AI in marketing: The benefits and limits

It’s almost impossible to escape talk of artificial intelligence right now. However, the discourse varies wildly. For some, AI offers endless promise to revolutionise the business world and beyond. For others, it is a sinister force of automation that threatens our livelihoods.

Somewhere in between, many view both arguments with a degree of scepticism. AI surely has many benefits that can be applied to the ways in which we work. But even with the undeniable recent advances - especially in generative AI - there are many things it cannot do to an acceptable standard, and it’s unwise to overestimate its uses and ability.

Artificial intelligence has some interesting applications for marketing, and the potential to develop further in the near future. But in many cases, you still can’t beat the power of the human mind and imagination. For the sake of your business and your customers, a realistic understanding of what AI can and can’t deliver is essential.

What AI can do

One of the most powerful functions of artificial intelligence is as a tool for idea generation and research.

Anyone who works in a creative industry will know that sometimes the ideas just refuse to come. You need a spark to get you going – something to build off.

This is where AI can be very helpful. Platforms like ChatGPT or Bard can generate long lists based on a simple prompt. Like an instant brainstorming session, this will leave you with a series of starters – at least one of which is likely to form the seed you need for inspiration.

Similarly, but on a more granular level, AI can help to generate short pieces of simple copy. When it comes to ad or social copy with a clear and concise message, AI is in its element. Sometimes it can be surprisingly challenging to manually create variations on a theme – you soon feel as though you’ve completely exhausted your vocabulary. In these instances, AI can serve as a sophisticated thesaurus, saving you time and effort. It can even help to create a framework for a longer piece – the limits of which we’ll discuss later.

While website chatbots have a reputation for causing more issues than they solve, AI can now be used to power them in ways that can improve efficiency and customer experience. It can help streamline and triage requests by interpreting questions and offering solutions, as long as customers are then connected to a human as soon as they need to be. Avoid trying to trick customers into thinking they are speaking to a real person when they aren’t, or trapping them in information loops that lead to no solutions – this is where chatbots can start harming your customer relationships.

Artificial intelligence can also be a powerful research tool. AI is able to assess data at a vastly faster rate than a human being, and so can extract insights from information that we might never discover otherwise. AI trained on a dataset can create written summaries of that data that are far easier to understand than huge spreadsheets awash with information. AI-powered intelligent search has emerged as a way of harvesting amazing insights from discrete datasets, uncovering information that was there all along but was lost, forgotten or overlooked.

Using AI in this way, we can discover ways of better understanding the market and target audience we are speaking to, making changes to our messaging and services so as to better meet their needs and desires. AI can help us develop deeper insights into customers and the performance of our work.

The limits of artificial intelligence

There is one key thing to remember when considering all the amazing applications of AI: it cannot be left to work unsupervised.

There is a temptation to think that, if AI can be used for writing and research, then we can begin to fully automate those processes. But organisations have already learned to their cost and embarrassment that artificial intelligence is not yet as smart or sophisticated as we might fool ourselves into thinking. Anything generated by AI needs to be monitored and checked by a human being, as it can get things very wrong indeed.

A major issue with AI is that it cannot distinguish truth from falsehood – nor understand contextual quirks of tone like satire or political bias. An AI platform trained on the internet has unbelievable amounts of data to work with, but it has no instinctive human understanding of what is valuable and what is nonsense. CNET discovered this when it was reported that, after the website secretly published content written by AI, more than half the articles created contained factual errors.

Any AI trained on the internet as a whole will be vulnerable to errors that riddle so much online content. As artificial intelligence evolves, finding the data to train it with will be key to its viability.

For these reasons, everything written by AI platforms needs to be checked by a human being, not only for factual errors but also to ensure that the writing is of good quality and makes sense. We are far from a place where we could even consider putting out AI-generated marketing materials that hadn’t been thoroughly checked by a human being.

As we said, there is scope for using AI to create a first rough draft framework to kickstart longer pieces of written content, but these should be carefully checked and rewritten in your own words to ensure the final piece is accurate, thorough and has the correct tone of voice. LinkedIn recently embarked on an interesting approach combining AI content creation with human input and opinion, essentially asking content experts to assess and add to AI-created articles. Not only does this type of approach highlight the contrast between artificial intelligence and the humans who do this for a living, but it also gives the AI platform high quality, accurate information from which to learn over time.

There is another reason that humans still have an edge over AI when it comes to writing, and that is our imagination. Artificial intelligence is trained on existing material, so it can never truly create something original. It doesn’t understand tone or wit. While the copy that generative AI platforms create might at first seem fantastic, you’ll likely find that after a while it tends to be extremely predictable or even nonsensical.

The issue as it stands is not that AI can write better than human beings, but that businesses might mistakenly believe that this is the case, and thus devalue the work of content creators, who remain vital to the continuing success of the marketing industry.

Another issue we’ve noticed is that AI tools can generate duplicate content based on similar prompts. Being trained on the same dataset (in this case, the internet as a whole), it’s perhaps not surprising that this should be the case. But what does it mean to ideas of plagiarism when separately sourced text ends up being near identical? Would the concept of online plagiarism lose all coherence? And that’s before you factor in the potential issues surrounding duplicate content and search rankings.

This also highlights the copyright concerns that are likely to be looming on the horizon – not just for text-based tools but particularly in the case of services like Midjourney and DALL·E that generate images. AI is often a black-box technology that may have been trained on anything harvested from the internet. Artists and other creators are rightly concerned about their work being plagiarised to power these money-making tools. The point is that no one knows whose work is being used or how – nor what lawsuits might be coming down the line.

The scope and uses of artificial intelligence is a fascinating topic, and we at Beettoo think there are many interesting applications for the technology now and in the near future. The key to getting your approach to AI right is to understand its strengths and weaknesses, so that it serves you instead of tripping your business up.