AI Is Saving You Money, But Might Be Costing You Customers

At Friday, we’ve been deep in it. Using it to do things for clients that simply weren’t possible at this scale before. We’re not neutral observers on this. We think it’s extraordinary, and we’ve built it into how we work.

Which is exactly why the way most brands are deploying it concerns us.

The Thales Digital Trust Index 2026 surveyed more than 15,000 consumers globally and found that 77% of them would not trust a company more for using generative AI.

More than a third said they would actually trust it less.

Only 23% believe companies can be trusted to use AI responsibly with their data. The industry read findings like these, filed them, and kept buying tools. If you want to understand where most AI marketing strategies go wrong, that sequence tells you most of what you need to know.

The case for AI in marketing is real. It saves time, it scales things that wouldn’t otherwise scale, and in the right places, it produces better outcomes than a human team working alone could manage.

The problem is that the places where it genuinely saves are not the same places most brands are deploying it, and the cost of that confusion tends not to show up until it’s already compounded. Not in the licence fee. In the customers who left without explaining why, in the rankings that quietly eroded, in the ad accounts that technically hit their numbers but somehow produced less of the thing the business actually needed.

The Saving That Isn’t

AI deployed to reduce the cost of caring is not a saving. It’s a deferral.

Content is the clearest example. A brand that uses AI to produce thirty blog posts a month because the old target was ten isn’t doing more content marketing. It’s doing more content, which is a different thing. The posts get published. They get indexed, sometimes. They answer questions in the flattest possible way, hitting the keyword, clearing the word count, ticking the brief. The brand becomes part of the noise it was trying to rise above, and it does so gradually enough that nobody notices the moment it happened.

Social follows the same logic. AI-generated captions auto-scheduled across platforms, optimised by a tool that has no understanding of why the brand exists or who it’s actually trying to reach. Impressions accumulate. The comment section goes quiet. Social media works when someone recognises something of themselves in what a brand puts out. A machine calibrated to maximise reach is solving a different problem than that, and the gap between those two problems is where the trust goes.

These register as efficiencies in a spreadsheet. To the person on the receiving end, they register as nothing at all, which is the version that’s hardest to recover from because there’s no moment of failure to point to.

Where the Maths Actually Works

In PPC, AI is largely uncontroversial because it’s doing things no human team could do at the required speed. Smart bidding, automated asset generation, audience expansion, real-time performance forecasting. Google has been making bid decisions faster than any media buyer for years and nobody lost sleep over it. The question was never whether to use it. It was whether the person overseeing the account understood the business well enough to set the parameters correctly and recognise when the automation had started optimising for the wrong thing.

A Performance Max campaign with no negative keywords, no asset-level review, and no understanding of which products actually carry margin is not an efficiency. It’s a process running unsupervised toward its own equilibrium, which may not match yours.

SEO is more complicated. The technical work has always been largely automated. Crawl analysis, log file review, internal link auditing at scale. AI has made all of that faster and more accessible, and that part is fine.

Where it gets expensive is content. The brands that flooded thin AI content into the index ahead of Google’s Helpful Content updates found out quickly that search doesn’t reward volume. It rewards usefulness. Rankings came back for the brands that had something real underneath. They didn’t come back for the ones that had been using content to paper over the absence of a genuine point of view.

GEO is where the most consequential question sits right now. AI assistants don’t surface the brand that published the most. They surface the source that answers the question most clearly and with the most evident authority. The content strategy that works for GEO looks almost identical to the one that worked before anyone started gaming anything. Be genuinely useful. Have an actual perspective. Write like someone who knows the subject rather than someone who has summarised a summary of it. AI can help produce that kind of content. It cannot substitute for having something worth saying.

The Audience Is Running Its Own Numbers

People can’t always name what’s wrong. They don’t say the interaction felt automated in a way that undermined their sense of being valued. They just don’t come back, or they come back with a lower ceiling on what they’re willing to do. Less likely to refer. Less likely to upgrade. Less likely to extend good faith when something goes wrong. That kind of attrition is slower and harder to trace than a bad review, and it tends not to get attributed correctly.

In social advertising, you see this in how audiences respond to creative that’s been generated rather than made. The targeting can be tight. The placement right, the bid efficient. And the ad lands like it was written by someone who had read a description of the brand rather than understood it. A weak CTR in that situation isn’t a targeting problem. It’s a creative problem that better audience segmentation won’t fix, because the issue is in the work itself.

AI personalisation creates a version of this too, in the moment it tips from feeling relevant to feeling like surveillance. An ad that appears for something mentioned in passing doesn’t register as clever. It registers as the brand being somewhere it wasn’t invited. The technology may be within its parameters. The experience it produces isn’t.

Three-quarters of consumers saying AI makes them no more likely to trust a brand is not a sentiment issue or a rounding error. It’s most of your audience telling you that something in how you’re using these tools is working against you. Most brands are logging the saving and not logging that cost, partly because it shows up in softer numbers, in a different quarter, in metrics that don’t sit next to the efficiency figure on the same slide.

The Actual Efficiency Play

The brands getting this right are using AI where it compounds and keeping humans in the work where presence holds value. Those two categories don’t always fall where you’d expect.

In SEO, AI earns its place on the technical side. Processing search data at scale, identifying gaps, running audits that would take an analyst team weeks. It holds less value in the writing, at least without meaningful human input, because the thing that’s hardest to fake in content is whether the person who made it actually understood what they were talking about.

In PPC, the compounding happens in bidding, testing velocity, and audience modelling. The human judgment matters in strategy. In understanding why the campaign is performing the way it is. In deciding whether the metric being optimised is the right one. That kind of thinking requires someone who understands both the platform and the business it’s running on, and that combination has become more valuable over the last five years, not less, precisely because it’s rarer.

In content and social, AI earns its place in repurposing, in first drafts, in getting something workable out of a brief that would otherwise wait in a queue. It holds less value in where the idea comes from, in the point of view, in the cultural specificity that makes something feel like it was made by a person rather than assembled from the category average.

The question running underneath all of it is not AI or no AI. It’s whether the efficiency lands on the brand’s side or the customer’s, and whether the people responsible for the work are actually directing it or watching dashboards that confirm everything is fine until it isn’t.

What the Numbers Don’t Capture

Most marketing dashboards are better at recording what happened than explaining why. AI has created a category of cost that sits outside what those dashboards measure. The audience that didn’t engage. The customer who didn’t return. The brand that shifted in perception over eighteen months and only noticed when something else changed and the recovery was slower than expected.

Search visibility declining over eight months because a content strategy prioritised output over depth. A social following growing on paper while the actual community dissolved. A PPC account that hit its cost-per-lead target while the cost per closed deal quietly climbed. None of these are hypothetical. They’re the shape that deferred costs take when the efficiency was real but the accounting wasn’t complete.

What You’re Actually Buying

AI in marketing is a genuine efficiency in the places where efficiency is the right tool. The question worth sitting with, before the tool gets purchased and the workflow gets rebuilt around it, is what’s on the other side of the saving.

The margin protected on one line is sometimes the margin being spent on another. The second line doesn’t announce itself. It shows up in a review where nothing is obviously wrong, but the trajectory isn’t right. In a pitch where the prospect looked the brand up and found content that reads like it was generated, ads that felt like they were following them, an SEO presence that looks like someone covering ground rather than building something.

It shows up eventually. It just rarely shows up on the slide that justified the decision.

The stat will cycle out of conference decks in due course. The dynamic it describes won’t. The brands that treated it as a real signal rather than an awkward data point will, over time, be fairly easy to distinguish from the ones that didn’t.

If you’re looking for an agency that understands where AI and GEO strategy actually move the needle, and where they don’t, Friday works with brands that want to get this right. Get in touch.