Last month, Forrester released its 2026 report examining the pace and direction of AI adoption within ad agencies.
It pointed to the technology being extremely pervasive, with 9 out of 10 agencies reporting to use generative AI and a further half using the technology for marketing campaigns. However, if we dig deeper, the report also found that the primary motivation in 83% of GenAI use cases and 63% of AI agent initiatives was to produce creative content, craft media strategies and improve general internal productivity.
The focus of using AI for ideation and content creation overlooks another high-impact use case.
AI agents are well suited to handling digital media buying and selling. Unlike creative work, where there is always an element of subjectivity and originality is highly prized, programmatic advertising is a data-driven process with clear-cut objectives and boundaries.
AI agents can query inventory and negotiate media deals in milliseconds in a way that supersedes any human agents. However, automated media buying is still largely in a more nascent stage of adoption due to concerns around commercial trust and risk control.
Letting AI make decisions without proper controls in place is a clear commercial liability that risks wasted budget and puts credit and pricing agreements at risk. Further, given the speed at which AI agents can make a decision, actions that have undesirable consequences for the business can rapidly snowball.
The current industry standard is to run post-campaign audits to identify issues and make adjustments. However, this approach will fail to provide robust risk controls due to the sheer speed of AI agents. Actions that have undesirable consequences for the business can rapidly snowball if reviews are left to a post-campaign audit.
Letting AI operate without commercial governance is a clear liability, but the ad industry needs a solution that doesn’t throttle the speed that makes AI agents a valuable tool.
Instead, the speed of AI agents demands an equally powerful solution for governance that provides structural protection across large ad agencies and media enterprises.
Speed, governance and guardrails
Unlocking true velocity in automated media buying isn’t a technical challenge; it’s a governance problem. Giving AI buying agents spending authority without real-time financial and contractual checks creates unmanaged liability.
Real commercial risks associated with letting AI agents buy inventory without real-time financial checks in place include overspend, “zombie RFPs,” unfunded automated queries, and even broken contracted rates.
Speed is useless if every fast transaction requires a slow, manual post-campaign audit to catch errors.
To scale autonomous commerce, the advertising industry must shift from post-campaign audits to pre-flight commercial governance, enforcing credit limits, and contracted pricing at the exact millisecond a deal clears.
Agentic AI for media buying has the potential to become one of the most valuable use cases for the ad industry, but its future won’t be viable without an intelligent way to manage governance at scale.
Rethinking campaign audits for the AI era
For years, the digital advertising industry has relied on a reactive approach to commercial governance.
Brands and agencies typically audit campaigns only after they have run, checking for overspending, incorrect contracted rates, and credit limit breaches weeks after media has been delivered.
While this audit process does identify issues and allow for adjustments to be made to the way future campaigns are structured, it can’t correct actions that have already taken place and money has already been spent.
Despite this, the retrospective approach has prevailed across agencies because executives have still been in the driving seat with these campaigns. This means that decisions are made with strategic context and commercial considerations, reducing the likelihood of serious risks to the business.
However, AI agents can negotiate deals in a fraction of a second. The post-audit approach means a simple error can snowball rapidly.
However, the way that buying and selling is managed across the industry means that this goes much deeper than the risk of a campaign overspend.
Agency CFOs and procurement teams spend years negotiating annual framework agreements, securing netto-netto pricing, volume rebates, and preferred commercial terms with media owners.
A widespread concern is that autonomous buying systems could inadvertently bypass these agreements, routing spend through channels that default to list prices and undermining carefully negotiated commercial relationships.
These serious commercial consequences that could occur with automated media buying have acted as a major barrier to adoption.
It’s unrealistic to govern AI buyer agents with post-campaign audits or vague prompts, but this doesn’t mean the technology isn’t viable for this use case. Instead, these agents depend on hard, real-time commercial rules built directly into the transaction layer.
This is why AI buying agents cannot be governed by post-campaign audits or vague prompts; they require hard, real-time commercial rules built directly into the transaction layer. Automated buying should never be a case of relinquishing control and letting AI agents run wild with campaigns. Instead, human-in-the-loop safety nets and equally powerful technology to manage governance at scale are critical pieces of the puzzle.
Scaling automation without losing control of commercial governance goes hand in hand with a guardrail engine that offers a structural safety net and human-in-the-loop controls that breed trust.
How guardrail engines protect commercial trust in AI
One of the biggest barriers to AI-powered media buying isn’t the technology—it’s commercial trust.
For AI agents to be viewed as a reliable tool to support media buying, agency leaders need watertight guarantees that automated AI won’t route around their annual framework agreements, force them to buy at list prices or lose their hard-won discounted terms.
That requires intelligent commercial infrastructure capable of recognising the buying entity in real time, automatically applying the agency’s contracted pricing, and ensuring every autonomous transaction contributes to existing annual spend commitments. AI should optimise within established commercial frameworks, not around them. By preserving negotiated terms and protecting agency economics, autonomous buying becomes an extension of existing business practices rather than a threat to them.
To achieve this, post-audit campaigns need to be replaced with pre-flight commercial governance where commercial controls are applied in real time, at the precise moment a transaction is cleared and before payment is authorized. This means validating credit limits and verifying contracted pricing before spend is committed, ensuring every transaction complies with agreed commercial and governance standards.
Next, leaders need to find a way to govern AI with human-in-the-loop controls that don’t result in a bottleneck.
The most effective AI systems combine speed with accountability by embedding human oversight where it matters most. Routine transactions can be executed automatically, while high-value, unusual, or policy-sensitive purchases are intelligently escalated to human approvers before completion. This human-in-the-loop model provides the reassurance that commercial judgement remains central to the process, even as AI accelerates execution.
ADvendio’s AI governance framework is designed to ensure autonomous agents operate within clearly defined commercial, operational, and security boundaries. Every AI agent is assigned a specific business role—such as Sales Enablement or Inventory—which defines its scope of responsibility and prevents it from performing tasks outside its designated remit. Agents are further constrained by topic-based skills, meaning they can only execute actions they have been explicitly configured to perform, regardless of broader user permissions.
Security is enforced through Salesforce’s native access controls, ensuring AI agents inherit the authenticated user’s permissions, role hierarchy, sharing rules, and data visibility. If a user cannot access a record, neither can the AI agent, maintaining enterprise-grade data governance without introducing additional security complexity.
At the platform level, ADvendio also includes technical guardrails through version management of the Ad Context Protocol (AdCP). Every AI interaction validates protocol compatibility before execution, preventing unsupported requests and ensuring predictable, reliable behaviour as the platform evolves.
Together, these layered controls ensure ADvendio’s AI agents are not simply autonomous—they are governed by role-based responsibilities, explicit operational permissions, enterprise security policies, and technical safeguards, enabling organisations to adopt AI with confidence while maintaining compliance, transparency, and control.
Commercial trust in automated media buying
The future of AI in media will be defined not simply by faster buying, but by trusted buying. Success depends on building systems that respect commercial agreements, protect agency relationships, and embed governance into every transaction.
When AI operates within these guardrails, it delivers the efficiency the industry is seeking without sacrificing the financial controls and accountability on which agencies and advertisers depend.
