Agentic AI threatens traditional DSP business models says industry veteran

Artificial intelligence poses multiple existential threats to demand-side platforms, according to analysis published July 21, 2025, by Ari Paparo, founder and CEO of Marketecture Media. Paparo's assessment suggests agentic AI could fundamentally disrupt the traditional programmatic advertising technology stack by automating campaign setup, targeting, and optimization functions currently handled by DSPs.

The analysis comes at a critical juncture for the programmatic advertising industry. According to Paparo, the modern DSP represents "one of the most complex categories of software ever invented," requiring massive scope of features and scale to support ad serving, creative management, data management, reporting, attribution, and fraud detection at millions of transactions per second.

However, technological shifts threaten this established model. "It's very hard to believe that AI won't be better than humans at campaign setup and targeting very soon," Paparo wrote in his Marketecture newsletter. The prediction aligns with broader industry trends toward automation and efficiency in digital advertising operations.

The timing coincides with significant market disruptions affecting transparency in programmatic advertising. Microsoft's announcement on May 14, 2025, that it would discontinue Microsoft Invest (formerly Xandr, originally AppNexus) effective February 28, 2026, removes one of the industry's most transparent DSP options from the market. According to Microsoft Advertising Corporate Vice President Kya Sainsbury-Carter, the company cited incompatibility between traditional DSP models and their vision for "conversational, personalized, and agentic" advertising futures.

The threat manifests through what Paparo terms "Agentic AI" - artificial intelligence systems that perform active campaign management tasks rather than passive analysis. These systems could theoretically deploy across multiple advertising venues within increasingly fragmented media environments, potentially eliminating the centralized role traditionally occupied by DSPs.

The transparency implications add urgency to these concerns. Industry studies reveal substantial financial leakage in programmatic advertising supply chains. According to the Incorporated Society of British Advertisers study, "a staggering 49% of the advertising dollar never reaches the hands of the publishers," meaning publishers receive only 51 cents for every dollar spent. The Association of National Advertisers found similar patterns, with 42% of programmatic spending going to "nonworking media" rather than actual advertising placement.

These efficiency losses become more problematic as fee structures directly impact campaign performance. According to analysis by Marcus Pratt of Mediasmith, different DSP fee structures can determine auction outcomes even when advertisers submit identical bids. In scenarios where two advertisers bid $5, varying fee structures and supply path costs can result in final publisher payments ranging from $2.40 to $3.83, affecting win rates and pricing dynamics.

The fragmentation challenge has accelerated since header bidding's emergence in the mid-2010s. According to Paparo's analysis, this development created supply-side duplication that removed "generic inventory access as a competitive advantage" for DSPs. The resulting environment saw supply-side platforms adding data capabilities directly, leading to the current "curation" paradigm.

Curation represents a significant shift in programmatic advertising operations. This trend has gained substantial momentum following the IAB Tech Lab's formal standards announcement in December 2024, with major platforms including Google Ad Manager and Microsoft Advertising integrating these frameworks into their operations.

The technical architecture changes become apparent when examining Paparo's framework diagrams. Traditional DSP models position the platform as a central hub managing connections between advertisers and multiple supply sources. However, agentic AI scenarios show direct connections between AI agents and various advertising touchpoints, potentially bypassing DSP infrastructure entirely.

The implications extend beyond theoretical concerns. Recent industry developments support Paparo's thesis about market fragmentation and consolidation pressures. Microsoft announced the closure of its Xandr DSP effective February 28, 2026, citing strategic pivots toward "conversational, personalized, and agentic" advertising futures incompatible with traditional DSP models.

Cross-media measurement requirements further complicate the DSP value proposition. Paparo notes that marketing mix modeling and other attribution methods have become "more important than single channel optimization," potentially undermining DSPs' traditional focus on programmatic efficiency within specific channels.

The privacy implications also favor agentic approaches. According to Paparo's analysis, AI agents present "less problematic" solutions for privacy and data movement compared to traditional DSP models that require extensive data sharing across multiple parties and platforms.

However, the transformation faces practical implementation challenges. Paparo identifies three key indicators for monitoring this transition: brands purchasing programmatic advertising without DSPs, client adoption of agentic solutions operating at scale across multiple inventory sources, and agentic platforms implementing last-mile features like frequency capping and pacing.

The competitive landscape reveals mixed signals about these developments. The Trade Desk's strong financial performance demonstrates continued demand for traditional DSP services, with revenue reaching $491 million in Q1 2024, reflecting 28% year-over-year growth. The company's investment in artificial intelligence through its Koai platform suggests established players are adapting rather than being displaced by external AI solutions.

However, the transparency legacy that distinguished platforms like AppNexus faces systematic erosion. Brian O'Kelley, AppNexus co-founder, had positioned the company as a transparency champion, revealing in 2017 that AppNexus charged just 8.5% to sellers compared to competitors with undisclosed higher fees. The platform's supply path optimization strategy routed demand to lowest-cost paths, often favoring its own exchange due to competitive pricing.

The Microsoft decision represents more than operational consolidation. According to Sainsbury-Carter's announcement, "Our commitment to more private and personalized advertising experiences for a more agentic and conversational world is not achievable with the industry's current DSP model." This philosophical shift toward AI-driven automation potentially reduces visibility into campaign mechanics and fee structures that advertisers have increasingly demanded.

Paparo acknowledges potential counterarguments through what he calls the "Taco Bell Defense" - the notion that advertising technology simply rearranges existing components without fundamental changes. However, he distinguishes current AI developments from previous technological shifts based on their scope and implementation speed.

The analysis identifies specific DSP functions most vulnerable to AI disruption. Campaign setup and targeting represent primary targets, as these activities involve pattern recognition and optimization tasks well-suited to machine learning algorithms. Creative management and attribution, conversely, may remain specialized functions where dedicated vendors already outperform most DSPs.

Industry data supports Paparo's fragmentation thesis. Recent programmatic expansion by major platforms demonstrates the increasing complexity of media buying environments. Netflix now operates programmatic capabilities across 13 countries through partnerships with The Trade Desk, Google Display & Video 360, Microsoft, and Yahoo DSP, requiring sophisticated coordination across multiple demand-side platforms.

The curation trend represents both threat and opportunity for traditional DSPs. According to industry analysis, programmatic curation builds upon traditional automated buying by incorporating strategic human oversight, enriched contextual intelligence, and advanced audience targeting. This evolution potentially creates new competitive advantages for platforms that successfully integrate these capabilities.

The technical efficiency arguments favor AI-driven approaches. Demand-side platforms have historically faced "relentless increases in computational effort measured in queries per second" required for integration with multiple parties and real-time supply filtering. Curation practices prove "naturally more efficient and therefore more sustainable," according to industry assessments.

The timeline for these changes remains uncertain. While Paparo suggests AI superiority in campaign management will emerge "very soon," the practical implementation requires significant infrastructure development and industry adoption. Current agentic AI solutions remain largely experimental, with limited production deployments at the scale required for major advertising campaigns.

Market consolidation trends may accelerate these transitions. The concentration of demand-side platforms in European markets, particularly following Microsoft's exit, creates opportunities for alternative approaches including AI-driven solutions that bypass traditional DSP architecture entirely.

The analysis suggests three potential scenarios for DSP evolution. First, traditional platforms could successfully integrate agentic AI capabilities while maintaining their central role in programmatic advertising. Second, new AI-native platforms could emerge to challenge established players through superior automation and efficiency. Third, the DSP category could fragment into specialized services as agentic AI handles core functions directly.

This evolution raises fundamental questions about the trade-offs between automation and transparency. The emerging "Outcomes Era" of advertising, as described by industry observers, prioritizes measurable business results over traditional metrics like impressions or clicks. However, this focus on outcomes often comes at the expense of visibility into how those results are achieved.

The human-versus-AI control debate extends beyond technical implementation to philosophical questions about professional competency. David Heinemeier Hansson, creator of Ruby on Rails, recently warned about similar risks in software development, arguing that over-reliance on AI tools could undermine fundamental skills. "I can literally feel competence draining out of my fingers," Hansson stated regarding AI-integrated development environments, drawing parallels to concerns about agentic AI replacing human decision-making in advertising campaigns.

This perspective highlights critical questions about maintaining human agency in AI-augmented systems. According to Hansson's analysis, learning requires active engagement that cannot be replicated through passive consumption of AI-generated solutions. The same principle applies to advertising optimization, where understanding campaign mechanics and fee structures may be essential for long-term strategic effectiveness.

Both Microsoft and Google are positioning AI-powered solutions as superior performance tools while reducing granular visibility into campaign mechanics. Internal Google communications revealed deliberate strategies to push advertisers toward fully automated solutions like Performance Max despite advertiser resistance, with executives noting "some real frustration that Google isn't listening and pushing 'full auto' solutions they don't want."

The broader implications extend to advertiser operations and agency models. If agentic AI can manage campaigns across multiple channels and inventory sources, the value proposition for traditional media buying services may shift toward strategic planning and creative development rather than tactical execution. However, this transition potentially eliminates the fee transparency and supply chain visibility that organizations like the World Federation of Advertisers and Association of National Advertisers have advocated as essential for advertiser protection.

However, implementation barriers remain substantial. Regulatory compliance, brand safety requirements, and measurement standards all require sophisticated systems that may favor established platforms with extensive infrastructure and partnership networks. The complexity of managing these requirements across multiple markets and formats presents challenges for newer AI-driven approaches.

The analysis acknowledges that specialist vendors already provide superior products compared to most DSPs in areas like attribution and creative management. This suggests the industry may be moving toward more modular approaches where AI agents coordinate specialized services rather than attempting to replicate all DSP functions within single platforms.

For marketing professionals, these developments signal potential changes in campaign management workflows and vendor relationships. The shift toward AI-driven optimization may require new skills and processes while potentially reducing operational complexity through increased automation.

The debate reflects broader questions about artificial intelligence adoption in business-critical applications. While the technical capabilities for agentic campaign management continue advancing, practical considerations including reliability, transparency, and control mechanisms will likely influence adoption timelines and implementation approaches.

Paparo's analysis contributes to ongoing industry discussions about the future of programmatic advertising technology. The assessment aligns with observations about market fragmentation, efficiency pressures, and automation trends while highlighting specific vulnerabilities in traditional DSP business models.

The convergence of these trends presents a critical question for the industry: Will agentic AI systems provide new forms of transparency that compensate for the loss of granular fee visibility, or does the shift toward automated, outcome-focused advertising represent a step backward in the industry's progress toward greater supply chain transparency?

The human competency argument suggests additional concerns beyond transparency. If agentic AI handles campaign optimization without requiring human understanding of underlying mechanisms, marketing professionals may experience skill atrophy similar to what Hansson describes in programming contexts. This raises questions about whether the advertising industry should prioritize maintaining human expertise and decision-making capabilities alongside AI assistance, rather than pursuing full automation.

As platforms like Microsoft pivot toward AI-driven solutions while emphasizing outcomes over process visibility, advertisers must evaluate whether superior performance justifies reduced control and transparency. The trade-off becomes particularly complex when considering long-term strategic capabilities versus short-term efficiency gains.

The implications for investment and strategic planning extend beyond immediate technology considerations. Companies operating in the programmatic advertising ecosystem must evaluate their positioning relative to these potential changes while balancing current operational requirements with future competitive landscapes. With over 88% of US display advertising flowing through programmatic channels, maintaining some form of supply chain visibility remains critical for advertisers seeking to maximize their working media investments.

Timeline

• Mid-2010s: Header bidding emergence creates supply-side duplication, reducing DSP competitive advantages
• 2018: AppNexus reworks publisher contracts to enable fee transparency
• 2021: Comcast acquires Beeswax, Paparo's previous company • 2023: IAB Tech Lab develops comprehensive curation framework standards
• 2024: Programmatic advertising demonstrates 18.4% expansion in Europe, reaching €15.5 billion
• December 2024: IAB Tech Lab announces formal curation standards
• May 2024: Netflix announces expanded targeting options across 100 interests and 17 categories
• May 2025: Microsoft announces Xandr DSP closure effective February 2026
• June 2025: Netflix adds Yahoo DSP as fourth global programmatic partner
• July 2025: Ari Paparo publishes analysis on AI threats to DSP business models

Key Marketing Terms

Demand-Side Platform (DSP): Technology platforms that enable advertisers and agencies to purchase digital advertising inventory through automated auctions. DSPs serve as the central hub for campaign management, targeting, bidding, and optimization across multiple supply sources. These platforms must handle millions of transactions per second while supporting complex features including ad serving, creative management, data integration, reporting, attribution, and fraud detection. The DSP category faces disruption from AI automation that could eliminate many of these core functions.

Programmatic Advertising: The automated buying and selling of digital advertising space through real-time auctions and algorithmic decision-making. This method replaced traditional manual insertion orders with efficient, scalable systems that can evaluate and purchase inventory in milliseconds. Programmatic advertising has grown to represent over 88% of US display advertising spend, but faces challenges including transparency issues where 42-49% of advertising dollars fail to reach publishers due to intermediary fees and technical inefficiencies.

Agentic AI: Artificial intelligence systems that actively perform tasks and make decisions rather than simply providing analysis or recommendations. In advertising contexts, agentic AI can handle campaign setup, targeting optimization, bid management, and creative decisions across multiple platforms simultaneously. These systems represent a fundamental threat to traditional DSP models because they can operate directly within various advertising touchpoints without requiring centralized platform infrastructure. However, concerns emerge about maintaining human competency and strategic oversight when AI systems handle complex decision-making processes that professionals traditionally managed manually.

Supply-Side Platform (SSP): Technology platforms that help publishers manage and sell their advertising inventory through programmatic channels. SSPs have evolved beyond simple inventory access to offer data enrichment, audience targeting, and curation capabilities that compete directly with DSP functions. The rise of header bidding in the mid-2010s made leading SSPs largely interchangeable for inventory access, prompting their expansion into value-added services that challenge traditional DSP roles.

Programmatic Curation: A practice where supply-side platforms leverage their capabilities to select, aggregate, and package premium inventory with relevant audience data and optimization features. This approach differs from traditional programmatic buying by incorporating strategic human oversight alongside algorithmic processes. Curation has gained significant industry momentum following IAB Tech Lab's formal standards, with major platforms including Google Ad Manager and Microsoft Advertising integrating these frameworks into their operations.

Header Bidding: A programmatic advertising technique that allows publishers to offer inventory to multiple ad exchanges simultaneously before making calls to their ad servers. This technology emerged in the mid-2010s and created supply-side duplication that reduced DSPs' competitive advantages from exclusive inventory access. Header bidding fundamentally changed market dynamics by making leading SSPs interchangeable for basic inventory access, forcing differentiation through other capabilities.

Supply Path Optimization (SPO): Strategies employed by advertisers and DSPs to identify the most efficient routes for purchasing advertising inventory, typically focusing on reducing fees and improving transparency. SPO became particularly important as programmatic supply chains grew complex with multiple intermediaries. AppNexus pioneered this approach by routing demand to lowest-cost paths, often favoring its own exchange due to competitive pricing structures that benefited advertisers.

Connected Television (CTV): Digital streaming content delivered through internet-connected devices including smart TVs, streaming sticks, and gaming consoles. CTV represents a rapidly growing advertising channel that operates largely outside traditional walled garden control, creating opportunities for independent DSPs and programmatic platforms. The format benefits from major streaming platforms expanding advertising capabilities while offering targeting precision unavailable in traditional linear television.

Transparency: The visibility and disclosure of fees, processes, and data flows throughout the programmatic advertising supply chain. Transparency has become a critical differentiator as industry studies reveal substantial financial leakage, with some advertisers seeing only 51 cents of each dollar reach publishers. Organizations like the World Federation of Advertisers and Association of National Advertisers advocate for greater transparency as essential for advertiser protection and campaign optimization.

Real-Time Bidding (RTB): The process of buying and selling advertising inventory through instantaneous auctions that occur each time a webpage loads or app opens. RTB enables precise targeting and dynamic pricing but requires sophisticated technical infrastructure to handle millions of bid requests per second. The complexity of RTB systems contributes to the technical challenges facing DSPs while creating opportunities for AI-driven solutions that can process auction data more efficiently than traditional approaches. However, as automation increases, maintaining human understanding of bidding mechanics and optimization strategies becomes crucial for long-term campaign effectiveness and strategic decision-making.

Summary

Who: Ari Paparo, founder and CEO of Marketecture Media and former Beeswax co-founder, published the analysis examining artificial intelligence threats to demand-side platforms.

What: The analysis argues that agentic AI systems could eliminate core DSP functions through automated campaign setup, targeting, and optimization, potentially bypassing traditional programmatic advertising infrastructure through direct connections to multiple inventory sources and curation strategies.

When: Published July 21, 2025, the analysis addresses current industry trends including market fragmentation, curation adoption, and AI automation developments that have accelerated since header bidding's emergence in the mid-2010s.

Where: The analysis focuses on the broader programmatic advertising ecosystem, particularly the "open web" market that Paparo estimates represents approximately €100 billion in Europe alone, encompassing display, connected television, audio, digital out-of-home, and retail media networks.

Why: Multiple factors drive this potential disruption including increasing market fragmentation that makes centralized DSP roles harder to maintain, AI superiority in pattern recognition and optimization tasks, privacy advantages of distributed AI approaches, the growing importance of cross-media measurement over single-channel optimization, and transparency challenges in programmatic advertising where studies show 42-49% of advertising dollars fail to reach publishers due to intermediary fees and technical inefficiencies.