Basis Compass Does in Five Minutes What Media Planners Bill Eight Hours For

The part of media buying that everyone assumed was safe from AI wasn't the bidding. Bidding has been automated for years. It wasn't the targeting, either. Lookalike audiences and algorithmic expansion already turned most manual audience selection into a polite suggestion the platform quietly overrides.

The part that was supposed to be safe was the thinking. The strategy. Taking a client brief, reading between the lines, matching channels to objectives, building a media plan, and presenting it in a deck that justified next quarter's spend. That's the work media planners actually bill for. And that's the work Basis just automated.

What Compass Does (and Why the 90% Number Matters)

Compass is an agentic AI tool baked directly into Basis's DSP. You feed it a campaign brief, along with whatever context you have (audience data, past campaign reports, brand messaging docs), and it produces a complete omnichannel media plan covering programmatic, direct, paid search, and paid social. Then it lets you activate that plan without switching platforms.

The numbers from beta testing: tasks that used to take an hour are finishing in about 5 minutes. Basis claims a 90% time reduction and a 15% drop in manual effort across planning processes. CEO Shawn Riegsecker put it bluntly: "For the first time, a DSP can take a campaign brief and generate a complete, customizable omnichannel media plan."

I'll give him this: he's not wrong about the first time. Most AI features in ad tech have focused downstream, on bid optimization, creative testing, audience expansion, that sort of thing. Compass goes upstream, into the planning work that happens before a single impression gets bought. That's a different kind of automation entirely.

The system runs on proprietary LLM agents trained on workflows from Basis's in-house media experts. Not a generic ChatGPT wrapper. It includes evaluation layers to monitor for bias and demographic sensitivity, which is... the minimum, honestly, but at least they mentioned it. Outputs export to Word and PowerPoint, which tells you exactly who this is for: the account manager who needs a deck for a client meeting by Thursday.

Three DSPs, One Race, Same Month

What makes this more than a product launch is the timing. Basis isn't the only one building an AI media planner. They're just the first to ship one.

Google folded Gemini deeper into DV360 at NewFronts in March. Marketers can now upload a media plan and the AI automatically translates it into a comprehensive campaign setup. Bill Reardon, Google Ads' enterprise platform GM, said Gemini "will be core to the success of the future of advertising." That's a big claim from the company that already controls most of the plumbing.

Meanwhile, The Trade Desk is running a closed beta that lets advertisers build campaigns using Anthropic's Claude. CEO Jeff Green framed it around scale: 20 million ad impression opportunities every second, 10 milliseconds per decision, millions of campaigns running at once. He plans to launch an agentic AI framework for partners later this year.

Three major DSPs. All moving upstream into planning. All within roughly a month of each other. That's not coincidence. That's a race, and the finish line is owning the entire workflow from brief to bid to report. I'd expect at least two of the three to ship planning-to-activation AI as a standard, non-beta feature by the end of 2026.

Where This Gets Uncomfortable for Agencies

If you work at an agency, this is the part that should sit a little heavy.

Agencies make most of their margin on two things: media commissions and planning hours. The commissions have been under pressure for years. The planning hours were the backstop. A junior media planner building a deck, a strategist reviewing channel allocation, a director walking a client through the recommendation. All of that is billable time.

If the DSP does the planning, what do you bill for? The supervision? A quality check? The "I reviewed what the AI recommended and it looked fine" meeting?

I don't think this kills agencies. Not even close. The clients who need strategic partners will still need strategic partners, probably more than ever, because someone has to decide whether the AI's plan is actually good. But the middle layer, the work of assembling a plan from known inputs and putting it into a presentable format, that's where the compression happens. And it's already started.

One agency user from the Compass beta, Tim Slater at SMZ, said Compass "closes the learning gap when exploring new verticals, media channels, or campaign goals." Which is a polite way of saying: it reduces the cost of not knowing something. The experience advantage senior planners have over junior ones just got smaller.

The Trust Problem the DSPs Created

There's a catch, and from what I've seen in programmatic, it's a significant one. According to IAB data cited by eMarketer, 60% of US ad industry professionals cite concerns about accuracy and transparency as the top barrier to adopting AI in media campaigns.

That number exists for a reason. When the platform that sells you inventory is also the platform that recommends how much to buy, you've got a structural conflict of interest. Google recommending you spend more on Google inventory through a Google AI is not exactly a neutral planning exercise. On paper, that sounds like an upgrade. And sometimes it is. But the incentive structure underneath it hasn't changed.

Basis has a slightly better story here because they're independent. They don't own inventory. Their recommendations don't directly increase their own ad revenue. But the DSP still benefits from higher spend flowing through its platform, so the incentive alignment isn't perfect either. Nobody's incentive is perfectly clean in this business, to be fair.

The advertisers who seem to do well with this kind of automation are the ones who use it as a starting point, not a final answer. If Compass spits out a media plan in 5 minutes and you activate it in 6, you've probably saved time and lost quality. If you take the 5-minute plan and spend 30 minutes pressure-testing it against what you actually know about the client's business, you might have something genuinely useful.

When Your DSP Becomes Your Strategist

There's a broader pattern worth noticing here, and it connects to a few things we've been covering around AI and ad platforms. The DSP used to be infrastructure. Pipes. You told it what to buy and it went and bought it efficiently. Now it wants to tell you what to buy.

That's not inherently bad. But it does change the relationship. When your DSP writes the plan and executes the plan, it becomes both the advisor and the executor. Same structural problem as when consulting firms started selling implementation services, or when agencies started owning media companies. The advice gets shaped by the incentives, even when nobody intends it to. The Publicis-Trade Desk split was partly about exactly this kind of pipeline control question.

Basis is positioning Compass as a starting point, with audience insights, third-party activation, and cross-channel optimization on the roadmap. Google is making Gemini the "operating layer" of DV360. The Trade Desk wants an agentic framework for its entire partner ecosystem. Each of these is a bet that the DSP becomes the strategic layer, not just the execution layer.

For the individual media buyer or planner reading this, I think the practical move is pretty straightforward. Get into these tools now, while your organization is still writing the AI policy. The people who understand how to prompt an AI planner, evaluate its outputs critically, and catch the spots where it gets the strategy wrong are going to be more valuable than the ones who build plans manually but slower.

It's a weird spot to be in. The tool that's supposed to threaten your role might be the thing that makes you harder to replace, if you learn to use it before someone above you decides they don't need as many planners.