Media buyers don’t buy media anymore - they babysit algorithms and run lab experiments at internet speed.

My AI research agent pulled the raw data on this, and the numbers don’t lie. Almost 9 in 10 display ad dollars are now programmatic. US internet ad revenue blew past 250B. This is where the budget lives, and the modern buyer is driving the bus.

Old school buyer: haggled TV rates and shook hands. New school buyer: lives inside platforms, flips switches, writes naming conventions, and argues with models. Think Meta Advantage+ as the daily grind - broad targeting, machine-picked placements, creative slots to feed, and a budget that needs constant pacing so the machine doesn’t set your money on fire. You edit inputs, the algo does the heavy lifting, and you keep it honest.

What does “keep it honest” look like?

• Incrementality - did the ad create sales you wouldn’t have gotten anyway. You run holdouts and lift tests, not vibes.
• ROAS - return on ad spend. Not the pretty dashboard number, the real one tied to cash.
• MMM - media mix modeling. Fancy words for a stats model that asks which channels moved revenue when everything overlaps.
• Privacy sanity - Apple’s “please don’t track me” prompt cut signals. Chrome cookies are in limbo. Either way, you lean into first-party data, broader signals, and clean rooms - a neutral room where you and a platform compare notes without swapping raw user data.

Here’s the part no guru course will tell you. Automation is a black box. It works - until it wanders. You still need brand safety, placement reviews, and the courage to kill your favorite creative because the numbers say nope. You’ll exclude junk inventory, watch for fake clicks, and push for log-level access whenever a partner gets cute with transparency.

Who is this actually for? Teams that can commit. If you can’t run weekly tests, fix tracking, and ship new creative like a heartbeat, the machines will eat you. Small teams still win - pick one stack to master first, like Meta plus one retail network - but you have to be boringly consistent.

The headline says media buying is AI. The truth - it’s half operator, half analyst, half editor. Yes, three halves. Because the job is bigger than the title. 🧪

My take: The new media buyer is a product manager for paid growth - use automation where it helps, and human judgment where it matters.

If you had to bet a year mastering one stack - Meta, Google, or a DSP - which horse are you riding and why?