Business leaders need to understand how artificial intelligence will change their business, not in a generation but inside the two to three planning horizons of most corporations.
Our recent Cognito survey included questions on AI. We frequently heard that there’s now a strong desire to to understand how advancements in artificial intelligence will lead to actual business transformation. These are starting to emerge – BT is planning to layoff thousands as it automates customer service and education platform provider Clegg saw its shares collapse on reports its business model could be eliminated by ChatGPT.
Experiments abound
At a recent staff meeting in the London office, our managing director chaired a discussion about how our employees were already using material. Nearly every person, across marketing and communications, had tried out and used artificial intelligence as part of their work.
None of this was mandated and we didn’t even at the time have a company policy around the material. Instead there was a near spontaneous attempt to see how ChatGPT might be able to help with assignments. I asked ChatGPT to shorten a piece of content that was too long. Others created tables, or looked for examples.
Quickly we found things that just didn’t seem to work with the current technology. ChatGPT is useless with anything timely, as at least the free version (without extensions) doesn’t have material from the last 18 months. And there’s a chance if you get too creative, the bot will respond simply by making up material.
Explanatory assistance
The financial markets are abound with complex terminology and very specific phrases. At Cognito, many of us come from financial backgrounds - having written on ABS, or originated DCM transactions or sold payments tech. But in truth, finance is several dozen sub segments all with their own language, and no one can be an expert on it all. Inevitably many of us, like most in FS marketing, have not actually used every specific product. I remember early in my career gog-eyed sitting through demonstrations about straight through processing and order management systems. This stuff isn’t easy and yet we need to develop a comfort with discussions around this material.
I see chatbots as an excellent way to ask more complex questions and have a true discussion. On Google, you can easily find out what SFDR means, or read a 1,500 word article aimed at professionals. AI allows you to drill down on exactly what you don’t understand – how is SFDR different from what is in America? What happens if I work at a particular company? You can ask it specifically to repeat and rephrase things that are hard to understand, and give examples. It is a teacher that never tires, never sees a question as stupid.
Mirco-targeting
We heard in our report that the place where there’s the most space for artificial intelligence to improve delivery is to simply create more content. I understand this. Too frequently we work with companies that want to reach eight or ten different audience segments, but there’s simply not the time and nor the money to create something bespoke for each of these. We’re faced with the choice of either creating something specific for a priority segment, or creating something general enough that it works from everything.
What we hypothetically do with AI is to take a master piece of content and then request re-writes that are based on new segments. We can specifically ask it to include specific message points for each segment, or even feed it ideas for quotes to incorporate into the text. These will need additional polish, but it may soon be possible to take a similar budget and really create more tailored, more relevant pieces.
This in turn can be used to hopefully turbo-charge the engagement rate to paid social advertising campaigns, as the material would be more relevant to audiences. Google Bard will soon be part of Google Ads services to help optimise campaigns and content creation and optimization, a big change from now where services just help with targeting, not customizing or optimizing content and copy.
Long-form content
An ethical and perhaps legal question will be in how and when agencies and freelancers – and also in-house professionals – disclose the use of AI in the creation of material. If someone has paid for a 2,000 word article, is it OK to allow ChatGPT to create the first draft of the material? We would absolutely cite sources and quotes if we used third party opinion, but general research and knowledge typically is not disclosed. Expect additional guidance from industry groups and experts this year.
Interestingly the FT’s editor said last week that no article would ever be written by AI, but that this did not mean that AI could not help in research. AI can help assemble information, prepare outlines and make longer, more bespoke projects easier to handle.
Where AI won’t go – corporate communications
Much of what I’ve been talking about is content that is the cut and thrust of B2B campaigns, across earned, paid, owned and shared media. Sometimes large, big budget and complex campaigns.
But we also (and our clients) write copy that is much more in the realm of corporate communications: a CEO speech, an acquisition press release, andop-ed in a major publication, a Q&A response document for a negative media story. By definition, such copy is extremely important, and so it is perhaps harder to see the scope for AI. It also, sometimes, needs to be produced under considerable time pressure – sometimes in crisis situations being drafted by multiple parties in real time. Certainly, AI can help speed, but in such situations you don’t want a “good enough” first draft to work from, you want the very best first draft in half an hour, not a reviewing process over days. Often such situations involve fewer but more important words, and that is perhaps harder for AI.
Moreover, the spoken word on broadcast or video formats, by identifiable spokespeople, increases in importance as you move upwards into corporate communications. While there are certainly roles for AI in broadcast (for newsrooms and corporates), its function may be more limited than with B2B marketing copy and content.
What’s ahead
There’s so much interest around this topic. We actually didn’t have a single person skip the question we asked about AI. It seems impossible not to have an opinion on the subject.
Where the conversation must now go is exactly where we see benefit. And then we need to have complicated discussions around value. It’s unfair to think we can charge the same price per unit or delivery when costs are falling. Instead we need to have competitive pricing (or costs for those in house) to show measurable improvement. We may also develop company specific AI products, so people are paying to access our proprietary generative algorithms in addition to our consulting services.
The party line among publicly quoted executives and experts is that AI will allow for a retrenchment towards both in house and agency comms people spending more time on strategic advice and creative delivery, and less on administrative tasks. A decade ago, agencies had small armies of junior staff who collected articles, did scans, and created reports. That still exists, but the vast majority of this has been taken up by specialty services and algorithms. Admin, however it is defined, definitely consumes less time and resource than it once did. But whether AI can make further gains is not yet clear.
We at Cognito are dedicated to an open, experimental approach that should yield useful and perhaps unexpected dividends. Yet, it’s important that as an industry we balance experimentation and a voracious appetite to create, with patience and caution. We are only at the beginning of a transformation driven by a technology that promises to be as meaningful as the internet and cloud technology.
Jon Schubin is a director in Cognito’s London office. This article adapts material in Cognito’s recent report “The Future of Finance & Technology: A 2023 guide for marketing and public relations,” which also includes material from Sam Barber and Andrew Marshall