AdCP and AAMP are still competing to define how AI agents buy media. The engineers either one needs are already scarce, and companies that wait for a winner will be hiring at the worst time.
US employers posted 19,297 openings for AI and machine learning engineers in the first quarter of 2026, almost twice as many as a year earlier, according to Veritone data published by Broadbean. Ad tech is hiring from that market while two standards efforts argue over how AI agents should buy media. Many engineering leaders treat that argument as a reason to hold off. They shouldn't. Whichever protocol wins, the work underneath it needs the same scarce engineers, and the companies that start hiring once a winner is clear will be competing with everyone else who waited.
Key findings
US openings for AI/ML engineers rose 94.5% year over year in Q1 2026, and AI roles paid a median $162,240, about 22% more than other IT jobs.
AI skills are now the hardest skills for employers to find worldwide, ahead of engineering and IT for the first time.
Two competing standards for AI ad-buying agents already require platforms to commit engineers, even though little spend runs through either.
In Australia, entry-level roles fell to 1% of digital advertising and ad tech vacancies in 2026, and about half of openings asked for six or more years' experience.
Two protocols, one set of skills
In October 2025, a coalition including Yahoo, PubMatic and Scope3 launched the Ad Context Protocol, an open standard for AI agents that find audiences and negotiate and execute media buys, Digiday reported. It runs on Anthropic's Model Context Protocol. IAB Tech Lab followed with its own framework, now called AAMP, built on OpenRTB, the standard most programmatic buying already uses.
Neither carries much money yet. Adform's head of technology said this year that very little ad spend flows through either one, according to PPC Land, but any platform that supports them still has to pay engineers to build and maintain the integration.
The two specs differ in design. The engineering work under them mostly doesn't. An agent that buys media needs clean audience and identity data, models that bid and pace correctly, and someone who can test whether the agent did what it was told. A team with those skills can integrate with either protocol. A team without them will struggle with both.
The engineers are scarce twice over
That work asks for two kinds of expertise in one person. The engineer has to know how models are built, tested and run in production, and also how the auction works: bidding and pacing at millisecond latency, attribution, consent signals, clean rooms and identity. Good ML engineers are hard to find in any industry. The ones who also understand programmatic are a small slice of them.
The slice is shrinking. AI skills overtook engineering and IT this year as the skills employers find hardest to hire, the first time that has happened in ManpowerGroup's 2026 Talent Shortage Survey of 39,000 employers in 41 countries. In the US, openings for ML and AI architects grew 196.5% to 2,283, the fastest growth of any AI title in the Broadbean data.
Ad tech is bidding against its biggest competitors
Microsoft, Amazon, Google and TikTok were all among the ten US employers with the most open AI roles in the first quarter, per Broadbean. Each runs a large ads business that needs the same mix of ML and auction knowledge, and each can outbid most independent ad tech companies on pay.
The median AI role paid $162,240 in the quarter, about $29,700 more than other IT jobs. AI software developers earned a median $169,052, 16% more than other developers. Openings are concentrated in California, New York and Texas, and New York is home to much of the ad tech industry. Hiring only there means paying the top of the market, which is one reason more engineering leaders now recruit ML and data engineers in Latin America, Eastern Europe and India, where working hours overlap with US teams.
Cutting junior roles makes the shortage worse
Some companies are responding by hiring only senior people. IAB Australia's 2026 talent review of 54 ad tech and media companies found entry-level roles had fallen to 1% of vacancies, Mediaweek reported, and about half of open roles asked for six or more years' experience, according to B&T. No comparable US figures exist, but US companies face the same cost pressure.
Hiring only senior engineers gets this year's work done. It also leaves every company bidding for the same small group, with fewer people being trained behind them, the trade-off NEAR Protocol’s Joseph Spano described when he warned that AI can erode the work that turns junior developers into experts. For agent engineering, a discipline that is barely a year old in ad tech, that is a poor bet. Very few people have years of experience in it, so most of the senior talent ad tech wants will have to be developed in-house.
Each hire takes about a quarter
Technical roles take about 10 weeks to fill, and candidates turn down roughly one offer in four, according to Ashby data we reported on. A three-person ML team hired one role at a time takes about seven months, before any offer falls through.
Teams that move faster change how they staff. They hire for several roles at once. They open roles to remote candidates, which raised offer acceptance for technical roles by 13% in Ashby's data. They bring in contract specialists for work with an end date, such as building a protocol integration, and pair each one with a full-time owner who keeps the knowledge after the contract ends. Many companies outside ad tech already work this way: the contingent workforce in the Americas was worth $4.6 trillion in 2025, according to Staffing Industry Analysts, part of a broader shift toward a four-segment workforce.
Agentic buying may take years to carry real spend, and neither protocol may ever win outright. Neither possibility changes what the work will require. At roughly a quarter per hire, the companies that start building those teams now will have them in place when the spend arrives.
Frequently asked questions
Is there a shortage of AI engineers in ad tech? Yes. AI skills are now the hardest skills to hire worldwide, per ManpowerGroup's 2026 Talent Shortage Survey, and ad tech needs the smaller group of engineers who also understand auctions, measurement and identity.
Which AI and data roles are hardest for ad tech companies to fill? ML engineers for bidding and pacing models, measurement and incrementality specialists, data engineers who handle consent, clean rooms and identity, and engineers who can build and monitor AI buying agents.
What is agentic advertising, and why does it need new engineers? It is AI software that finds inventory, negotiates and places media buys on its own. Two competing standards, the Ad Context Protocol and IAB Tech Lab's AAMP, each require platforms to dedicate engineers to build and maintain integrations.
Should ad tech companies wait for AdCP or AAMP to win before hiring for agentic AI? No. Both need the same underlying skills: audience and identity data, bidding and pacing models, and testing agents in live buying. Those engineers take about a quarter each to hire.
How much do AI engineers cost in the US? AI roles had a median salary of $162,240 in Q1 2026, and AI software developers a median of $169,052, according to Veritone data published by Broadbean.
Where can ad tech companies find AI and data engineers outside the US? Latin America, Eastern Europe and India all have strong ML and data talent in time zones that overlap with US teams.
Should ad tech companies hire AI engineers on contract or full-time? Full-time for the people who own core models and the data platform; contract specialists for builds with an end date, such as a protocol integration, paired with a full-time owner.
How long does it take to hire an ML engineer? About 10 weeks to fill a technical role on average, with roughly one in four offers declined, according to Ashby data.
How can an ad tech company scale an AI team faster? Hire in parallel, open roles to remote candidates, and use contract specialists for time-bound work while full-time hires close.
Talent Observer is sponsored by SiiRA.
