Let’s stop pretending.

DSPs are the flip phones of digital advertising—nostalgic, clunky, and increasingly embarrassing to be seen using in public. Sure, they technically still work. Yes, they made sense in 2009. But in 2025, insisting your media buying strategy needs a full-stack DSP is like arguing your Razr is still “just as good” as an iPhone because it makes calls. It’s delusion dressed up in a dashboard.

We’ve been dragging these aging platforms along like a dead dog in a baby stroller—polishing the reporting UI, slapping on “AI” labels, adding more acronyms, and telling ourselves it’s innovative. Spoiler: it’s not. It’s just more overhead and more excuses for why your campaign went sideways while a dozen vendors took a cut on the way down.

And then—finally—someone said it out loud.

Ari Paparo, our industry’s very own reluctant prophet, the not-technically-Jewish-but-he-has-a-Jewish-first-name-so-I’m-claiming-him prince of programmatic, lit the match in his Marketecture newsletter and dropped the line that broke the spell:

*Oh, turns out I am wrong, he is indeed a Sephardi Jew.

“Ad tech is like Taco Bell—the same seven ingredients served hundreds of ways.”

If you’ve been ignoring the noise around Agentic AI, let me break it down without the sugar coating: we’re not talking about fancy dashboards with better reporting. We’re talking about fully autonomous systems that do what DSPs claim to do—but actually deliver.

They build the campaign, choose the audience, optimize the budget, cap the frequency, and pace the spend. And they do it across walled gardens, CTV apps, mobile, web, you name it—without ever needing to boot up the Frankenstein's monster we still call a DSP.

Ari laid it out clearly: these agentic systems don’t just help—they replace. And if your business model relies on DSP fees, markups, or explaining why your seat-based license is “worth it,” you should be sweating. Profusely.

There was a time when DSPs were genuinely necessary. I’ll give them that. They had to do everything: serve the ad, manage the data, report the numbers, filter the fraud, optimize the bids—and all at massive scale. They were the Swiss Army knife of digital media buying.

But here’s the thing: nobody uses a Swiss Army knife when they have a toolbox. And in adtech, that toolbox now looks like Agentic AI.

Then came header bidding, which blew the illusion of inventory scarcity wide open. SSPs started acting like DSPs, adding data and curation layers, and suddenly DSPs had nothing proprietary left. Generic access to inventory stopped being an advantage—and started being a commodity.

The foundation cracked. Agentic AI just drove a bulldozer through it.

Ari even pre-empted the usual eye-rolling from people who think every new model is just a reorganization of the same mess. He calls it the “Taco Bell Defense”—the idea that ad tech just keeps remixing the same junk and pretending it’s new.

But Agentic AI is different, and he nails why:

This isn’t just a rebrand. It’s a regime change.

Ari laid out the early warning signs, and guess what? They’re already here.

It’s not a theory. It’s a transition. And it’s already underway.

Maybe.

Maybe not.

He’s been liking some of my posts lately, even throwing in the occasional comment.

I think I’m growing on him—or at least irritating him in an entertaining way.

And yeah, I’ll admit it: we’ve had our differences. A couple years ago, I was calling out AdTech God for hiding behind anonymity while boosting clear nonsense.

Some of that got tied up with Ari, maybe unfairly, maybe not. Either way, that was two years ago. I need to let it go.

Stupid me for holding onto it this long.

Because at the end of the day? Ari’s right.

And when someone’s calling it like it is—calling out DSPs for what they’ve become, laying down the roadmap for what’s next—I’m not just going to stand on the sidelines because of some old beef.

We need more Ari.

Even if he never follows me back.

While the industry races to declare the death of DSPs, platforms like SmartyAds are pumping the brakes—and offering a more measured take.

They’re not denying AI’s potential. But full replacement? Not so fast.

Think Excel, not extinction.
SmartyAds draws a parallel: Excel launched in 1987. It’s still here—updated, smarter, faster—but fundamentally the same tool. Why? Because users still want to work with their data directly. AI made Excel more powerful, not obsolete. Same logic, they argue, applies to DSPs.

Google Ads? Same story. In 2009 it had manual setup, graphs, budgets. Today it’s loaded with AI enhancements—but the UI, the structure, the user control? Still central. It’s evolved, not disappeared.

So what’s the future, according to this camp? Not replacement. Augmentation.

They even joke about Black Mirror-style neural implant advertising—but only half kidding.

For SmartyAds, AI is a co-pilot, not the plane.
It can optimize and suggest, but buyers still want the cockpit. They want filters, graphs, toggles—and visibility. You don’t “talk” through a $500,000 buy. You touch it. You verify. You trust what you see.

As they put it: nobody wants to ask a chatbot where their money went.

SmartyAds draws a line between the flashy LLMs everyone’s talking about and the quiet, industrial-strength machine learning that powers predictive bidding, fraud detection, and pacing inside DSPs.

Want to simulate 1M queries per second with contextual signals, PMPs, and 1st-party data? They say that takes thousands of GPUs and an infrastructure that still makes centralized systems like DSPs relevant.

Try setting up complex retargeting with a chatbot? Painful. With a UI? A few clicks. For SmartyAds, interfaces are value-adds, not relics. They’re how control is retained, scaled, and refined—especially when stakes are high.

In their eyes, the future isn’t set-it-and-forget-it. It’s human + machine.
Smart dashboards. Smarter algorithms. And a media buyer still in the driver’s seat.

Demand-Side Platforms (DSPs) were originally built to simplify and scale digital media buying. In theory, they streamlined fragmented inventory, empowered buyers with real-time bidding, and offered granular control over targeting. In practice, they evolved into sprawling software fortresses—simultaneously overpowered and under-delivering.

Over the past decade, DSPs absorbed everything: ad serving, data integration, fraud detection, campaign management, and creative optimization. The result? Platforms so complex they required full-time teams just to operate—and full-time consultants to explain the billing.

By 2020, what once felt like efficiency began to look a lot like entropy. Transparency eroded. Margins thinned. Marketers, faced with bloated dashboards and unclear ROI, started asking uncomfortable questions.

“Curation” arrived as a euphemism for fixing what the DSP broke. Rather than rely on open exchanges, brands and agencies turned to Supply-Side Platforms (SSPs), publishers, and data partners to pre-package high-quality inventory deals—stripped of the junk, fraud, and wasted impressions that haunted open programmatic.

These curated Private Marketplaces (PMPs), often delivered via Deal IDs, began recentralizing power. But this time, it shifted away from the DSP and toward the sell side. SSPs layered on first-party data, contextual intelligence, and performance metrics—offering what many DSPs promised but failed to deliver.

It wasn’t just a workaround. It was a warning shot. The more buyers leaned on curated deals, the clearer it became: DSPs were becoming less platform and more pass-through.

Enter Agentic AI—a concept that makes even the best-curated PMP look like a holdover from the dial-up era.

In the agentic model, autonomous AI agents handle planning, targeting, buying, and optimization—end to end. No dashboards. No human toggling. No fragmented supply chains riddled with markups and inefficiencies.

Inventory access happens directly, via APIs. Optimization decisions are made in real time, based on dynamic input. Frequency caps? Handled invisibly. Budget pacing? Self-regulated. Channel allocation? Calculated continuously.

Agentic AI doesn’t just eliminate overhead. It collapses the need for a DSP altogether.

This isn’t theoretical. It’s visible in the market right now:

What makes Agentic especially disruptive isn’t just the tech. It’s the complete incompatibility with the legacy adtech business model.

DSPs thrived on complexity. They monetized bloat. Platform fees, managed service markups, trading desks, and consulting overlays all worked—as long as clients needed human help to navigate a maze.

But Agentic AI flattens the maze. It automates what DSPs charged a premium for. It doesn’t need account managers or a deck of “strategic recommendations.” It just works. Which means the revenue tied to navigating complexity evaporates.

Operational costs can’t keep pace. Human optimization teams? Redundant. Large UI-centric engineering budgets? Obsolete. The financial foundation of the legacy DSP is cracking—and no amount of ChatGPT-powered product updates will save it.

Audigent, once positioned as a top-tier curation and audience data player, provides an almost too-on-the-nose case study. According to several industry insiders, the company recently dismissed the bulk of its operational staff, shifting fully to AI-powered curation.

Gone are the hand-built audience packages and curated deals assembled by strategy teams. In their place: autonomous agents orchestrating audience targeting, inventory pairing, and bid strategy at machine speed.

This isn’t a one-off. It’s part of a larger trend across adtech in 2025: companies quietly retaining a handful of salespeople to maintain relationships—while AI takes over the actual work.

Expect the 2026 model to be lean. A skeleton crew of commercial leads. A few product owners. And an agentic layer doing the job that used to require forty people, six platforms, and three status calls.

What DSPs once called "managed service" is being automated out of existence.

Curation wasn’t a trend. It was the alarm bell.

And now Agentic AI has kicked the door in.

The old model—built on complexity, people, and layers—isn't just evolving. It's collapsing. And anyone still investing in the maintenance of that structure is buying more fax machines in the age of iPhones.

In Part 3: The Great Fragmentation — Why Modular AI Will Replace Full-Stack DSPs.

Stay Bold. Stay Curious. And Know More Than You Did Yesterday.

Because your DSP isn’t evolving—it’s eroding. And most of the industry is still acting like a UX facelift will fix the existential rot.

Let’s be blunt:
The king is dead—and he’s still charging you a seat license.

At ADOTAT+, we don’t write eulogies. We write post-mortems. And the “From Kingmaker to Commodity” series didn’t just throw shade—it flipped on the overhead lights in the basement of programmatic’s crumbling cathedral.

We don’t just observe—we interrogate. Why are platforms still pretending agentic automation is a feature, not a total rewrite of their value? And who’s quietly trying to become a middleware layer before the stack swallows them whole?

What buyers are actually saying when the mics are off? Hint: It’s not “we love dashboards.” It’s “why am I paying six vendors to do what one agent can handle in 30 seconds?”

We’re mapping which DSP functions have already been automated and which ones are on borrowed time. (Spoiler: strategy & planning? You’re next.)

If you still believe a DSP is “where the magic happens,” you’re reading the wrong newsletter.
But if you’re ready to see who’s quietly rewriting the playbook, automating the hard parts, and turning dashboards into plumbing?

You belong at ADOTAT+.

👉 Subscribe now.
Because midlife crises don’t fix themselves. And the DSP you love is about three quarters into a bottle of gin and quoting its Q2 roadmap like it still matters.

Stay bold. Stay curious. Know more than you did yesterday.
(And maybe stop pretending your toggles are anything more than digital decoration.)