In this guest article, Alexandra Balashoiu, Vice President of Product at programmatic adtech platform Onetag, distinguishes automation from true agentic workflows, identifies where agents can add value for buyers and publishers, and argues that shared standards will define whether agentic advertising succeeds.
Agentic AI has quickly become one of the advertising industry’s biggest talking points. New AI capabilities are emerging across the ecosystem, from campaign optimisation to media planning and decision-making. IAB UK’s The State of AI in Advertising reports that whilst only 4 percent of marketing operations are currently agentic by design, 74 percent of members are starting to experiment. Yet despite this growing attention, there is still limited clarity around what an agentic workflow actually is and what it means for the future of programmatic advertising.
Much of the discussion has focused on what AI agents can do. The more important question is how they should operate within real advertising environments where commercial decisions, privacy obligations, legal responsibilities and ethical considerations all intersect. As agents begin moving beyond recommendations towards taking actions on behalf of organisations, questions around transparency, accountability and governance become just as important as technical capability.
The next phase of advertising will depend on finding the right balance between AI-driven autonomy and human judgement. Agents can take on more operational complexity within clearly defined boundaries, while people continue to shape strategy, set objectives and oversee decisions.
Recent developments such as IAB Tech Lab’s AAMP 2.3 reinforce this principle by introducing approval gates, pricing verification and clear points where human authorisation is required before certain commercial actions can proceed. Rather than limiting autonomy, these controls make it possible to deploy agentic workflows with greater confidence.
Understanding these distinctions will be essential as the industry moves from individual AI-powered features towards more connected agentic workflows.
From Automation to Agentic Workflows
Automation has always played an important role in advertising. Programmatic technology already manages many processes, from bidding decisions to audience activation. However, most automated systems still rely on predefined rules and workflows, carrying out specific tasks based on instructions they have been given, and typically stop when they encounter situations outside those parameters, such as troubleshooting unexpected issues.
Deal synchronisation provides a useful example. Today’s Deal Sync APIs automatically activate deals between platforms using predefined fields and fixed business rules. If that process fails because of an unexpected API issue or missing information, the automation simply stops and reports an error. It cannot investigate the cause, gather additional context or attempt an alternative path.
Agentic workflows introduce a different approach.
An AI agent can be given an objective, assess available information, plan potential actions and execute tasks within a defined set of permissions and constraints. Instead of simply following a fixed sequence, it can adapt its actions based on changing conditions and feedback.
Rather than replacing Deal Sync itself, an agent could take over the troubleshooting process around it. It could investigate why a deal failed, compare information across multiple systems, identify where the breakdown occurred and recommend or execute the appropriate resolution within approved guardrails. To put it simply, automation executes predefined processes, while the agent brings reasoning, context and continuous learning to the surrounding workflow.
This distinction matters in advertising because campaigns operate within constantly evolving environments. Performance shifts, inventory availability changes, and audience behaviour develops over time. Once automation stops, managing these variables manually across multiple platforms creates significant operational complexity.
Agentic workflows can help teams respond more effectively by supporting faster decisions, reducing repetitive tasks and bringing greater consistency to campaign management.
Autonomy Within Human Guardrails
One of the biggest misconceptions around agentic AI is that greater autonomy means removing people from the process. In reality, the most effective agentic workflows are designed around collaboration between humans and AI systems.
People remain responsible for defining the foundations of a campaign, including business objectives and campaign goals, KPIs and success metrics, risk tolerance and operational constraints, and the level of autonomy an agent should have. Crucially, they are responsible for creating the knowledge bases, operational guidance and governance frameworks that agents rely on. Without well-structured institutional knowledge, even the most sophisticated agent has little context from which to learn, reason or execute reliably.
Agents then operate within those boundaries. They can plan actions based on objectives, execute approved tasks, identify optimisation opportunities and learn through feedback loops.
This creates a model where human expertise guides decision-making while AI extends the scale and speed at which teams can operate.
For advertising professionals, this means spending less time managing repetitive operational processes or troubleshooting issues, and more time applying strategic thinking, creative judgement and market knowledge.
Where Agentic Workflows Can Add Value in Programmatic
Many opportunities for agentic workflows exist across the programmatic ecosystem.
For buyers, agents can provide support throughout the entire campaign lifecycle. During planning, a buyer could provide a campaign brief and desired business outcome, allowing an agent to recommend an activation strategy, suitable supply paths and execution plan before presenting those recommendations for approval. Once a campaign is live, agents can continuously analyse campaign signals, identify optimisation opportunities and recommend adjustments based on specific objectives. Instead of teams manually reviewing multiple dashboards and interpreting fragmented information, agents can surface relevant insights and actions within the right context. This enables buyers to spend less time monitoring campaigns and more time making strategic decisions.
For publishers, agents can support more efficient supply management by helping identify valuable opportunities while maintaining quality standards. Seller-side agents could also package inventory into suitable deal opportunities by evaluating demand signals, audience characteristics and inventory quality, helping commercial teams bring higher-value opportunities to market more efficiently.
Across the wider ecosystem, agentic workflows can enable more dynamic and efficient interactions between platforms. This includes agent-to-agent trading, where systems communicate with one another to identify suitable opportunities, or automated troubleshooting, where agents detect issues and help resolve them before they affect performance. This addresses a long-standing operational challenge across programmatic. As highlighted by The Trade Desk, around 90 percent of campaigns using structured private marketplace deal IDs struggle to scale effectively, often because relatively simple technical issues require lengthy manual investigation between buyers and sellers. Rather than relying on emails, Slack messages or repeated retries, agents could identify the root cause, validate information across connected systems and resolve many of these issues automatically.
This represents a significant change in how technology supports advertising teams, moving beyond tools that simply organise information and automate processes towards systems that can actively help manage complex workflows.
To make this possible at scale, however, agents need a reliable way to connect with the wider advertising ecosystem. Programmatic advertising itself only became possible because shared standards such as OpenRTB allowed buyers, sellers and platforms to communicate through a common language within milliseconds. As agentic workflows become part of programmatic, the industry faces a similar challenge. AI agents also need a shared way to discover tools, exchange context and interact consistently across platforms.
Why Standards Will Define the Agentic Future
The success of agentic advertising will depend on alignment across the wider ecosystem surrounding AI models, particularly the standards that allow different systems to communicate and operate reliably.
The IAB Tech Lab has highlighted the importance of developing common frameworks for agentic advertising, including initiatives designed to establish protocols, registries and standards that allow agents to operate effectively across the ecosystem.
This approach requires shared infrastructure that simplifies collaboration between systems and creates a reliable foundation for agent-driven workflows.
As agentic technology develops, interoperability will become just as important as intelligence. Agents need to understand the context in which they are operating, communicate consistently with other systems, and follow clear rules around access and decision-making.
MCP and Connecting Layers for Agentic Advertising
Achieving this level of coordination is challenging in today’s fragmented technology landscape.
Digital advertising operates across a complex network of platforms, tools and data sources. For agents to deliver meaningful value, they need a consistent way to understand, access and interact with these systems.
This is where the Model Context Protocol (MCP) becomes important.
Described by the IAB as the “USB-C for an AI world”, MCP provides a universal connection layer between AI agents and the tools they need to use. Similar to how USB-C created a common standard for connecting different devices, MCP creates a shared approach for AI systems to communicate with external applications and data sources.
The wider industry is actively developing standards for agentic advertising, including initiatives such as IAB Tech Lab’s Agentic Advertising Management Protocols (AAMP) and the Ad Context Protocol (AdCP), each addressing different aspects of how agents communicate and transact. While discussion continues around the evolution of these higher-level frameworks, MCP sits beneath them as the foundational connectivity layer. Regardless of how agentic workflows evolve, agents still require a consistent method for accessing systems, retrieving context and requesting actions. That is the problem MCP solves.
At a technical level, MCP is an open standard based on a client-server model. It enables agents to discover available tools, understand what those tools can do and request permission before taking action.
For advertising, this creates the foundation for more connected agentic workflows.
MCP removes the need for every AI system to rely on separate integrations with individual platforms by providing a common interface that allows agents to work across different parts of the adtech stack. While the industry may continue refining how the best advertising agents should behave or which higher-level protocols ultimately gain the broadest adoption, the MCP’s standardised interface already addresses the challenge of securely connecting those agents to the tools they rely on.
This can enable capabilities such as agents accessing live campaign context instead of relying only on static dashboards, systems communicating with one another to support transactions, and automated processes that identify and resolve issues more efficiently.
As the industry moves towards more advanced AI workflows, this type of interoperability will become increasingly important.
Building Trust Through Transparency and Oversight
As AI systems move from making recommendations towards taking actions, governance becomes increasingly important.
Businesses need confidence that automated decisions are explainable, accountable and aligned with their objectives. This requires clear guardrails around what agents can access, what actions they can take and when human approval should be required.
The goal is to create systems that expand what teams can manage while keeping people connected to the decisions that matter most.
The future of programmatic advertising will depend on how effectively humans and AI agents collaborate and, critically, on platforms, buyers, publishers and technology partners aligning on shared standards. Programmatic’s evolution, through inventions like OpenRTB, has shown that industry-wide protocols create far greater value than isolated innovation. Agentic advertising is unlikely to be any different.
Agents can help teams process complexity at greater speed and scale, while human expertise remains essential for strategic decisions, judgement and accountability.
By combining agentic workflows with open standards and responsible governance, the industry can build a more intelligent, transparent and efficient foundation for the next generation of advertising.
Originally published on: Futureweek
