Media buying has shifted from manual knobs to machine-led systems that reward signal-rich inputs and fast iteration. Meta’s Andromeda, Google’s Performance Max, and TikTok’s Smart Performance Campaigns now handle auction decisions, creative matching, and budget routing in real time—provided you feed them the right data and diverse assets. The job of the modern buyer is less about micromanaging audiences and more about architecting the learning environment.
What do platforms now optimise for?
Meta: Andromeda upgrades the ads retrieval layer—personalising which ad is eligible before ranking—supercharging Advantage+ automation and improving ROAS through better creative–person matching. Practically, this favours broad setups with more creative variety over narrow segmentation.
Google: Performance Max centralises inventory and bidding across Search, YouTube, Discover and more. 2025 updates add tighter controls and transparency while leaning further into generative AI for asset creation and testing.
TikTok: Smart Performance Campaigns automate targeting, bidding, and even creative combinations to maximise outcomes with minimal inputs—ideal for always-on performance.
Your operating model for AI-led buying
1. Start broad, structure simple. Use platform-native automations (Advantage+, PMax, Smart Performance) with minimal fragmentation. Over-segmentation starves the AI.
2. Win on assets, not audiences. Ship a modular creative library (multiple hooks, angles, lengths, CTAs). Treat UGC and product demos as “data”—they’re the strongest signals the systems can learn from.
3. Instrument for learning speed. Optimise early for attention and intent proxies (thumb-stop, view-through, engaged-view). Graduate winners to CPA/ROAS targets once stable.
4. Close the loop with modelling. Use incrementality tests and MMM to arbitrate spend across channels and seasonality. Meta’s open-source Robyn is a practical entry point for semi-automated MMM and budget allocation.
5. Automate the grunt work, keep humans on strategy. Let the platforms handle bids and routing; keep human judgment on brand safety, messaging, offer strategy, and creative QA. (Even Google’s recent upgrades emphasise AI for speed with controls for governance.)
A sharp, scalable UGC operating model
1. Brief for moments, not demographics. Map the buyer journey (scroll-stopping hook → demo → proof → offer) and script modular beats creators can mix and match.
2. Ship volume with standards. Weekly drops of 10–20 variants built from a fixed “beat sheet” (5–7 hooks × 3 benefits × 3 CTAs) give Andromeda enough surface area to learn.
3. Measure for learning speed. Optimise to cost per thumb-stop and hook hold for week one; graduate winners to CPA/ROAS targets in Advantage+ once they clear your payback guardrails.
4. Close the loop. Turn top-performing UGC into evergreen templates, then re-shoot quarterly with fresh faces and product updates.
Guardrails that protect performance
Data hygiene first. Clean conversions, deduped pixels/SDKs, consent mode, and first-party event quality directly impact how the AI explores your auctions.
Budget pacing with intent. Constrain learning with phased caps and clear payback windows; don’t yank budgets daily.
Diversity over volume. Ten meaningfully different creatives beat fifty near-duplicates.
Regulatory awareness. Expect more scrutiny on automated ad products; build optionality (whitelists, exclusions, brand rules) into your default setups.
The business case
AI doesn’t replace media buyers—it amplifies them. Teams that standardise simple account structures, systemise high-velocity creative, and validate with MMM/experiments are seeing faster optimisation cycles and more durable ROAS.
The competitive edge now comes from how effectively you design for learning and translate platform feedback into next iterations—weekly, not quarterly.
We build that operating cadence: creator sourcing, modular scripts, disciplined testing, and modelling to direct every next dollar.
If you want your media dollars compounding instead of stalling, it’s time to put AI at the centre of your buying practice—and put a team behind it that knows how to make these systems work.
