Automation and the UA manager's new job
With Advantage+, Smart+, Unity Vector and AI bidding handling the auction, here is what is actually left for a UA manager to own.
Ask a UA manager what they did five years ago and most of the answer was bidding: adjusting targeting, setting budgets by placement, then watching a dashboard all afternoon and nudging the numbers. Ask what a platform's automated system does with that same job today and the honest answer, across Meta's Advantage+, TikTok's Smart+, Unity Vector and most DSP bidding stacks, is nearly all of it, and better than most humans ever managed by hand. That isn't a controversial claim anymore. The controversial question is what's left for the person whose job title still says media buyer.
The bidding layer is not coming back
Unity's own numbers make the point plainly: Vector passed a billion dollar annual run rate two quarters earlier than the company had guided, and it now represents the large majority of Unity's Grow advertising revenue. AppLovin's growth over the same run of quarters tells the same story from the demand side. These aren't marginal efficiency gains sold with a marketing deck. They're systems that see more of the bidding surface and update more often than a human team monitoring a dashboard ever could, correlating signals no analyst would think to line up, and the platforms have priced and structured their products on the assumption that buyers will hand over the auction-level decisions rather than fight the algorithm for control of them. A UA manager who spends the week manually adjusting bids inside one of these systems is, in most cases, working against the tool.
What is genuinely still a human decision
Budget allocation across channels and platforms remains a human call, because the automated systems inside each platform optimise within their own walls and have no visibility into, or interest in, whether the marginal dollar would work harder somewhere else entirely. Deciding how much goes to Meta versus Unity versus a DSP versus owned channels, and rebalancing that mix as costs and results shift, is portfolio strategy rather than bidding; no platform's automation will do it for you, because no platform benefits from recommending a competitor.
Creative strategy has if anything grown in importance, since an automated bidding system is only as good as the assets it has to choose from. Deciding what concepts to test and how many variants a portfolio needs is judgment that sits upstream of anything the algorithm does. So is calling the difference between a concept that has genuinely fatigued and one that has just dipped for a week. When the bidding layer itself is roughly equivalent across competent teams, this is the lever most likely to separate a strong result from a mediocre one.
Measurement design matters more under automation, not less. An automated bidding system optimises toward whatever event it's told matters, and if that event is a weak proxy for actual value, the system will very efficiently produce the wrong outcome at scale. Choosing the right optimisation event is a specialist skill. So is building the value models that feed it, and so is auditing whether the audience the system has chosen actually correlates with long-term value; automation has made all of that more valuable precisely because it now operates with less day-to-day human oversight of the mechanism itself.
Vendor negotiation has become a bigger part of the job, not a smaller one
As bidding decisions move inside a small number of large automated systems, the bargaining power a studio has left sits mostly in the commercial relationship: how much data access a studio gets, what test budget the platform will fund, whether exclusivity is on the table, and how much visibility a platform is willing to give into why its system made a particular call. Axon's move to global self-serve removed one kind of gatekeeping. It didn't remove the negotiating table, since larger spenders still extract better terms than self-serve alone provides, along with better support and better access to data. That negotiation increasingly sits with the UA manager rather than a separate commercial function, because nobody else on the team understands what a platform's black box actually needs pushing on.
The honest version of this
The job hasn't shrunk. It has moved up a level, from adjusting dials to deciding which dials matter, and a UA manager who hasn't made that shift explicitly, who is still measuring their own value by how well they can out-bid an algorithm, is optimising for a skill the market has already stopped rewarding.
