement director delegates to a category manager with authority up to £500,000 per transaction. The category manager configures a procurement agent with authority to negotiate and commit up to £50,000 per order, with automatic escalation for larger amounts. The procurement agent then interacts with supplier agents, each of which has its own delegation chain from its own organisation.
The design challenge is ensuring that authority is preserved, constrained, and auditable at every level of the chain. The procurement agent must not exceed its delegated authority. It must escalate appropriately when a transaction exceeds its threshold. It must maintain an audit trail that shows exactly what authority it exercised and why. And it must do all of this in real-time, during negotiations with supplier agents that are operating under their own delegation chains with their own constraints and escalation protocols.
06 - The Procurement Agent and the Sales Agent
The most distinctive interaction in B2B agentic commerce is the agent-to-agent negotiation. When a procurement agent negotiates with a supplier’s sales agent, both parties are autonomous systems acting within delegated authority. Neither is a human. Both are constrained by their respective organisations’ delegation frameworks. Both must make real-time decisions about trade-offs, concessions, and commitments.
This creates a market dynamic that has no precedent in human commerce. Human negotiation relies on social cues, emotional intelligence, relationship building, and strategic ambiguity. Agent-to-agent negotiation operates on structured protocols, verifiable declarations, algorithmic evaluation, and explicit constraint communication. The “negotiation” is not a conversation but a structured exchange of proposals, counter-proposals, and constraint declarations, evaluated against each agent’s outcome specifications and authority boundaries.
The AXD Institute identifies three design requirements for agent-to-agent negotiation. First, protocol standardisation: agents from different organisations must be able to communicate using shared protocols - such as those tracked in the AXD Institute's Protocol Tracker - that enable structured proposal exchange, constraint declaration, and agreement verification. Second, authority transparency: each agent must be able to declare its authority level and constraints, so the other agent can calibrate its proposals accordingly. Third, outcome alignment: the negotiation protocol must enable both agents to identify mutually beneficial outcomes within their respective constraint sets - not through strategic deception but through structured optimisation.
07 - Compliance, Governance, and the Audit Trail
B2B commerce operates within regulatory frameworks that constrain agent autonomy in ways that B2C does not. Procurement regulations, anti-bribery laws, sanctions compliance, data protection requirements, and industry-specific governance standards all impose constraints on what agents can do, how they can do it, and what records they must maintain.
The audit trail is the critical design element. Every B2B agentic transaction must produce a complete, verifiable record of what the agent did, what authority it exercised, what alternatives it considered, what trade-offs it made, and what compliance checks it performed. This audit trail must be structured for both human review (by compliance officers, auditors, and governance teams) and machine processing (by compliance agents, regulatory reporting systems, and automated audit tools).
The AXD framework addresses this through what the Institute calls governance-embedded delegation - a design approach in which compliance and governance requirements are not external constraints applied after the fact but structural properties of the delegation itself. The procurement agent does not first negotiate and then check compliance. Compliance is embedded in its delegation: the agent cannot negotiate terms that violate regulatory requirements, because those requirements are encoded as non-negotiable constraints in its authority framework. This approach transforms compliance from a post-hoc audit function into a real-time design property.
08 - Designing the Agent-to-Agent Marketplace
The ultimate expression of B2B agentic commerce is the agent-to-agent marketplace - a commercial environment in which autonomous agents representing different organisations discover each other, evaluate each other’s offerings, negotiate terms, and execute transactions with minimal human involvement. This is not a theoretical construct. It is the logical endpoint of current trends in enterprise AI adoption, and it is approaching faster than most organisations realise.
Designing the agent-to-agent marketplace requires extending every element of the AXD framework. Signal Clarity must address how organisations make their offerings legible to other organisations’ agents. Trust architecture must address how inter-organisational trust is formed, verified, and maintained at machine speed. Delegation design must address how authority cascades through organisational hierarchies and across organisational boundaries. And the Five Pillars of AXD Readiness must be extended to assess organisational readiness for agent-to-agent commerce, not just agent-to-human commerce.
The organisations that will thrive in the agent-to-agent marketplace are those that invest now in the trust architecture, delegation frameworks, and legibility infrastructure that agent-mediated B2B commerce requires. The organisations that wait will find themselves in a marketplace where their competitors’ agents are already negotiating, transacting, and building institutional relationships - while their own procurement teams are still reviewing spreadsheets.
B2B agentic commerce is not a future scenario. It is the present trajectory of enterprise AI adoption. The AXD Institute exists to provide the design frameworks, the conceptual vocabulary, and the practical tools that organisations need to navigate this transition. The design of agent-to-agent commerce is not a technical challenge. It is a trust architecture challenge - and trust architecture is the defining competence of the agentic age.