Brands need a human touch more than ever

The consequences of AI run deeper than a few embarrassing headlines.

When the world locked down in 2020, AI felt like a promise on the horizon rather than a tool on the desk. Tech leaders spoke of a future where intelligent systems would cure disease, eliminate drudgery and free people for higher pursuits. That promise sharpened in late 2022 when ChatGPT arrived, followed by a wave of generative tools that could write, design and code on command. For marketers and communications professionals in particular, the appeal was obvious. Here was a way to produce more content, faster, for a fraction of the cost of a creative team.

Four years on, that bargain is looking far less straightforward, and the consequences run deeper than a few embarrassing headlines. They touch how much campaigns actually cost, what creative work is even worth once anyone can generate a version of it in seconds, and how exposed a brand becomes the moment something AI-produced turns out to be wrong, derivative or simply tone-deaf. None of this means AI lacks value. It means the businesses that treat it as a replacement for judgement, rather than an assistant to it, are paying for that mistake in cash, reputation, or both.

When the savings disappear

The pitch behind most enterprise AI adoption has been simple: replace expensive human labour with cheap automated output. That pitch is starting to fall apart under its own economics, and the fallout lands directly on marketing and communications budgets.

Starbucks scrapped its AI-powered inventory tool across North America after nine months, once it became clear the system could not reliably recognise products sitting in plain view on a shelf, an embarrassing reversal for a company that had built its public turnaround story around exactly this kind of efficiency gain. A Pizza Hut franchisee running more than 110 stores is now suing parent company Yum! Brands for $100 million, alleging that a mandated AI dispatch system caused such operational chaos that it actively damaged the customer experience the technology was supposed to improve. For brand teams watching from the sidelines, the lesson is not that AI failed at counting milk bottles. It is that operational AI failures become public, fast, and they are absorbed as brand failures regardless of which department actually owns the technology.

The maths behind the savings story is also wobbling. PwC's 29th Global CEO Survey found that 56% have seen neither increased revenue nor reduced costs from AI, and 22% reported costs going up instead. Microsoft's own internal data shows that once licensing, compute and integration costs are properly counted, running AI agents can cost more than employing the people they were meant to replace.

This matters for any brand or agency that has quietly built its content production model around the assumption that AI-generated assets are simply cheaper. Uber found this out the hard way, exhausting its entire 2026 AI budget by April, while one unnamed enterprise client reportedly ran up a $500 million bill for Claude licences in a single month. The pattern is the same wherever it shows up: unlimited AI use does not scale down cost, it scales it up unpredictably, which is a difficult thing to explain to a finance director who was promised the opposite.

That unpredictability is becoming structural. OpenAI is preparing for an IPO that could value it above $1 trillion, yet remains deeply unprofitable. Financial analyst George Noble has noted the company needs $200 billion in annual revenue by 2030 to justify its current trajectory, a fifteenfold increase in five years while costs keep climbing, with compute commitments alone projected to reach roughly $600 billion by then. Meanwhile, Google has responded by cutting its entry-level Gemini subscription by almost 40%, putting pressure on every competitor to match it on price.

For brands, this price war is not simply background noise in the tech press. It is precisely why the industry is shifting from flat subscriptions to token-based pricing, billing by the unit of compute actually used. A model that means agencies and in-house teams can no longer budget for AI tools the way they budget for software seats. Costs will move with usage, and usage tends to move faster than anyone expects once a team is rewarded for using more of it.

What gets lost when everything is optimised

Even where AI is cheap and reliable, it creates a different problem for anyone whose job is to make a brand stand out. Generative models work by predicting the statistically likely next word, image or frame, drawing on enormous datasets of what already exists. That makes them, by design, machines for finding the average.

Creativity lives in the opposite place: in the outlier, the unexpected choice, the joke that should not work but does. This is not a minor stylistic quibble. It goes to the heart of what strategy actually is. AI is a master of hindsight, brilliant at telling you what a market bought last week and what it is statistically likely to buy next. It cannot tell you why, and why is where strategy lives. It cannot grasp that a campaign succeeds in one country because it signals the brand has arrived, while the identical message bombs in another because it reads as trying too hard. It cannot explain why a joke lands with a British audience or gets them cancelled. Recognising a pattern is mechanical. Understanding what that pattern means to actual people, in their actual cultural context, takes a person.

Marketers themselves are increasingly aware of this gap. Dentsu's global survey of chief marketing officers, published through the World Economic Forum, found that 79% agreed algorithm-driven optimisation is making brands look identical, and 87% believed today's strategy demands more creativity and human texture, not less. That is a striking admission from the people who control marketing budgets.

Algorithms chase the next click because that is the only objective anyone built them to pursue. But brand loyalty is built over years of consistent, recognisably human storytelling, not optimised in a single session. A category where every competitor uses the same handful of AI tools, trained on largely the same data, will inevitably produce campaigns that converge towards a flat, forgettable sameness, exactly when audiences are most fatigued by it.

The reputational exposure nobody priced in

This sameness is already curdling into open hostility, and it is creating a new category of brand crisis that communications teams are not yet equipped to handle.

When Coca-Cola tried to recapture the warmth of its classic "Holidays Are Coming" campaign using generative video, audiences saw straight through it, and the 2025 version was savaged online, racking up roughly 24,000 dislikes against 3,000 likes and turning decades of festive goodwill into a punchline. Levi's discovered a different version of the same trap when it announced it would use AI generated models to increase diversity in its imagery rather than actually hiring diverse human models, a decision that converted a campaign meant to signal inclusion into proof of the opposite. New Zealand label Huffer found out how quickly that kind of misstep can escalate when a former model accused the brand of recreating his likeness through AI without his consent, a dispute the brand handled so badly it ended in legal threats and reputational damage that dwarfed whatever the original shoot saved.

None of these were technology failures so much as judgement failures, the kind that happen when no one in the room stops to ask whether the output actually fits the brand, or whether it will be read very differently by the people watching.

The risk, however, goes beyond aesthetic missteps into outright misinformation, and here the exposure is arguably even greater because errors carry the appearance of authority. Air Canada was ordered by a tribunal to honour a discount its own chatbot wrongly promised a grieving customer, after the airline tried and failed to argue the bot was a separate legal entity it bore no responsibility for. South Africa had to withdraw its entire draft national AI policy after journalists found six of its 67 academic citations were AI hallucinated, referencing journals and studies that do not exist.

Deloitte was forced into a partial refund of a $290,000 Australian government report after it emerged the document contained fabricated quotes and nonexistent research.

These are not obscure technical failures. They are exactly the kind of error that ends up in a customer service transcript, an annual report or a press release, the precise channels through which brands communicate with the world, and once discovered, they do not read as a software bug. They read as a brand that cannot be trusted to tell the truth.

What this actually means for brands

Put together, these are not isolated stories about a chatbot here or a Christmas advert there. They describe a single structural problem.

AI compresses cost in the short term and expands risk in the long term, and brands that build their strategy around the first half of that equation while ignoring the second are setting themselves up for exactly the kind of crisis this article has catalogued.

The answer is not to abandon AI, which remains genuinely useful for sifting through information a human has already gathered, summarising it and making it easier to digest, the unglamorous operational work that frees people up for better thinking. The answer is to keep human judgement at the centre of anything that touches a customer, a campaign or a reputation. That means a person checking the fact before it becomes a hallucinated citation, a strategist deciding whether a joke actually works in this market rather than trusting an average drawn from every market, and a creative director willing to make the unexpected choice that no model would ever predict.

The brands currently retreating from AI overreach are not rejecting the technology. They are rediscovering, often at considerable expense, that the thing customers actually connect with was never the efficiency. It was always the human behind it.

Bylined by Nicole Jacobs, Senior Research Strategist at Irvine Partners