Infillion this week published a guide titled "15 Questions You Should Ask Your DSP for the Agentic Era," directed at media buyers who are evaluating or re-evaluating their demand-side platform partnerships as artificial intelligence takes on a larger role in programmatic campaign management. The guide, distributed by email on April 29, 2026, to the PPC Land editorial address, frames DSP selection as a decision with long-term structural consequences - not merely a platform preference.

The core premise is a projection: according to the guide, by 2028, more than $100 billion in programmatic spend is expected to be managed by AI-driven systems. That figure sets the stakes. A DSP chosen today will be handling campaigns in a fundamentally different operational environment within a few years, and the guide argues that most buyers are not asking the questions that would reveal whether their current or prospective platform is equipped for that environment.

Infillion is a New York-based advertising platform assembled from the unification of four previously independent companies - MediaMath, TrueX, Gimbal, and Drawbridge - with a stated investment of $750 million in technology. The company relaunched as an agent-native composable platform in January 2026 and acquired Catalina in February 2026, gaining access to a deterministic purchase intelligence database covering 130 million U.S. households and $600 billion in annual consumer spending.

Pricing transparency and hidden fees

The first question in the guide targets cost structures. According to the document, buyers should ask whether the DSP charges separately for campaign or data onboarding, bid shading, log-level data access, custom curated inventory, or connecting the platform to AI tools and automation systems. The guide states: "As buying gets more sophisticated, pricing models become more complex, and not all costs are immediately visible. Hidden fees add up fast and can quietly erode campaign performance."

This is not an abstract concern. As programmatic workflows become more automated and more layered - incorporating audience enrichment, identity resolution, and agentic execution - the number of billable integration points multiplies. A platform that charges separately for each connection can add meaningfully to effective campaign costs before a single impression is served.

AI architecture: native versus bolt-on

Question four addresses what the guide calls a critical distinction in platform architecture. According to the document, "there's a big difference between a platform designed from the ground up to work with AI and one that has added AI features as an afterthought." A genuinely AI-native platform, the guide argues, lets automated systems - with humans in the loop - handle planning, bidding, and optimization end-to-end. The guide's analogy is blunt: "AI bolted onto a legacy platform is like putting a new engine in a broken-down car. The fundamentals still hold you back."

That framing tracks with a broader industry conversation that PPC Land has been documenting throughout 2025 and into 2026. Yahoo DSP integrated agentic AI capabilities on January 6, 2026, enabling advertisers to automate campaign setup, troubleshooting, and optimization through natural language. AdRoll and PubMatic demonstrated MCP-powered agent-to-agent diagnostics on April 23, 2026, allowing demand-side agents to query supply-side systems in real time without switching platforms or waiting for manual support responses. The pattern is consistent: platforms are building AI into transactional workflows, not just reporting interfaces.

Openness and interoperability

Question five in the guide asks whether buyers can bring their own AI tools, or whether they are locked into the DSP's proprietary systems. According to the document, "many agencies and brands are building or buying their own AI systems to manage campaigns." Open platforms give flexibility and interoperability; closed ones create dependency and risk. This question has become structurally important as the Model Context Protocol - originally developed by Anthropic and subsequently donated to the Linux Foundation - has emerged as a connective standard across advertising platforms.

The guide positions Infillion's own approach here through its Infillion Agent Connector, described as the first solution in the industry built specifically to let any AI agent - commercial, custom-built, or internal to an agency - plan, buy, and optimize campaigns directly through an MCP connection. The document states that this approach "removes proprietary lock-in and avoids the need to rebuild workflows for every new tool." Amazon opened its advertising APIs to AI agents through MCP in February 2026, and FreeWheel launched an MCP server into premium video ad infrastructure in March 2026. Infillion's Agent Connector enters that ecosystem as a buy-side implementation.

AI transparency and explainability

Question six raises an accountability issue that sits at the intersection of platform design and client relationships. According to the guide, when AI is making decisions about budget allocation, buyers need to be able to see and understand those decisions. The document asks: "Can you access a clear record of what the AI did, when, and why? Can you explain campaign decisions to a client without guessing?" The guide's framing is terse: "'The algorithm decided' isn't a sufficient answer when a client asks why performance changed."

That is a practical operational problem, not a theoretical one. As automated systems take on more campaign responsibility, the ability to reconstruct a decision chain becomes essential for client reporting, troubleshooting, and regulatory compliance - particularly in verticals subject to GDPR or CCPA scrutiny. The guide returns to this theme in question 11, which asks whether optimization can be proven through case studies that tie execution to real business outcomes - sales, store visits, conversions - rather than media metrics like impressions and clicks.

First-party data and audience control

Question seven covers first-party data ownership and control. The guide states that buyers should confirm the platform lets them bring their own customer data, control who sees it, and ensure it is not co-mingled or used for the DSP's own gain. According to the document, "your audience data is a competitive advantage. The right DSP activates it for you. It doesn't lock it in, co-mingle it, or use it for their own gain."

First-party data has become the central asset in programmatic targeting following the extended debate over third-party cookie deprecation. Although Google ultimately reversed its cookie deprecation plan in April 2025, the infrastructure investments made by platforms in first-party identity resolution have not been unwound. If anything, the competitive value of verified first-party data has increased as the industry recognized how fragile tracking infrastructure built on third-party signals had become.

Omnichannel and supply ownership

Questions three and eight address supply-side dynamics. On supply ownership, the guide warns that if a DSP also owns ad inventory, its optimization system - including any AI - risks prioritizing that inventory over better-performing options. The document advises buyers to ask "whether their optimization is truly performance-based, or whether owned inventory gets preferential treatment." The incentive structure question is direct: "are their incentives aligned with your outcomes, or are they making money regardless of whether your campaigns perform?"

Question eight covers omnichannel reach. According to the guide, a genuinely connected campaign should be able to run across CTV, display, digital video, audio, and digital out-of-home (DOOH) from a single campaign structure, without requiring manual budget reallocation between channels each time conditions change. "Constantly stitching together separate channel buys wastes time and creates gaps in your data," the guide states.

Identity resolution without cookies

Question 12 is the most technically specific in the guide. With third-party cookies largely gone from Firefox and Safari - and the Chrome situation still unresolved despite Google's reversal - buyers need to understand exactly how their DSP identifies and reaches audiences across devices and channels using privacy-safe methods. The guide recommends looking for a DSP that is "agnostic to any identifier, including those from competitive providers, and compares all solutions to optimize toward maximum reach for your campaign."

The practical implication is that a DSP tightly coupled to any single identity provider - whether a publisher-controlled login graph, a hashed email solution, or a probabilistic device graph - carries concentration risk. If that identifier loses coverage or regulatory approval, targeting performance degrades with it. Identifier agnosticism, combined with the ability to optimize across multiple signals simultaneously, is described as the more durable architecture.

Data ownership and log-level access

Question 13 addresses what happens to campaign data - both during and after a buyer's relationship with a platform. According to the guide, buyers should confirm they can access raw log-level data, not just pre-built dashboards. They should also ask whether API access and reporting are updated frequently enough to act on, and - critically - what happens to audience insights, campaign history, and performance data if they leave the platform. The guide states: "Data you can't access is data you can't use. And data you don't own becomes leverage someone else controls."

This question applies particularly to buyers who have built audience models or attribution frameworks inside a DSP's proprietary environment. If those assets cannot be exported, they represent a lock-in mechanism as powerful as any contractual arrangement.

Vertical support and compliance signals

Question 10 uses vertical support as a proxy for compliance sophistication. According to the guide, whether a DSP can run campaigns in regulated categories - pharma, COPPA-compliant environments, gambling, alcohol - signals how seriously the platform takes privacy compliance including GDPR, CCPA, and COPPA more broadly. The reasoning: "A DSP that operates in the most tightly regulated verticals has had to build robust privacy controls. That's good news for everyone on the platform, not just regulated advertisers."

That logic is persuasive for large agencies managing clients across multiple categories. A DSP that has invested in compliance infrastructure for pharmaceutical or children's advertising has, by necessity, built consent management, data minimization, and audit trail capabilities that benefit all campaigns on the platform.

Support and the agentic shift

Question 14 asks about customer support - specifically, 24/7 availability and whether buyers get a dedicated contact or a generic help desk. The guide is unambiguous: "A DSP that isn't available when issues arise functions more like a vendor than a partner."

The agentic context makes this question more urgent. As AI systems take on more operational responsibility, the failure modes change. An underpacing campaign managed by a human trader generates a support ticket. An underpacing campaign managed by an AI agent may generate cascading automated responses before a human intervenes. The question of who picks up the phone - and how quickly - becomes a system reliability question as much as a service quality one.

The differentiation question

Question 15 - what can a DSP offer that is genuinely unavailable elsewhere - is the guide's most open-ended prompt. According to the document, "with so many platforms making similar claims, this question cuts to the heart of what really differentiates a DSP. Look for a platform that is building for where the industry is going, not just where it's been."

The guide uses Infillion MediaMath as its own example of a platform built for this model, citing the composable platform structure, privacy-safe identity graph, high-attention ad formats, and real-world attribution capabilities alongside the Agent Connector. Infillion partnered with Yobi in January 2026 to bring Performance Max-style AI optimization to the open web, providing a practical illustration of the composability it describes.

Why this matters for the marketing community

The guide lands at a moment when the DSP market is in genuine structural flux. Microsoft shut down its Invest DSP (formerly Xandr) in February 2026, removing one of the industry's more transparent independent options. Kochava opened its StationOne platform to public beta in March 2026 with IAB Tech Lab's agentic advertising infrastructure, giving buyers a place to test agentic workflows before committing real spend. Optable and Goodway Group announced a partnership on April 27, 2026, with more than 70 Goodway team members already operating agentic workflows through the platform. The pace of change is fast enough that a DSP selected 18 months ago may now have a substantially different capability profile - or may no longer exist.

PPC Land has tracked the expansion of MCP-based advertising tools throughout 2025 and into 2026, with implementations spanning Yahoo, Amazon, FreeWheel, PubMatic, AdRoll, Dstillery, and now Infillion. What distinguishes Infillion's guide from typical vendor collateral is that its 15 questions are genuinely adversarial - designed to surface weaknesses in any platform, including the one publishing them. Whether buyers use it to evaluate Infillion or a competitor, the framework asks questions that the industry has historically avoided asking loudly.

Timeline

Summary

Who: Infillion, a New York-based advertising platform formed through the unification of MediaMath, TrueX, Gimbal, and Drawbridge, published a buyer education guide targeting media buyers and agencies evaluating DSP partnerships.

What: A 15-question framework covering hidden fees, AI architecture, openness to external AI tools, AI transparency, first-party data control, omnichannel support, inventory curation, vertical compliance, identity resolution without cookies, data ownership and log-level access, customer support, and platform differentiation. The guide cites a projection that more than $100 billion in programmatic spend will be managed by AI-driven systems by 2028, and introduces Infillion's Agent Connector as the first solution built specifically for MCP-based AI agent access to a DSP.

When: The guide was distributed by email on April 29, 2026.

Where: The guide was distributed to industry contacts and published by Infillion. It was sent to the PPC Land editorial address from [email protected].

Why: Infillion frames the guide as a response to an industry-wide gap between the sophistication of modern programmatic buying and the questions buyers typically ask before committing to a platform. With AI automation projected to manage over $100 billion in programmatic spend by 2028, and with the DSP market in structural transition following the closure of Microsoft Invest and the rapid deployment of MCP-based agentic capabilities across multiple platforms, Infillion argues that buyers who do not ask these 15 questions risk selecting platforms whose architecture, incentive structures, or data practices will constrain performance as complexity increases.

Share this article
The link has been copied!