Pitch-deck preachers. LinkedIn prophets. “Strategic consultants” who think DSP stands for Don’t Show Proficiency.

Your boss paid them $5,000 to “upskill” the agency. They used it on bottle service, bad decisions, and a “training deck” stolen from someone else’s keynote. Then they mumbled something about “funnel velocity” and couldn’t diagram a single stage if you handed them crayons.

Now they’re quoting ADOTAT+ like they came up with it.

Meanwhile, DSPs? They’re not platforms—they’re padded cells for aging media logic. They used to command the buy. Now they sit in the corner, taking orders from upstream AI like a middle manager clinging to their last Slack login.

Execution stayed behind. Intelligence left the building.

That’s why we built ADOTAT+.

We expose the underbelly:
🧾 The late payments.
📉 The phantom impressions.
📂 The backchannel whisper campaigns.
💣 The fraud tools that couldn’t detect a bot waving a red flag in Times Square.

This isn’t a newsletter. It’s a crowbar. And we’re using it to pry open the black box of this industry, live on stage, receipts in hand.

So if you’re done pretending your LinkedIn feed is “intel,” and you want real ammo to pitch smarter, spend cleaner, and stop getting punked by your stack—subscribe to ADOTAT+.

Or don’t.
And let your competitor keep quoting us while beating you on every call.

I’ve been spending the past few weeks talking to people who are way smarter than me about where the ad stack is going. And the more I listen, the more I realize we might be staring at one of those “the floor just moved” moments — the kind that takes an entire industry from steady to chaotic in less than a year.

Maybe it’s 24 months away.

Maybe it’s six.

But when it hits, the business models we’ve been running for the last decade won’t survive the month.

For years, DSPs were the nerve centers of digital advertising — stitching together fragmented supply, giving us the levers to pull, making programmatic not just possible, but scalable.

They still do that. They still matter.

What’s changing is where the thinking happens. The decisions about what’s good inventory, what’s on-brand, and what’s worth bidding on are increasingly being made before the DSP even logs in for the day.

Brian O’Kelley put it bluntly when we talked: “The entire business of just classifying inventory as safe or not is a dead business… You can build something richer upstream, connect the agent directly to content, and you’ll see better performance before the DSP even touches it.”

That stuck with me because it’s not about replacing DSPs — it’s about changing where the intelligence sits in the stack.

When programmatic began, the promise was simple: transparency, control, and scale.
For many campaigns, DSPs still deliver on that promise.

But Gareth Capon of Grabyo told me what a lot of buyers are starting to whisper: “The DSP ecosystem has become increasingly murky. There’s not a lot of differentiation, and brands are starting to ask, ‘What am I paying for exactly?’”

That’s not cynicism. That’s curiosity — the kind that leads to big shifts. If the same placement is being sold in three DSPs, maybe it makes more sense to decide earlier in the process — before it’s packaged, repackaged, and marked up twice.

When I log into a DSP, the familiar sliders, toggles, and targeting panels are all still there — the toolkit that’s moved billions in ad spend.

But I’m hearing more talk about adding layers above the DSP — contextual pre-filters, curated marketplaces, and AI models that narrow the target list before the bid fires.

One buyer didn’t mince words: “Three DSPs claiming exclusive access, yet I see the same garbage inventory in all of them. Unless you manually blacklist every overlapping exchange, you’re bidding against yourself.”

That’s not an anti-DSP statement. It’s a push for smarter filtering earlier in the chain.

Here’s something I hadn’t fully appreciated until I started digging in recently: DSPs are also financing the supply chain. They pay SSPs in 30 days while brands often pay them net-90 or net-120.

For some, that makes DSPs indispensable — a partner who can bridge the cash gap. For others, it’s a red flag. Do you want your ad platform to be a bank?

It’s part of why Microsoft stepped away from Xandr, why MediaMath collapsed, and why Tim Vanderhook at Viant is betting on identity infrastructure. As Vanderhook said recently, “Identity infrastructure is still very important for where we have to go in the future.”

Agentic AI doesn’t replace DSPs. It changes what they’re asked to do.

Jules Minvielle of Olyzon is building agents that can read show transcripts in real time, infer tone, and create inclusion lists — then hand that list to the DSP for execution.

O’Kelley is wiring Scope3’s AI directly into marketplaces to identify sustainable, high-quality inventory before it even hits the DSP’s radar. As he told me, “This isn’t about making brand safety better. It’s about making it part of a much bigger conversation.”

Here’s what I’ve learned:

Some buyers are already there. Others will wait until the last possible moment. But when the tipping point hits — whether in two years or by next summer — the industry’s center of gravity will shift almost overnight.

Execution may stay in the DSP.
But the most important decisions? They’re moving somewhere else.

And that’s where things get really interesting.

Once, it was the alpha of the adtech pack — fast, loud, and smart enough to know where every bone was buried.

Brands fed it budgets like raw steak, agencies followed its bark without question, and entire careers were built on keeping it happy.

Now the coat’s dull, the bark is hoarse, and the leash is held by something colder: AI-led planning layers. The power has moved upstream.

In a cramped conference room with stale air and burnt coffee, Tyler Kelly, President of Basis Technologies; Jules Minvielle, Co-founder and CEO of Olyzon; and Suds Roy, a global commercial and growth leader who once turned down the Global COO seat at a major DSP, are staring at the same tired dog and arguing about whether it’s still worth feeding.

Kelly starts with the calm of a man who’s defended this animal before….

The DSP isn’t dying, he says — “they’re not being replaced by AI, they’re already part of it.” In his view, “the best DSPs embody AI at their core” and still offer automation, control, and precision “that matter when you’re actually running campaigns.”

Roy barks back before Kelly’s finished. “DSPs aren’t making decisions anymore — they’re just carrying them out,” he says, leaning forward. “The intelligence has moved upstream into AI-led planning layers. What’s left is a glorified API endpoint charging 4–8% for the privilege.”

Minvielle jumps in, nodding toward Roy. “Exactly. The real value has moved upstream to platforms capable of interpreting briefs, understanding context, orchestrating channels, and sequencing ads.” DSPs, he adds, “are efficient at bidding and pacing, but blind to brand meaning, audience emotion, and campaign intent.”

Kelly tries to pull it back to execution, but Roy isn’t letting go: “The much-hyped DSP AI — Kokai, for example — isn’t discovering value for advertisers. It’s optimizing bids to drive platform EBITDA margins. That’s execution for profit, not strategy.”

Minvielle points at the DSP UI on the projector screen like it’s a bad crime scene photo. “It’s proof these things are back-end utilities. You don’t see water pipes with fancy dashboards — you just expect them to work after the real planning happens upstream.”

Kelly waves the jab away. “Real value lies not in the interface but in the DSP’s ability to plug into AI, clean supply, and decisioning tools at scale.”

Roy shakes his head. “When we had to build an entire logic engine and agent model outside DV360 — audience, context, modular creative, granular targeting — because DV360 couldn’t do it, it begs the question: why pay 2.5–4% DSP fees at all?”

Kelly leans in again: “Publishers want to optimize yield across every partner, and buyers want to hedge against fragmentation and redundancy. A multi-DSP setup can make sense.”

Roy smirks. “It’s a mirage. Your second DSP after the first delivers marginal, overlapping inventory while escalating non-working media costs. Without exclusive supply, multi-DSP setups are economic sabotage.”

Minvielle nods toward Roy. “Redundancy is complexity disguised as control. Every extra DSP means extra contracts, extra reconciliations, and extra retraining cycles.”

Roy’s voice sharpens. “The DSP has turned into a short-term lender rather than a technology platform. That’s not a cycle — it’s a death spiral. Friends of mine are picking up $10–30 million in liabilities a month because a DSP can’t pay on time.”

Kelly says the stronger players — “the ones not operating on caffeine and wire fraud” — can weather it. “Meta and Google shifting to 60-day payment cycles has forced discipline into the system.”

Minvielle shakes his head slowly. “Thin margins and delayed payouts accelerate consolidation. And consolidation just pushes more leverage upstream.”

Roy delivers the kill shot. “DSPs are on the wrong side of a secular attention shift. User discovery is moving upstream into AI-driven assistants, walled gardens, and retailer ecosystems, leaving DSPs extracting fees from collapsing inventory quality.”

Kelly still sees hope if DSPs “evolve into AI-infused orchestration partners.”

Minvielle’s already looking past that. “The coming attention shift will only speed up the move toward upstream strategy platforms. And when that happens, the DSP is just a workhorse taking orders.”

The room is quiet now, save for the whir of the projector fan and the faint clink of a pen against the table.

Three men.

One damn tired dog.

Three different fates: train it, put it down, or leave it in the yard while the real work happens inside.

The fight’s fictional. The quotes are real. The leash is still up for grabs.

For years, demand-side platforms (DSPs) have served as the primary on-ramps to programmatic advertising. Built to optimize real-time bidding on the open web, these platforms were born in an era dominated by cookies, clicks, and single-user devices. That era is rapidly fading.

And nowhere is this more evident than in Connected TV (CTV), where the very nature of content, context, and identity has outgrown the capabilities of traditional DSPs.

CTV was meant to be digital advertising’s great convergence—a way to marry television’s scale and storytelling with the precision of digital targeting. Instead, it’s exposing the foundational limitations of DSP infrastructure.

Connected TV introduces a fundamentally different data environment. In a browser, targeting is deterministic: cookies, device IDs, user logins. In CTV, these signals become sparse or irrelevant. The household device is shared, privacy regulations are tightening, and the available data—transcripts, content metadata, viewer behavior—is probabilistic at best.

DSPs weren’t built to interpret this. They were engineered to match bid requests to audience profiles based on pre-existing identifiers, not to analyze the emotional tone of a documentary, the pacing of a scene, or the real-time context of a sports broadcast.

In short: CTV demands meaning. DSPs were designed for math.

Today’s leading DSPs—including The Trade Desk, DV360, and Amazon DSP—trace their architectures back to web-first models. They excelled in environments where one device equaled one user, where tracking scripts ran freely, and where the value proposition was optimization at scale.

That model fails on a smart TV mounted in a living room with five viewers and no cookies.

The result is a mismatch: DSPs still serve as the transactional layer for programmatic CTV buys, but they increasingly rely on external systems—contextual agents, AI models, and inclusion tools—to make sense of what they’re actually bidding on.

Rather than scrap DSPs, the market has begun to layer intelligence on top of them.

Companies like Olyzon, led by Jules Minvielle, are at the forefront of this shift. Their AI-powered platform interprets live content, constructs dynamic inclusion lists, and executes activation through DSPs—but only after the agent has determined what’s relevant. Olyzon currently integrates with platforms like Amagi, Fubo, LG Ads, and Plex, and its logic is designed to interface directly with The Trade Desk, DV360, and Amazon DSP.

This is not a theoretical roadmap. It’s already live and operating across major brand campaigns.

Agentic media buying—the concept of using autonomous software agents to make contextual advertising decisions—is gaining traction not just as a concept, but as infrastructure. These agents, often AI-driven, ingest a wide array of signals: genre, viewer attention cues, content tone, and more. The result is a system that doesn’t rely on identity, but instead operates on real-time understanding.

As Brian O’Kelley, co-founder of AppNexus and current CEO of Scope3, told ADOTAT “We don’t need another DSP. We need smarter decisioning, better optimization algorithms, and systems that can interpret context natively.” He’s argued that agentic systems will play a central role in CTV because they “solve for the hard parts”—namely, the absence of user-level identity and the complexity of household viewing.

O’Kelley’s broader thesis is simple: CTV is where real innovation in advertising must happen. Not because it’s trendy, but because it exposes the weaknesses of the legacy stack. He’s been especially critical of existing brand safety tools—calling many of them “medieval nonsense”—and says agentic systems will redefine suitability and performance by making decisions based on the content itself, not just where the ad appears.

This isn’t a requiem for DSPs. They continue to serve essential functions: centralized campaign management, frequency capping, cross-channel attribution, and cost efficiency. But in the CTV ecosystem, they’re increasingly dependent on intelligence layers they weren’t built to provide.

What we’re witnessing is a bifurcation of function: DSPs remain the transaction layer, but intelligence—contextual, emotional, even environmental—is migrating upstream. Buyers still need DSPs. They just can’t rely on them alone to navigate CTV.

CTV will not conform to the rules of web-based programmatic. If anything, it will demand new standards—some already emerging—that prioritize consent, relevance, and media quality over scale and precision. The winners in this new era won’t be the platforms with the most knobs and toggles. They’ll be the ones who understand the room: not just who’s in it, but what they’re watching, how they feel, and whether the ad makes sense at that moment.

That’s a tall order. It also may be the only one that matters.

If you thought the DSP wars were just about better dashboards, you’re missing the real fight. In the full ADOTAT+ series, we go deep—really deep—into the last days of the demand-side platform as we know it, why Viant and The Trade Desk are clawing at the same 97% of inventory, and what happens when AI agents make the UI itself irrelevant.

Here’s a taste of what’s inside:

This isn’t a quick blog post—it’s nearly a book on the tectonic shifts happening in media buying. You’ll see the names, the numbers, the timelines, and the real-world examples that big agencies, holding companies, and publishers are already acting on.

Stay bold. Stay curious. And if you want to know what’s really replacing the DSP, you’ll find it in ADOTAT+.