The Infrastructure Layer Nobody Is Talking About
On 9 July 2026, the International Telecommunication Union, the United Nations agency for digital technologies, announced a Focus Group to develop frameworks for AI agent identity, trustworthiness and human control. The ITU specifically highlighted impersonation, unauthorised decisions, financial transactions and critical infrastructure as the primary risk areas. Its first meeting is scheduled for Paris in November 2026.
The announcement received a fraction of the coverage given to GPT-5.6 and ChatGPT Work. That is a mistake. The ITU initiative points to the infrastructure problem that will determine whether agentic commerce can scale at all.
The marketing conversation about AI agents is currently focused on capability: can the agent research accounts, draft content, buy media, negotiate pricing, send communications? The harder commercial question is not whether the agent can do these things. It is whether the systems it interacts with can verify that the agent is authentic, authorised and acting within a defined mandate.
The Agent-Identity Gap
Human identity infrastructure was built for humans. A login credential represents a person. An API key represents an application. Neither adequately represents an autonomous agent acting on behalf of an organisation within a defined scope.
When a human employee buys media, the platform knows who they are, which organisation they represent and, through approval workflows, what they are authorised to spend. When an AI agent buys media, the platform currently has no reliable way to verify the agent's identity, its organisational affiliation, its spending authority or whether its actions are within its mandate.
This creates several problems that become more serious as agentic commerce scales. Impersonation risk arises when an agent claiming to represent an organisation may be acting without authorisation, may have been compromised, or may be a malicious actor impersonating a legitimate agent. Authentication risk means that even if an agent's identity is verified, its authority may not be — a verified agent can still act outside its mandate if there is no machine-readable representation of what it is permitted to do. Accountability risk means that when an agent makes an error, exceeds its authority or produces an output with commercial or legal consequences, the audit trail needed to assign liability and remediate the action may not exist.
For B2B marketing operations, these risks are not theoretical. They are already present in any organisation using AI agents to send communications, buy media, update CRM records or publish content.
The Delegated-Authority Problem
Authentication is necessary but not sufficient. An authenticated agent still needs explicit limits on spend, data access, communications and commitments.
Consider a B2B marketing team using an AI agent for account-based marketing. The agent has been given access to the CRM, the email platform, the LinkedIn advertising account and the content management system. It is authenticated. But what is it authorised to do?
Can it send emails to prospects without human review? Can it increase campaign budgets when it identifies a high-performing audience? Can it publish content to the company blog? Can it update account records based on its own research? Can it make commitments in communications that imply a commercial relationship?
Each of these questions requires an explicit answer before the agent is deployed. The answers need to be machine-readable so that the agent's actions can be constrained at the platform level, not just at the instruction level. Instructions can be misinterpreted or overridden by the model. Platform-level constraints cannot.
This is the delegated-authority problem. It is not a technology problem. It is a governance design problem that organisations need to solve before they scale agentic marketing operations.
The Commercial Trust Layer
The ITU's initiative is directionally important because it recognises that agentic commerce requires a new infrastructure layer. The analogy is instructive: ecommerce could not scale until payment infrastructure was reliable, standardised and trusted. Buyers needed to know that their payment would be processed correctly, that their data was protected and that disputes could be resolved. Sellers needed to know that payments were authentic and that chargebacks were manageable.
Agentic commerce needs an equivalent trust layer for agent identity, authority and accountability. Businesses interacting with AI agents need to know which organisation owns the agent, which person authorised it, what it is permitted to do, whether its actions can be audited and reversed, and who carries liability when it acts outside its mandate.
That verification layer may become as important to agentic commerce as payment infrastructure was to ecommerce. The organisations that build it first will have a structural advantage in deploying agents that other systems will trust and interact with.
What B2B Marketing Teams Should Implement Now
The ITU framework will not be finalised until 2027 at the earliest. That does not mean organisations should wait. The governance controls that make agent deployment commercially safe today are the same controls that will be required by future standards.
Start with scope definition. Before any agent is given access to external systems, document explicitly what it can and cannot do. Define spending limits, data access permissions, communication authorities and commitment scope. This documentation is the foundation of delegated authority.
Build audit trails. Every action taken by an agent should be logged with sufficient detail to reconstruct what happened, why, and what the consequences were. This is essential for liability management, quality improvement and regulatory compliance.
Establish human review checkpoints. Identify the actions that have significant commercial or legal consequences and require human review before execution. These checkpoints are not a sign of distrust in the agent. They are the governance layer that makes delegation commercially reliable.
Test your liability exposure. Review your current agent deployments against the question: if this agent takes an action outside its intended scope, who is liable and what is the remediation path? If you cannot answer that question clearly, the governance framework is incomplete.
Integrated.Social's Agentic AI services include governance framework design as a core component of every agentic programme. The agent is only as valuable as the governance that makes its actions commercially reliable.
About the Author
Modi Elnadi is the Founder and Director of Marketing and AI Growth at Integrated.Social, a London-based B2B AI marketing agency specialising in Agentic AI lead generation, Answer Engine Optimisation, and AI-native website builds. Modi has been building performance marketing systems since 2014, working with FinTech, SaaS, and B2B brands across the UK and USA. He advises enterprise marketing leaders on AI governance, agentic commerce frameworks and the accountability structures that make autonomous agent deployment commercially safe. Connect on LinkedIn or explore Integrated.Social's AI governance and Agentic AI services.
