TL;DR: Agencies have been automating programmatic buying since the technology existed; agentic advertising is not the introduction of automation to agency workflows but a change in which layer of the buying stack gets automated. Rules-based automation (bid caps, dayparting, frequency limits) has been standard since DSPs launched. Machine learning optimisation became mainstream through the 2010s. Agentic advertising introduces a third stage in which an agent executes buying decisions against a configured mandate, rather than applying rules or models to bid prices within a human-managed campaign structure. The discipline this demands from agencies is mandate design: translating a client brief into agent instructions precise enough to produce the intended behaviour without exploitable gaps.
Agencies have been automating programmatic buying since programmatic existed. The suggestion that AI agents represent automation arriving in media buying misreads a decade of development. What agentic advertising introduces is not automation. It is a change in the level of the buying stack that gets automated, and that change has real consequences for what agencies need to know how to do.
What did programmatic automation look like before agents?
The first layer of automation that agencies adopted was rules-based: bid caps to prevent overspend in specific placements, frequency limits to prevent ad saturation, dayparting to restrict delivery to defined hours. These were deterministic rules applied at the campaign or line-item level. A human configured the rule; the DSP applied it without human involvement at each impression. This was operational automation: removing the human from individual bid decisions while keeping the human responsible for the rule that governed those decisions.
The second layer was machine learning optimisation. DSPs introduced dynamic bid pricing that adjusted the maximum bid for each impression based on predicted conversion probability. Audience modelling tools built lookalike segments from seed populations and expanded reach without manual audience list management. Attribution models fed back into bid strategy to weight bid prices toward placements that historically contributed to conversion. By the mid-2010s this layer was mainstream at any significant agency. A human set the optimisation objective; the model determined the bid price.
In both layers, the human retained direct control of the campaign structure: the line items, the audience segments, the creative assignment, the flight dates and budgets. The automation applied within that structure, not above it.
Where are agencies deploying agents today?
Agentic advertising operates above the campaign structure layer. A media-buying agent, given a mandate, can negotiate deal terms with publishers, manage pacing across a campaign period, and respond to inventory signals without a human configuring each operational parameter. The agent acts on the mandate, not within a campaign a human has built.
Deployment at this stage is limited. Andrew Mole of pubX confirmed in an ADOTAT investigation published in August 2026 that live agentic media spend sits at approximately $3,000 a day from one named operator. This is early-stage commercial deployment, not mainstream adoption. The agencies building the capability now are doing so while the volume consequences of errors are small.
The IAB Tech Lab's agentic advertising working group specifications define the governance requirements for this layer: what constitutes a mandate, how approval thresholds trigger human review, and what a shared deal record must contain for both buy-side and sell-side agents to reference it. These specifications set the boundary conditions within which agencies are designing their first mandate frameworks.
What is the discipline of mandate design?
Mandate design is the skill set that agentic advertising introduces as a distinct agency competency. A mandate is an instruction set that defines what a media-buying agent may do, within what constraints, and at what approval threshold. Writing a mandate that produces the intended agent behaviour requires precision that campaign management historically did not demand, because the consequences of ambiguity are different.
A vague bid cap in a DSP campaign produces mildly inefficient spend. A vague mandate for an agent that can negotiate deal terms directly with publishers and commit budget against those terms produces an agent that finds every gap between what was intended and what was written. The mandate is the governance document. Its precision is the primary determinant of whether an agent behaves within the intent of the client brief.
Mandate design requires the agency to make explicit several things that campaign management kept implicit: which data assets the agent is authorised to use and under what legal basis; what creative format constraints apply to any placement the agent may win; what spend authority the agent has before human approval is required; and what brand-safety parameters govern contextual placement. Translating a client brief into each of these specific, unambiguous constraints is a new discipline for most agency teams.
How does agentic advertising change what an agency offers?
The shift from campaign management to mandate management is a genuine change in agency value, not a rebranding of existing work. An agency that was expert at configuring bid multipliers and managing audience lists was offering operational expertise: knowing how to extract performance from DSP tooling. That expertise had value because the tooling was complex and the performance difference between good and poor configuration was measurable.
Mandate design is a different kind of expertise. It requires understanding what an agent will and will not interpret correctly, how to write constraints that are unambiguous to a machine, how to set approval thresholds that maintain human oversight without generating approval requests so frequently that the agent cannot operate efficiently, and how to design the audit trail that will allow the agency to demonstrate compliance with the client's instructions and with data protection obligations.
Agencies that build this competency create a durable advantage. The mandate frameworks developed for early agentic deployments become the institutional knowledge that governs larger-volume deployments. The agencies that wait until agent mandate design is a procurement requirement will find themselves buying that expertise from outside.
*This article references the IAB Tech Lab's published agentic advertising working group specifications and volume data from Andrew Mole of pubX, published in an ADOTAT investigation in August 2026.*