Large language models (LLMs) have made significant strides in digital advertising, permeating various aspects such as planning, creative optimization, and reporting. However, a notable exception persists: the moment when actual spending occurs. This distinction is crucial, as many platforms utilize LLMs for workflow automation, yet they refrain from granting them autonomous spending authority. The reluctance to cede this control reflects deeper industry concerns regarding accountability and the technical foundations of programmatic advertising.
At the recent Consumer Electronics Show in Las Vegas, Jeffrey Hirsch, CEO of ad tech platform QuantumPath, underscored this reality. His platform can reduce complex workflows from two hours to just ten minutes with minimal errors but intentionally avoids making budgetary decisions. “At QuantumPath, we want to automate the workflow, not the buying decisions,” he stated. This stance is emblematic of a broader industry approach that maintains a clear boundary between accelerating human workflows and replacing humans at the point of expenditure.
Industry insiders have cited various reasons for this caution, ranging from institutional self-preservation to more technical concerns related to programmatic advertising’s architecture. The fundamental incompatibility between LLMs, which operate in an open-ended semantic realm, and the deterministic logic demanded by programmatic auctions creates a significant barrier to fully autonomous spending. “It’s not going to be broadly deployed yet because of cost, readiness, and unresolved use cases,” remarked Michael Richardson, vice president of product at Index Exchange. He envisions a future where more advanced autonomous systems could emerge as infrastructure evolves and computing costs decline.
The hesitation to integrate LLMs into the spending decision process also stems from concerns about data quality. “Unreliable inputs produce unreliable decisions,” cautioned Tom Swierczewski, vice president of media investment at Goodway Group. The modern data landscape in advertising is riddled with challenges such as last-click bias and siloed metrics that resist auditing. Training autonomous systems on these flawed inputs could exacerbate existing blind spots rather than enhance decision-making capabilities.
The focus of many current LLM investments lies in enhancing foundational infrastructure rather than executing autonomous campaigns. For instance, platforms like Yahoo DSP have embraced LLMs for orchestration layers and dashboards while maintaining deterministic bidding logic at their core. “Nothing that we’re doing at the moment would suggest that agentic or an LLM will take the place of bidding logic,” stated Adam Roodman, general manager at Yahoo DSP. This careful positioning allows for improvements in workflow efficiency while preserving human oversight over spending decisions.
As various platforms like Amazon and Google roll out their own agentic capabilities with a focus on efficiency, they still implement safeguards that require human approval before any budgetary decisions are made. The emphasis remains on transparency and control, signaling a cautious approach toward automation at the critical point of spending. Both companies launched new features aimed at streamlining campaign management, yet they stop short of granting LLMs full autonomy in spending decisions.
Recent industry analyses suggest that the evolution of autonomous systems could pose fundamental questions for traditional programmatic business models. As the advertising ecosystem continues to grapple with the right balance of automation and human control, the technical infrastructure required for these systems is also advancing. The IAB Tech Lab’s introduction of the Agentic RTB Framework aims to establish standardized protocols for integrating agentic capabilities without fragmenting the ecosystem.
Shifting consumer behavior adds urgency to these developments. A recent survey indicated that 67% of consumers use AI weekly, underscoring the need for brands to rethink their advertising strategies. As AI reshapes how information is discovered, brands must ensure their content is structured and accessible to algorithms, potentially paving the way for future interactions that do not rely on traditional advertising methods.
In this evolving landscape, the distinction between enhancing infrastructure and granting spending authority remains a critical focal point. While the industry is making strides in workflow optimization and decision support through LLMs, it appears unlikely that autonomous spending authority will be granted until measurement accuracy improves and accountability frameworks are established. For now, LLMs are being integrated into multiple facets of advertising, yet they remain locked out of the spending moment—an intentional stance as the industry navigates the complexities of accountability and data integrity.
See also
Sam Altman Praises ChatGPT for Improved Em Dash Handling
AI Country Song Fails to Top Billboard Chart Amid Viral Buzz
GPT-5.1 and Claude 4.5 Sonnet Personality Showdown: A Comprehensive Test
Rethink Your Presentations with OnlyOffice: A Free PowerPoint Alternative
OpenAI Enhances ChatGPT with Em-Dash Personalization Feature
