Why AI in Google Ads Makes Commercial Judgment More Important, Not Less

There's a dangerous narrative spreading through the Google Ads world: that AI and automation have made human expertise obsolete.

The pitch goes something like this: let the algorithm learn, trust the machine, step back and watch the magic happen.

It's compelling. It's also wrong.

The Automation Paradox

Here's what most advertisers miss: the more automated Google Ads becomes, the more commercial judgment matters—not less.

Why? Because automation optimises for what you tell it to optimise for. And most advertisers tell it the wrong thing.

When you set a target ROAS of 400% and let Performance Max run, the algorithm will dutifully find you that 400%. What it won't tell you is whether that 400% is coming from:

• Customers you would have acquired anyway
• Low-margin products that hurt your business
• One-off purchasers who never return
• Brand searches that cannibalise organic

The algorithm doesn't care. It hit the target. Job done.

AI Amplifies Your Mistakes

Think of Google's AI as an extremely efficient executor with zero commercial sense.

If you point it at the wrong target, it will hit that target with terrifying efficiency. If your account structure is flawed, AI will scale those flaws. If your bidding strategy ignores margin, AI will happily burn through your most profitable products to chase volume.

This isn't a flaw in the AI. It's working exactly as designed.

The flaw is in assuming the AI understands your business. It doesn't. It understands signals—clicks, conversions, values—and it optimises for whatever combination you've configured.

What AI Cannot Do

Here's a partial list of things Google's AI will never understand about your business:

Cash flow timing. AI doesn't know that a sale today at 300% ROAS beats a sale next month at 400% ROAS if you need working capital now.

Customer quality. AI doesn't distinguish between a customer who returns monthly and one who buys once on discount and disappears.

Margin variation. Unless you're feeding margin data (and most aren't), AI treats a £100 sale of a 60% margin product the same as a £100 sale at 15% margin.

Competitive context. AI doesn't know that your competitor just ran out of stock, creating a temporary opportunity worth chasing aggressively.

Brand value. AI will cheerfully cannibalise brand equity to hit short-term targets, because brand equity doesn't show up in conversion data.

The New Skill: Commercial Translation

The advertisers winning today aren't the ones who've mastered Google's interface. They're the ones who've mastered the translation between commercial reality and algorithmic instruction.

This means:

• Knowing when a ROAS target is actually the right metric (rarely)
• Understanding how to structure value signals that reflect true business value
• Recognising when the algorithm is optimising for the wrong thing
• Having the judgment to override automation when the numbers look good but the business outcome is bad

This is harder than the old way of doing things, not easier.

The Expertise Shift

Ten years ago, Google Ads expertise meant knowing which match types to use and how to structure ad groups.

Today, Google Ads expertise means understanding the commercial implications of algorithmic decisions that happen at a scale and speed no human could match.

It means asking questions like:

• Is this account structured to give the algorithm the right signals?
• Are we optimising for efficiency or growth, and do we know the difference?
• What is the algorithm actually learning from our conversion data?
• Where is automation hiding problems that will surface later?

These are commercial questions disguised as technical ones.

The Agency Implication

This is why the "set it and forget it" agency model is so dangerous.

When an agency tells you they've "let the algorithm learn" and everything is running smoothly, ask them:

• What is the algorithm optimising for, and why is that the right target?
• How do you know the conversions it's finding are good for the business?
• What would you change if margin mattered more than revenue?
• Where is the algorithm potentially making decisions you wouldn't make manually?

If they can't answer, they're not managing your account. They're just watching it.

What We Look For

When we audit accounts, this is one of the first things we examine: is the algorithm being given the right instructions?

We look for misalignment between what the business actually needs and what Google is optimising for. We look for structures that accidentally reward the wrong behaviour. We look for cases where automation is working perfectly—and hurting the business.

Because in modern Google Ads, the risk isn't that AI won't work.

The risk is that it works exactly as configured—and nobody notices the difference.