With the likes of ChatGPT and other AI-powered tools hitting the headlines during the latter part of last year and this one, many of the questions asked about these language processing tools are based around the scope of what they can actually do, what they can be used for, and which professions may need to be worried about being replaced by a bot.
Back in June, we put ChatGPT through its paces with a range of academic questions to see how it fared in answering them. You can read our article reviewing its responses to the first four of those questions here. Our findings from this test included that although the bot was able to understand the questions put to it and offer up knowledge based on fact, it struggled in giving a full and insightful explanation of the sort you might expect from a thoughtful human counterpart. It was useful for an initial response to the questions, but fact-checking and sense-checking was needed to elevate the answers to a higher standard.
Now that more people have dipped their toes into the technology, we wanted to look at how this tool and others like it were actually being used in businesses or the workplace and how we ourselves may use it in the future. We are avoiding using it to write copy because of our concerns around the quality and uniqueness of content as well as with a view to keeping our own editorial voice. We are further aware that moral dilemmas have been raised over the charge that AI can produce output that indirectly plagiarises the work of others, and over the risk that it could reduce demand for skilled creatives such as copywriters and designers. This said, we have experimented to assess the capability of modern AI, and are potentially open to using it for sourcing initial ideas or for creating conceptual artwork to illustrate our articles in the future, but we feel that at present, we need to set definite limits on the processes for which it is used.
In this article, we have collated some of the business areas that could use this technology to their advantage.
Let’s start with one close to our line of work. While as previously alluded, we have both moral and quality-related qualms over whether it would be right and desirable to use such technology for our own copy or that of our clients, we shall here explore the areas of marketing where AI seems to have the most potential.
AI-powered tools such as ChatGPT and Google Bard could be used for idea generation for article posts and email subject lines by prompting the bot to provide a list of titles and catchy subject lines around a specific theme or topic. This could speed up the process of developing concepts in both of these areas, and also has potential to increase email open rates if you are able to use AI for analysis of previous subject lines used and open rates for those by using an AI-powered email marketing platform.
Product descriptions and social media posts could be another area where AI might be useful, especially if there is a need to create a number of product descriptions that are similar or to promote the same piece of content across multiple social channels. AI could do both of these tasks quickly and efficiently, ensuring no duplication across product descriptions, and helping to tailor social content to the requirements of specific platforms.
AI-powered Chatbots are already in widespread use by businesses seeking to cut costs on human customer service operators. We will all have come across a pop-up chatbot on a website offering their services over a real-life interaction; and it is now impossible to start a live online chat with many large companies without first navigating through a chatbot input process, as these are used as gateways to filter out people who have questions that can be answered automatically with reference to an existing knowledge base.
In our experience, some of these chatbots are better than others, and many fail to provide apposite responses to questions that do in fact require the intervention of a customer service agent in relation to a particular problem or concern with the provision of a product or service.
What can’t be argued with, however, is that AI chatbots can negate the need for a human interaction when all that is needed is a statement of company policy or a similarly simple informational answer. They will often direct customers to pre-written web pages on topics related to the enquiry, although in our experience, it is very much a matter of hit and miss as to whether those pages actually provide adequate and easy to follow responses to the specific questions asked; and often they don’t.
A well-programmed chatbot should also be able to identify when a conversation does need to be transferred to a real-life customer sales advisor for a conversation, although most seem to be designed to put the customer off from pursuing this option for as long as possible.
The use of AI chatbots as a customer services screening gateway can improve a company’s economic efficiency by reducing workload and freeing up staff time for more important tasks. Tools that offer these services include Zoho SalesIQ, Tidio and Intercom.
The downside is that when these chatbots do not provide relevant answers or are faced with a customer service question requiring a personal investigation of the status of a purchase or other problem with an order, they can cause immense frustration to the user, reducing customer satisfaction levels.
Among the worst user experiences come from companies that offer a chatbot routine that ends with no live chat or email alternative when the customer’s question has not been adequately answered. Some companies make access to customer service both difficult and expensive by offering only premium-rate telephone calls held in a long queue to anyone who is not satisfied with the response of a chatbot. This affords a very poor customer experience by making the individual wait on a call while paying money for the privilege, often raising the true cost of using the company’s service or purchasing one of their products to an uncompetitive level. This is a fast way to lose customers permanently.
To summarise, it is well worth bearing in mind that it is far preferable to use chatbots only judiciously and moderately, and not as an outright replacement for free-of-charge personal customer service contact either by live online chat or by email.
AI chat can also be programmed to offer any multilingual support to customers who are more fluent in other languages than English. Rather than employing a number of different skilled language translators at a premium cost, AI can detect when a different language is being used and respond accordingly.
As well as chatbots making waves, many text-to-image AI tools are causing a similar stir. These tools allow the user to provide a prompt for an image they want created and wait just a few minutes while the technology works to create an impressive version of what you asked for. Midjourney and DALL-E 2 (developed by OpenAI, the same company behind ChatGPT) both offer these services. Midjourney offers a free trial period before various monthly payment plans take effect. DALL-E 2 works by using credits available to buy for $15 – this will give you 115 prompts, with each prompt offering up 4 images. Regardless of the paid models, we think these tools could be good additions for designers to have in their kit in terms of creating desired images fast and effectively.
Photoshop AI Generative Fill is another example of a tool that designers could use to their advantage. Rather than creating a whole image from scratch, this AI tool allows you to both clean existing images and add to them. You can either prompt the tool to add something to an area of the image you have selected, or leave the input field blank for it to remove it. If you have a subscription to Creative Cloud, Adobe Firefly or other Adobe services, this tool is currently free to use. However, this will change from November 1st when generative credit limits will come into play. You can read further information on this here.
The moral considerations around AI creating text content are also relevant to its creation of images and designs, as they are another area of creative output that conventionally depends on skilled human workers. Establishing the line between what is and what isn’t morally right may be a subjective task that each business must make for themselves, bearing in mind how others, including their clients and target markets, will perceive their use of AI if they know how they are using it.
With all AI content generation, at least for now, human oversight is needed, both at the input stage and in determining what to do with the output.
To take the input stage first of all, these tools won’t work without first being programmed and prompted in order to give them directions on how to act and what to deliver. In some case people are now getting AIs to prompt other AIs.
When it comes to the output, none of these tools are perfect or proofed against serious problems, so a critical review and sense check before it is put to use is essential. There is additionally a risk that commonly used AI platforms such as Chat GPT will generate more or less identical responses to different users from different companies asking it the same questions, and even if no-one else asks the exact same question, the output will be written in the AI’s voice and not your own brand voice or the voice of one of your creative copywriters. As a result, in terms of creating unique copy, it is sensible at the very least to think in terms of editing its output, and preferable to view it only as a jumping off point for a copy in your own words.
We rate the technology as having major potential to save time, improve efficiency and increase productivity for businesses that work with creative content of all kinds. However, this technology needs to work alongside skilled humans with a good understanding of the target market for each piece of content, in order for its output to be successfully put to use in practical business applications. It does not replace creatively skilled members of the human workforce, although it may be that many medium-sized and larger companies will be looking to leverage it to cut their costs of employing writers and artists, so job losses in some creative industry sectors seem likely.
We would be interested to hear your thoughts on this discussion and the philosophical and moral dilemmas raised by the use of AI content-generation technology. For example, what are the defining characteristics of creativity? How would you feel as a paying client if your digital agency was leveraging AI for content creation and design? And how would you feel as a website user if you became aware that copy on the website had been generated by AI bots and not by real people? Would you feel it was a waste of your time to read it, or would you still feel you could learn something from it and be happy reading it?