Lloyds Says Not Every AI Use Is Worth It. I Agree.

Faryal Raza Bhatti

3 days ago

6 min read

Image Generated with Chatgpt showcasing this Blog Intent

Honestly, this was one of the most refreshing things I have heard a big brand marketer say all year.

At Marketing Week's Festival of Marketing on 14 October, Lloyds' chief customer and experience officer, Suresh Balaji, stood on stage and asked the audience to put a hand up if they thought the bank's AI videos were low-quality AI content. Plenty of hands went up, and he told them to be honest.

Then he admitted the bank had got it wrong. As Marketing Week reported, he said: "We very quickly realised that we were degrading our brand by creating this."

A major UK bank, saying out loud that its AI content experiment hurt the brand. That does not happen often. And I agree with him completely.

What Happened With the Lloyds AI Videos?

In June 2026, Lloyds posted a series of AI-generated videos on its social channels featuring its famous black horse. In them, a miniature AI horse was carried into an office in a tote bag, drank water from a glass and knocked things out of a pen pot.

Here is the interesting part. Many customers loved them. On engagement alone, the videos probably looked like a win.

But the team felt the content did not fit what the brand stands for, which is stability and trust. So they decided they would not do it again.

That is a hard call to make when the numbers look fine. It is also the right one.

Why Can AI Content Damage a Brand Even When People Like It?

Because engagement and brand meaning are not the same thing.

A brand like Lloyds has spent decades building one idea in people's heads: this is a safe place for your money. A cute, slightly uncanny AI horse being silly in an office can get likes and still chip away at that idea. Every piece of content either strengthens what people think about you or weakens it. Nothing is neutral.

There is a second problem. AI-generated video now has a recognisable look, and audiences are getting good at spotting it. When people spot it, a lot of them read it as cheap or lazy, whatever the actual cost was. For a brand built on trust, "cheap" is a dangerous signal to send.

Richard Shotton has written about this in Marketing Week as AI's "beige trap" for creativity, and it is worth a read if you are planning AI creative work.

So Where Does Lloyds Think AI Does Belong?

This is the part of the talk that got less attention, and it matters more.

Balaji was not anti-AI at all. He told the room that marketers who have not built an AI agent yet are falling behind, and that you can learn how to do it on YouTube. His advice was to find the people in your organisation willing to lead on AI, and to learn through experiments.

Where Lloyds has gone big is anti-fraud. Balaji said the bank has invested £100m in anti-fraud technology, using agentic AI, and has stopped more than £1bn of fraud.

This is not a bank that is cautious about AI. Earlier this year, Lloyds said generative AI delivered £50m of value to the group in 2025, with more than £100m expected in 2026. In one internal test, three teams worked on the same product launch brief: one with no AI, one using AI as a helper, and one using only AI. The AI plus human team won.

So the lesson is not "avoid AI". It is "be very deliberate about where you use it".

Where Should AI Sit in Your Marketing?

I trained as a software engineer with an AI major before I moved into marketing, so I am not someone who needs convincing that AI is useful. But that background also makes me more careful about where it goes, not less.

Here is the split I use:

Use AI confidently	Use AI carefully	Keep humans in charge
Data analysis and reporting	First drafts of copy	Brand-defining creative
Search Console and CRO insights	Ad variations for testing	Your brand's mascot or icon
Research and summarising	Image concepts and mood boards	Anything about trust or values
Workflow automation and agents	Personalisation at scale	Crisis or sensitive comms

The left column is where AI saves real time with little brand risk. I wrote a whole guide on using Search Console reports in Claude for CRO and SEO, and that kind of work is where AI earns its place.

The right column is where your brand lives. If an AI output is the thing customers will remember you by, a human needs to own it.

Do Marketers Need an AI Policy?

Yes, and Balaji made this point too. In the absence of industry-wide rules, he said, brands need their own governance. Lloyds has an AI ethics committee that reviews how the company uses AI.

You do not need a committee if you are a team of one or five. But you do need a simple, written answer to three questions:

Where will we use AI, and where will we not?
Who checks AI output before it goes public?
Will we tell customers when content is AI-generated?

Platforms are already moving on that last one. I covered LinkedIn's approach in LinkedIn AI Content Policy 2026.

The data suggests structure pays off. Marketing Week reported that high-performing marketing leaders are four times more likely to restructure for AI, and twice as likely to work from a central AI roadmap.

My Honest Take

"Can AI make this?" is the wrong question in 2026. The answer is almost always yes.

The better question is: "If our best customer knew AI made this, would they think more or less of us?" If the answer is less, do not publish it, however good the engagement looks.

I respect Lloyds for saying this publicly. Most brands quietly delete the experiment and move on. Admitting it failed, explaining why and sharing what worked instead is exactly the kind of honesty our industry needs more of.

What Should UK Marketers Do This Week?

List every place AI touches your marketing. Content, ads, reporting, customer service, everything.
Sort each one into the three columns above. Be honest about anything customer-facing that defines your brand.
Review your last five AI-assisted posts. Would you be comfortable if customers knew how they were made?
Write a one-page AI policy. Answer the three questions above. It does not need to be perfect.
Build one small agent or automation. Pick a boring, repetitive task such as weekly reporting. That is where AI pays back fastest.

alaji summed it up well: AI can lift the floor of what your team produces, but only people lift the ceiling.

Frequently Asked Questions

What happened with the Lloyds AI black horse videos?

In June 2026, Lloyds posted AI-generated social videos showing a miniature version of its black horse in an office. Many customers liked them, but the bank decided they did not fit its brand values of stability and trust, and said it would not repeat the approach.

Can AI-generated content hurt a brand?

Yes. AI content can perform well on engagement while weakening what people associate with a brand. Audiences increasingly recognise AI-generated video, and many read it as low effort, which is risky for brands built on trust or quality.

Where is AI most useful in marketing?

AI is most useful for data analysis, reporting, research, workflow automation and AI agents. Lloyds, for example, uses agentic AI in anti-fraud work and reports more than £1bn of fraud stopped after investing £100m in the technology.

Should marketing teams have an AI policy?

Yes. Even a one-page policy should say where AI will and will not be used, who reviews AI output before it goes public, and whether customers are told when content is AI-generated.

Does AI plus human beat AI alone in marketing?

In Lloyds' own test, three teams tackled the same launch brief: no AI, AI-assisted and AI only. The team combining AI with people produced the best result.

What is AI slop?

AI slop is a slang term for low-quality, generic content produced quickly with AI tools. It usually feels repetitive or uncanny and adds little value for the audience.

Quick Reference: Lloyds and AI in Marketing

Fact	Detail
AI black horse videos posted	June 2026, Lloyds social channels
Where Lloyds admitted the misstep	Festival of Marketing, 14 October 2026
Lloyds investment in anti-fraud tech	£100m
Fraud stopped	More than £1bn
Generative AI value to Lloyds Banking Group, 2025	£50m
Expected generative AI value, 2026	More than £100m
Best result in Lloyds' three-team test	AI plus human team
High performers restructuring for AI	4x more likely than under-performers