Artificial intelligence (AI) has emerged as perhaps the most talked-about topic among marketers this year. No longer a future concern, AI is transforming marketing right now. In this interview for WARC’s 2024 Marketer’s Toolkit, Cathy Taylor from WARC talks with long-time CMO Shiv Singh, who most recently held that role at US online lender Lending Tree.
WARC: Let’s start at the 30,000-foot level. What’s the current reality in terms of the marketing industry and AI?
Singh: I see a huge dichotomy developing. On the one hand, you have folks – and it begins often with the CMO – who are diving in deep. They're pushing their teams, and themselves, to do any and every kind of experimentation, because they see it as transformative, and with the ability to both directly affect the top line, as well as the cost structure within their function.
On the other hand, you have the laggards, and they believe AI, and, more recently, Generative AI, is really about being a tool that can help or enhance work the way the next version of Microsoft Office is going to, but not in any meaningful, transformative sense. My view is in four years' time, there will be 25% fewer jobs in the marketing industry because of AI, or – if it doesn’t reach that amount – at the very least 25% of jobs will have changed dramatically. When you look at every piece of an AI function, from insights-gathering, to brand definition, to audience segmentation, and of the marketing factory flow, it is increasingly getting automated thanks to Generative AI.
Wow, that's rather a large prediction.
Two things are going to happen: there are jobs that are going to completely go away; and there are jobs that are going to get enhanced, and you're just going to be able to do the work a lot quicker, so you need fewer people on the teams. The silver lining is it's creating new jobs and new types of skills. Therefore, hopefully it won't be 25%, if there's enough retraining and there's enough redefinition of a marketing role within a corporation.
What jobs do you think are most at risk?
One of the first examples is in performance marketing, where Google is the giant. Probably 70% or 80% of spend goes to Google in some form or other in the performance marketing space. Every morning, performance marketers in Search are faced with a choice: do they use Google’s automated tools?
It used to be if you were hands-on-keyboard, managing your Google campaigns, you'd perform better than Google would with its automated tools. That is fast-changing. We human beings cannot beat the Google automated ad planning and buying tools, which are getting very sophisticated because they're using AI to create real-time audiences much better, and then to execute against them, based on keyword choices they're making, without, necessarily, human oversight.
The second piece, which is also important, is whether it's Google, or Meta, or Snap, or TikTok, they had tools that enhanced the headline and ad copy optimization. You put in one headline and one set of ad copy, and they would create five other versions, and then do a whole bunch of multivariate testing against it. That's gotten so much more sophisticated, that if you point Google to your landing page, based on that, it will create ten or 15 or 100 versions of ad copy and headlines. It'll go run those and then come back and tell you which is more efficient.
It's smarter, it's quicker, and it's obviously much cheaper to do using the automated tool. What does it mean in turn? One thing is more money can go towards working dollars, so the actual spend. The second is it can move marketers up the value chain where we're less doing the tactical planning and buying and can do more strategic activities. Then, third, not in every organization do you have enough runway to do more strategic things. It means you just need fewer people. No one wants to talk about it, but it is headed in our direction.
For all these years, people have said, “You can optimize against this, that, the other,” but it was way under-utilized before because there just weren't enough people to do it. Now it's possible, but not because of humans.
That's a big area where we're going to see the earliest and the most dramatic change separating those leaders and the laggards. The laggards will want to be in denial. The working teams will downplay the change themselves. If I was on the team, and I was early or mid-career, I would, too, because it puts my job or my team at risk.
But the leaders will go ahead first, because they're the ones who will be chasing revenue and margin and profitability. And it will be on them to drive the retraining for their teams to help take everyone into this dynamic future.
Can you talk about how you have used AI tools on teams, and what you've seen as the progression versus, say, two years ago?
Until recently, AI and marketing was primarily in two domains. In analytics, there were AI tools used to help sift through the mountains of data, and glean insights, and take us in the direction of doing more and more predictive analytics. Sometimes, the responsibility would sit with the CMO and the marketing function; sometimes, it would sit outside and then be with a separate data analytics team. But that was the emphasis of the use of AI.
Outside of that, it was a slow, steady progress: whether it was Google, or Bing, or TikTok, there was AI behind the scenes, helping and enhancing the planning, the buying, the segmentation, the copy, for headline optimization and multivariate testing, all pretty basic stuff.
On November 30 last year, with the launch of ChatGPT and Generative AI coming into the forefront, we started to see some very significant shifts. This is having an impact, and some of the most critical areas are around SEO, so any company that has a large SEO presence or depends on that to drive acquisition, they're finding everything from the most basic scenarios of keyword management, headline creation, article writing – even on, let's say, the first draft – changing to more AI co-piloting help.
Now, there's a huge range in terms of the quality of the output from those first drafts using AI tools. What we've also seen is quality is heavily dependent on our prompt engineering skills; they are probably nowhere near the level of sophistication they can be. But the more sophisticated we get in prompt engineering, the less actual creation of text, images or video we need to be doing ourselves.
A very different example is when it comes to brief writing. If you take ChatGPT-4, and upload your customer research, and ask it: “What are the four or five sharpest insights that you would glean that could lead to a differentiated campaign down the road?” Then, guess what? You will get back thoughtful insights that then can go into a brief. We're seeing that more and more now.
Will it be like having a new team member who is better than everyone else? Not yet. But the right AI tool, with the right prompts, with the right large language model behind the scenes, and the right data, will produce an above-average result. And that's still incredibly powerful and incredibly frightening.
What that then translates into is, when you're going and creating the work, there are all kinds of manifestations of how it can happen. A simple, differentiated example is: I was talking to a friend at one of my agencies, and he mentioned they were pitching for a big piece of creative work.
They had a very short timeline. It was August. They did not have enough people in the office to actually do the work, so they turned to a Generative AI solution, and they actually, through a series of prompts, over a three-hour period, produced creative they felt strongly enough was in great shape to present to that client as a part of that pitch. Now the trickier question, which I did not get a clear response on was, did they tell the client this was AI-generated or not? I suspect they actually did not, but that's a huge game-changer now for an agency. Was it good enough work to run? Probably not, but definitely good enough for a pitch.
The other question is, did they win the pitch?
So, what becomes the differentiator, then, for agencies? Obviously, they're not going to all have the same data to input, they're not going to use the same AI tools, but you can see the playing field potentially level a bit.
There was a study led by Harvard which addresses this issue. The conclusions weren't specific to creative, but these AI tools collapse the spread between extraordinary talent and low performers who shouldn't be in an organization. It takes everyone who would be a low performer and makes them just above average. So, you still need your “A” players, but it lifts up everybody else and brings them closer to that “A” talent. What this means is, among agencies, there will still be differentiation.
In terms of, say, media planning, what are you seeing that gets to the nuts and bolts of that discipline?
It's doing a few things. All of a sudden, the media planners and buyers are doing more creative themselves, because of the tools I mentioned earlier. They're doing more work with the ad copy. They're doing more work with the visual creative and, definitely, the headline pieces. Secondly, a new crop of tools have surfaced, which literally collapse the three pieces of audience creation across platforms, creative output, and the planning and buying into one seamless flow using machine learning to analyze the data much more deeply and provide better recommendations.
Now, obviously, these used to be separate pieces. You now have these tools that manage the entire journey and don't need a human, other than to sign the purchase order.
That's a very different future.
Another area is when you use the AI as a true CMO co-pilot, not for your teams, but for yourself.
On 9/11 this year, the betting platform Draft Kings ran a campaign, with the hashtag #neverforget, encouraging people to bet on three New York sports teams winning their 9/11 games. I get the whole thing about culture-jacking and real-time marketing, but it was trivializing a sacred, really difficult day.
It took a couple of hours for Draft Kings to get pushback around it, and they pulled the plug on the campaign, as they should have.
As an exercise, I took the exact campaign, and I asked Chat GPT, Perplexity.ai, and one or two others, “Is this a good campaign to run?” And they all said it's a terrible idea.
The AI also listed out why it was a bad idea, and dos and don'ts for any future campaign. When you think about that, well, firstly, make sure I use this tool now forever onwards, but then, isn’t this supposed to have been coming from a team member?
What do you see as the best uses of AI as a CMO co-pilot?
Where it's about to get extremely interesting is when you truly use it for idea and strategy generation. This can sound like science fiction, but there’s some research that came out of the Wharton School recently that shows it isn't. This research gave 100 MBA students homework, and they had 24 hours to work on it, which was to come up with one idea for a product that will be $50 or less, that you'd sell to other college students.
At the same time, they asked GPT-4 the same question, but they only gave it two hours, after giving it a series of prompts to get there. At the end of this, there were 200 ideas, 100 from the MBAs, 100 from GPT-4. They then took these 200 ideas and put them in an anonymous online survey, statistically significant, to see which ideas bubble to the top. The AI won. It had more, better ideas than the MBA students. That's frightening, because it turns the whole notion that these are just tools like Microsoft Word on its head. So, then you say, “If it's doing all of this for me, what other ideas can I request of it to get stronger thinking out of it?”
What do you think the marketing industry is missing right now in terms of AI?
As an industry – and this is one of the reasons I'm currently doing a book – is that it has a structural challenge. We're nowhere near as data-driven as we ought to be by now.
At the heart of all these large language models, like ChatGPT, and the AI space more broadly, is access to data, and having strong data, whether it's publicly available or private in your organization.
As an industry, we don't care about organizational data strategies. As CMOs and marketers, we need to care about that. We need to advocate to the CTO and the CIO, so that our needs and use cases are front and center, so we can say, “We need these five repositories stitched together so that we can run these prompts and get outsized results.”
Another big challenge, and sometimes we don't like to admit it, is we get a lot of our education from the platforms: It's Meta, it's Google. They're not invented to transform marketing organizations; they are invented just to sell their narrow suite of goods. So, we're not learning as quickly as we normally would.
As a performance marketer, if you're investing in Google, it's going to tell you what the best things to do are within Google, but that doesn't always mean that's the best thing to do for your brand.
And it certainly won't tell you how AI can be a CMO co-pilot, though it will tell you a whole slew of things you should be doing as a junior or mid-level performance marketer. At a C-level, we won't see that.
The third challenge is we don't want to change.
AI is going to be even more complicated, with even deeper, ethical questions than the social media marketing transformation. It's going to not just change how your consumers are consuming media, but how you need to organize internally, the number of full-time employees you need, and where you put emphasis and what your annual planning looks like. We don't want to do that. It's a hard change.
We've seen this movie before, but this is even bigger. If everybody's worried about defending their jobs, a lot of emphasis is going to go into saving one’s skin.
And, compared to the previous decade, we're in a really high interest rate environment, which, in very practical terms, means the CFOs are more focused on the expense structure. They're the ones who will probably drive more of the AI marketing changes than the marketers themselves as a result.
What's going to happen is a lot of CFOs are going to read the headlines around Generative AI, and they’ll say the two functions that will be most impacted by Generative AI are customer service and marketing. And they're going to come after the CMO and say, “If it's doing all this stuff for you, it means you need fewer people.” Unfortunately, that's going to be the most immediate, practical effect.
But, still, this is incredibly mind-blowing technology. Generative AI, but even before that, AI in general, is life-changing. Channeled in the right fashion, with appropriate co-piloting, and the right leadership, it can help us resolve a lot of different challenges in the country.
AI is now helping us solve some of the greatest problems of our times in medicine, in health. There are challenges, but it has the potential to make all of our lives and our society much stronger if managed right.
Singh publishes the "Being Savvy in the AI Era," Substack, which can be accessed here.