Go back to the mid-to-late-2010s and though artificial intelligence wasn’t in every other news bulletin like it is now, there was still a lot of talk about AI in marketing. As predictive analytics and natural language processing (NLP) became more commonplace, there was excitement around emerging case studies, such as KLM’s use of automation in customer service. The airline integrated with Facebook Messenger in 2016, sending flight updates and boarding passes, but also using NLP tech to support agents with suggested answers to customer questions. Each answer came with a confidence score and could be sent out automatically if this score passed a certain threshold.
At the Festival of Marketing 2016, KLM’s Karlijn Vogel-Meijer described this tech as “the best of both worlds – a timely answer, a correct answer, and a personal answer. The best of humans and the best of tech.”
For all the innovation, there was also healthy cynicism around this time towards AI hype, particularly the number of martech SaaS products that seemed to have so suddenly become ‘AI-powered’. Or perhaps it would be fairer to say that many businesses didn’t yet have their data in the right shape to let the algorithms loose.
AI, of all different stripes, is now a regular feature in company strategies and annual results outside of the tech industry.
Fast-forward to 2023 and things have moved on. The multitude of feature releases in the world of SaaS martech has been remarkable as vendors seek to add generative AI functionality. When I log in to Canva, for example, one click of a button will transform my creative assets, turning a banner ad into email copy (unfortunately, it will also write poetry if you want it to).
AI, of all different stripes, is now a regular feature in company strategies and annual results outside of the tech industry. In Next’s half year results, for example, the retailer noted how its website modernisation has included improved and personalised product recommendations. More ambitiously, the retailer also references a project that uses AI to help “teams communicate more clearly with customers, in a way that more accurately addresses the nature of each individual enquiry,” adding, “It feels like we are only scratching the surface of what might be achieved.”
Though some may point to the lack of a killer use case for generative AI amongst consumers, on the back-end, some businesses are making strides.
“How can we make AI more productive?”
At last month’s Festival of Marketing, Michael Cato, Product Manager, Campaign Management at the John Lewis Partnership told Salesforce’s Jonathan Beeston, AI is “nothing new to us” and for “the last decade” the business has been using AI across its campaigns.
“If you walk into a Waitrose,” said Cato, “the reason there’s fresh bread on the shelves is because there’s a model [that predicts] when customers are walking into the branch to [buy fresh bread].”
The next step, Cato says, is to ask, “How can we make AI more productive… sitting behind [our] campaigns? And also making sure we can get these out of the door quicker and to the customer. And… then how do we make sure we can have a better conversation with customers, that is more relevant (to them).”
The John Lewis Partnership has KPIs and OKRs in place in its AI strategy for both of these goals, says Cato, productivity and relevance.
If you walk into a Waitrose, the reason there’s fresh bread on the shelves is because there’s a model [that predicts] when customers are walking into the branch…
-Michael Cato, John Lewis Partnership
On the progress in AI models, and the growth in generative AI, Cato remarks that, “AI has been there for people who’ve worked in data, insight and analytics… the core difference that’s happened in the last 12 to 18 months is now… [with generative AI, anyone] can sit there and write some text into a box and get an outcome.”
The marketer’s intelligent assistant has always been a tantalising prospect. Beeston introduced a demonstration of Salesforce’s Einstein within Salesforce Data Cloud, showing how users can talk to a segment generation module in natural language, for example, “create a segment of customers who are loyalty members and fitness enthusiasts”. The segment would then be created by, in this example, looking at customer hobbies in the database and inferring which are fitness related. The segment can then be edited, or Einstein can be asked to explain its actions.
Salesforce Marketing Cloud can also generate imagery, through its integration with Typeface (e.g. adapting product shots to suit a campaign), as well as subject lines and copy (even specifying tone of voice).
One of the arguments against personalised marketing in the past has been that, in a world of dynamic content, the marketer would have to spend a lot more time creating and versioning. With generative AI, there seems to be a solution in sight, even if human oversight and proofing is still part of the workflow.
AI for surfacing institutional knowledge
Cathrine Levandowski is Global Head of Operations at Quintessentially, a lifestyle group best known for its concierge service, organising everything from travel plans to birthday parties for clients. Levandowski oversees technology, including CRM and customer data.
“We’re using AI in our (Salesforce) Service Cloud… it’s our source of truth, for all of our data, [which]… is pulled into Marketing Cloud. So, there we’ve been using predictive AI for quite a while. It’s usually ‘next best action’, or… intelligent search [and] logging. [In the Marketing Cloud], it’s content selection, send-time optimization, subject line generation, we’re playing around with that,” said Levandowski, speaking to Salesforce’s Beeston.
“So, we’ve sort of covered the predictive space, we definitely want to continue to develop that, [but] we’re… definitely looking at the generative space.”
Levandowski described how the business is thinking of generative AI when building its new app member portal on Salesforce, “so that everything we can learn will… contribute to the marketing and service clouds”.
“One of the key things [that’s] different about our approach,” she said, “is that all of our copy is going to be written in a way, in a sort of knowledge article, which will then feed AI later. So that we can be learning from everything that the marketers are doing.”
This approach will help to prevent a “disconnect between customer service… and marketing and how [each department] is communicating [with the customer],” added Levandowski.
“And so this will help us marry up [approaches], and will allow the customer service agents to really understand and leverage… what our marketing team is trying to do and vice versa.”
Trust and humanity are vital
Levandowski emphasises the importance of trust for Quintessentially, both in terms of customer trust (keeping their data safe), and trust in marketing outputs (protecting the voice of the brand). The marketer should be “in the pilot seat”, using “generative AI to assist them, but always being able to refine (the output),” she said.
“Our USP is actually being human, we never plan on having any of our members talking to just a chat bot… For us, [AI] means maximizing efficiency and optimising everybody’s role within the business.”
As marketers adopt intelligent tools and assistants in this way, future Festivals of Marketing will surely see AI start to become just a little less remarkable.
Econsultancy offers a short course in AI for marketing