The harsh reality of AI in marketing

Can I be honest about AI? It looks incredibly promising, but it doesn’t quite deliver what the flashy demos and overenthusiastic hype would have you believe. Like many new technologies riding the giddy crest of a new wave of media excitement, AI hasn’t reached the shoreline.

When I was young, technology promised so much. Sure, we now carry the equivalent of a Star Trek communicator in our pockets — devices that also double as our Hitchhiker’s Guide to the Galaxy — but so much else has yet to materialize.

For example, I don’t (yet) have a jetpack. I saw a team using jetpacks last year, and one of them crashed into a fence, so I may give the jetpack a miss. You get the point — not every technology delivers on its promises. Some just crash you into a fence.

Unlike your postman, AI fails to deliver

Recently, Apple researchers published a paper with the snappy title, “The Illusion of Thinking: Understanding the Strengths and Limitations of Reasoning Models via the Lens of Problem Complexity.” A better title might have been “AI gives up when the going gets tough.” The phrase they use is “complete collapse.” That doesn’t sound like something we can rely on!

Marketing may not be the high-complexity environment where Apple observed AI failures, but are marketers reaping the promised rewards? Not really.

What’s the most common use of AI in martech? Making the systems easier to use. That’s why many AI deployments are simply a chat layer designed to hide yet another poorly designed martech UI.

It’s not all bad

AI is not all bad. It has helped me several times this week alone. AI-generated summaries of meeting notes are a great way to get things done. Even if it’s often another human being who gets tasked with those meeting actions, at least it’s not me. It’s also beneficial for explaining topics I don’t fully understand, which gives me a veneer of expertise.

Image generation is another area where AI excels. My PowerPoint decks would be a snoozefest without some clever AI images.

AI is also buried in the murky depths of many everyday physical products and software-based tools. We’re unwittingly benefiting from AI in various activities, from running Google Ad campaigns to using online translation tools. 

The truth is that there is often nowhere near as much AI as the tool vendor would like you to think. If AI is delivering in some areas, why am I moaning? Because I’m frustrated.

Dig deeper: Your AI strategy is stuck in the past — here’s how to fix it

So near, yet so far

My frustration is that AI gets close but doesn’t quite manage to replace or improve upon what I can do in most marketing tasks. Sure. It will help me navigate the vagaries of a new martech tool, but it sucks when it comes to writing a marketing email. You can’t tell me that you send AI-generated emails and everyone thinks they were written by a human being or worse, that they were written by you personally. You need to improve your communication skills if they believe the latter. You sound like a bot.

That said, AI can generate (notice I didn’t say write) moderately passable text, but it lacks the necessary dose of common sense, awareness and personality that every marketing email needs. I’ve received many AI-generated emails that:

• Claim to be personalized by pulling a line from my company’s website.
• Include text telling me I work for one of our clients (technically accurate, but I work for all of them).
• Or randomly pull a blog post excerpt that is irrelevant to the email.

And I’m sorry (not sorry), often cringey AI-speak just doesn’t sound human. There. I said it. Did that sound human to you?

In my world of B2B technology, AI’s ability to see and replicate patterns isn’t a great way to write about a new technology. It struggles to generate appreciably different articles on any topic because it quickly uses up its limited supply of inspiration and uniqueness. It has only been trained on a limited data set and has nowhere else to go.

I could cite many examples, but you get the point. At first blush, AI looks impressive, but you soon discover it can’t deliver quality at scale. And as it’s a machine, delivering quality at scale should be its most significant benefit. So far, not so much.

There is a lot of discussion about measuring the effectiveness of AI compared to humans. The reality is that AI tends to improve until it approximates the level at which humans perform and then plateaus. That shouldn’t be surprising since human output has trained the AI.

There are attempts to train AI to train AI, but this is fraught with problems. One is overfitting, which means that it’s often the case that using AI-generated training data makes AI’s performance worse.

Do humans have a secure future?

I’m neither stupid nor a Luddite. Unlike AI, I am fully aware that AI will continue to improve. However, the rate of improvement is already slowing dramatically. So much so that, on the evidence, it’s likely that we’ll soon see AI stabilize at the level of an average human for many tasks.

And this is the problem. We’ll end up with average AI models cranking out average marketing campaigns. They’ll be OK and most likely will not bomb, but none of them will stand out and deliver exceptional results, unless there is additional input from an above-average person. 

While AI continues to generate and send embarrassingly wrong marketing emails and suggest formulaic ad headlines, we need to rise. If we can bring above-average creativity and ideas to campaigns, AI will remain a tool that we can exploit in ways that augment our best campaigns rather than replace them with yet another OK.

Ultimately, the benefits AI has been touted to deliver across marketing will be realized, but it will never replace great marketers. However, if you’re not focused on improving your skills, when the AI marketers come knocking on your virtual door, anyone who is OK will have a tough time justifying their continued employment.

Dig deeper: AI fatigue is real, and it’s costing brands more than engagement