Brands and the Growing Flirtation with Artificial Intelligence (AI) Copywriting

Reports are surfacing everywhere (written perhaps by AI copywriters themselves) of a new form of Artificial Intelligence copywriting software or service that, while not actually threatening to take the place of human copywriters just yet, might be uncomfortably nipping at their heels. It started when the internet became all a-twitter on the subject of AI copywriting when JP Morgan Chase first announced in a press release a five-year, enterprise-wide deal with Persado, the leading company to use AI to generate better performing marketing creative. ‘The expansion to cover marketing creative that will reach,’ in their words, ‘millions of current and potential customers,’ comes in the wake of a successful pilot programme of the new software. According to reporter Michelle Cheng, in tests, JP Morgan Chase found that Persado’s machine-learning tool crafted better ad copy than its own writers could muster, as measured by the higher click rates – more than double in some cases – on digital ads for Chase cards and mortgages.

It’s All Semantics

From a distance, AI copywriting seems simple. An artificial intelligence engine helps precisely guide marketing managers to identify the best semantic choices in relation to the target audience and in relation to the message that brands want to convey. Then, an AI-powered message machine which understands language, breaks down marketing creative into its critical elements, namely narrative, emotion, descriptions, calls-to-action, formatting, and word positioning. Every word makes the right emotional appeal with Persado’s AI-powered knowledge base of more than 1 million tagged and scored words, phrases, and images in 25 languages. (Point of note: the average human, i.e. copywriter possesses a word count of between 20,000 and 35,000 words.) The end-result is especially successful in writing email subject lines, headlines, calls-to-action, etc., especially as it pertains to SMSs and email campaigns, and way more successful than those written by their human counterparts, but less so when it comes to bodycopy.

A Monkey with Time on His or Her Hands

The debate surrounding non-humans occupying the writer’s seat – I include here the copywriter’s Herman Miller Aeron chair – has raged on for as long as anyone can remember. The ‘Infinite Monkey Theorem’ aside (the idea that a monkey hitting keys at random on a typewriter keyboard for an infinite amount of time will almost surely type any given text, such as the complete works of William Shakespeare), AI copywriting is not as far-fetched as it seems. Software generated journalism, for one, is already updating Americans daily of the latest Baseball scores. An automated wordsmith programme appropriately named ‘Wordsmith’ by tech company Automated Insights, is programmed (by humans) to write corporate earnings reports as well as minor league baseball recaps for the Associate Press, thereby, in their words, ‘freeing up journalists to focus on more newsworthy reports’. If you have read a recap on MiLB.com, you have inadvertently doffed your hat to a ‘bot journalist.’

According to ‘real’ journalist Will Oremus, an automated minor-league baseball recap resembles anything written by a competent human reporter, but without the ‘tell-tale signs of abject boredom, self-loathing, or stifled literary ambition.’ Wordsmith, while not well-schooled on the anecdotal front, if at all, is nonetheless trained to quickly convert data into prose, and is best suited to stories whose key elements can be quantified in a spreadsheet. Missing mostly, however, are anecdotal mentions which AI software seems as yet unable to achieve with any real believability. Qualitative observations about the world are harder to accomplish for a machine. Automated Insights admits this fact. Wordsmith’s output, as is the case with any ‘bot journalist,’ is mostly quantitative in output, and struggles to make the type of qualitative observations which humans make easily. While bots may excel at ‘who,’ ‘what,’ ‘where’ and ‘when’ (the stuff of spreadsheets), it is the ‘why’ that humans excel at.

In a Word

So, can a monkey type King Lear? Not really. But a monkey with infinite time on his or her hands (not to mention infinite data storage) probably could – in theory at least. Can a machine write persuasive headlines? On that score, too, we are inching closer. Turns out without realizing it, AI and ML are already prevalent in much of what we do in our day-to-day life. Grammarly, a popular writing program for one, utilizes machine learning in the same way that Persado does. And every time you compose a new email in Gmail you are utilizing AI. In fact, Grammarly even boasts a so-called tone-detector, appearing to make qualitative assumptions about the tone of a text. And then there’s Facebook’s DeepText, a text-understanding engine that Facebook claims can understand ‘with near human accuracy the textual content of several thousand posts per second.’ It is used, according to the platform, to detect intent – again a qualitative emotion utilizing quantitative data. And hot off the linguistic press, there’s the nascent but promising Translatotron system which claims an increasing ability to translate speech from one language to another, while still maintaining the voice and tone of the original speaker, an attribute sorely missing from most speech translating systems.

The Cockpit is the Limit

Of course, there are many aspects in life in which AI and Machine Learning have become enmeshed and which are becoming too ubiquitous to mention. Take self-driving cars and machine learning programs that determine ETA for rideshare companies like Uber or Lyft. And what about airplanes, those ubiquitous pre-Covid-19 contraptions that we used to fly on with great regularity and most likely will again. It has been reported that a typical flight of an airliner involves only seven minutes of human steered flight, reserved typically for take-off and landing. The rest is left to AI. In fact, it is perhaps this analogy that might help inform our prediction on AI copywriting: A percentage of composition (the take-off and landing part) will still require human oversight, for now, provided by humans with an inbuilt system which still boasts the most advanced known qualitative ‘machine learning’ capability. The rest will be left to our non-flesh and blood compatriots.

* 

Tactis is a strategic customer experience agency specializing in the integration of all aspects of customer experience, from customer experience strategy – enabled through user-centric digital products and advanced contact center services – to the employment of emergent artificial intelligence-based technologies. Craig Strydom is a human copywriter and creative director working to enhance customer experiences through a combination of ideas and words. For more information on how Tactis's suite of artificial intelligence-based technologies can help your brand, contact us today.