← The Agile Brand with Greg Kihlström®: Expert Mode Marketing Technology, AI, & CX
The Agile Brand with Greg Kihlström®: Expert Mode Marketing Technology, AI, & CX
#809: Jacquard Chief Product & Growth Officer Toby Coulthard on sounding verifiably human in an AI world
00:28:32
Loading summary
Transcript48 lines
- [00:00]
A
When wildfire risk is high, every second matters. That's why at Pacific Power, we use enhanced safety settings that shut off power in fractions of a second when a potential hazard is detected, helping reduce the risk of sparks. We know losing power can be disruptive, but in moments like these, safety comes first. It's just one of the proactive measures we take to help protect you and your community. Learn more@pacificpower.net wildfire tired of overpaying with DirecTV?
- [00:34]
B
Dish offers a reliable low price every month without surprises. Get the TV you love and start watching live sports news and the latest movies, plus your favorite streaming apps all in one place. Switch to Dish today and lock in the lowest price in satellite TV, starting at $89.99 a month with our two year price guarantee. Call 888Add dish or or visit dish.com today. The agile brand. Welcome to Season eight of the Agile Brand Podcast. This season we're going all in on Expert Mode, MarTech AI and Customer Experience, talking with the people and platforms behind the brands you know and love. I'm Greg Kilstrom, your host and I help Fortune 1000 companies make sense of martech AI and marketing ops. Hit subscribe or Follow to make sure you always get the latest episodes and leave us a rating so others can find us as well. And make sure you check out our sponsor Tech Systems, an industry leader in full stack technology services, talent services and real world applications. For more information, go to teksystems.com now let's dive in as every brand rushes to adopt generative AI. What if the greatest competitive advantage is no longer about speed and scale, but about something unique, uniquely, verifiably human. Agility requires moving beyond the hype of new technology to strategically apply it for true differentiation. It's about being smart and selective, not just fast. Today we're going to talk about a paradox at the heart of modern marketing. Generative AI has promised unprecedented scale and personalization, but for many it's delivering a sea of sameness where brand voice gets lost. We're going to explore how to break free from this generic output, moving from a reactive test and learn model to to a predictive one. And discuss the critical balance of combining AI's power with essential human expertise to maintain brand soul, safety and performance across countless channels. Tell me Discuss this topic I'd like to welcome Toby Coulthard, Chief Product and Growth Officer at Jacquard. Toby, welcome to the show.
- [02:48]
A
Hi Greg, thanks for having me.
- [02:50]
B
Yeah, looking forward to talking about this. Before we dive in though, why don't you give a little background on yourself and your role at Jacquard?
- [02:56]
A
Yeah, my name's Toby. I've been in this space for probably about a decade now. I've worked for a number of kind of CPS or ESPs and joined Jacquard a couple of years ago. My role is Chief Product and Growth Officer. And in that. Well, in terms of what Jacquard does, we generate predictive performance and distribute high performant on brand content for brands and. Yeah, responsible for the product part of the business as well as growing it as well.
- [03:28]
B
Wonderful. Well, yeah, let's dive in. And I want to start with the strategic view here is this concept I teed up in the intro, which is, you know, we're all talking about efficiency gains and just the ability to scale with AI, but what does that do in a negative way to erode the uniqueness of the brand voice? So you've talked about chat, GPT and sameness. From your perspective, what's the underlying mechanism that causes these, you know, admittedly powerful AI models, yet you know, they're converging on such similar, often generic outputs. And what's the, what's the business risk for, for brands that fall into this trap?
- [04:11]
A
Yeah, it's a great question. You know, it's interesting you say these kind of models converging. I think there is an element of these models converging on a similar output. But the reality is 85% of marketers are just using ChatGPT to come, you know, to generate content. And so it's, we're not even talking about using some of these other models. And that, and that 15% includes those who aren't using any AI model at all. And so the risk is, and it's such a hard thing to put your, your finger on, but we can all kind of feel AI generated content. When you see it, if you go on LinkedIn, there's a certain something about it. There's a tone of voice, there's a cadence, there's usage of different linguistic devices that you can just tell that it's, it's AI written and there's this convergence of different people sounding the same. You know, my, my LinkedIn feed is very much a lot of very similar posts and my emails are starting to sound similar. The subject lines that I see or the push notifications I receive are all starting to sound a little bit like ChatGPT. And the risk is not that your marketing becomes less effective, it becomes ineffective because you can't differentiate Your brand, there's nothing unique about it. Brands rightfully are very discerning about how their brand sounds. They've got all these brand guidelines, but they're foregoing that, that quality for that productivity gain. And that's where Jacquard comes in, where we very much care about an authentic brand voice, but we care about performance and what people want to receive. I don't want to receive AI generated content. I want to receive, well, content that feels AI generated. I want to receive content that I resonate with and that feels unique and feels human. And so that's where we come in.
- [05:56]
B
Yeah, yeah. And you know, even for seasoned CMOs, that that idea of maintaining authentic brand voice can often feel a little abstract. You know, how do you translate a strategic goal into a concrete, defensible asset when your teams are using AI tools that they're by design pulling from the same data pools as everyone else? As you mentioned, you know, just the, the extent of using ChatGPT for instance.
- [06:26]
A
Yeah, so that's, that's a good question. So we're not really pulling from the same data set actually. So on one hand, and the way that we're architected is that we have the generative side of coming up with content, which is this, and I won't go too much into it, but this kind of neuro symbolic architecture where we take the best creativity of a large language model and actually a range of large language models, but also apply some kind of deterministic guardrails to it. So that's one side coming up with the content, but the secret sauce is really on the models that we train. So we have predictive models. And so rather than being a generative model, we have a natural language understanding model that uses a huge number of data points, including the responses from the customers, messages that they sent, the opens, the clicks, the conversions, and we use that to really hone the content in a way that means that it's going to drive performance and it's using a unique data set and a unique model that isn't the same as what all these other LLMs are doing. LLMs themselves, ChatGPT, Gemini, whatever they are using doesn't understand what works well. It can come up with some content, but it doesn't know what, what's good, what's bad. So applying this layer of predictive AI on top of that, that can kind of separate the wheat from the chaff, so to speak, means that you can actually take a very data driven approach to coming up with content and refining that output to Be very highly performant and on the authentic voice. Like I say CMOs or brands, they have very comprehensive brand guidelines. They know what they want to sound like. And so it's shaping that and adding an extra layer on top of that and aligning the AI output to those guidelines and ensuring that it is aligned. That's the secret sauce. And that means that we can come up with content that is both on brand safe, performant and distributed to their customers where they need to distribute it to.
- [08:31]
B
Yeah, yeah. So let's get a little more tactical here then as well and talk a little bit about testing and experimentation as well as predicting. So the test and learn mantra is pretty deeply ingrained with many digital marketing teams. You've argued for predicting before you pay. Can you walk us through how that actually works? And how can an AI confident predict the performance of creative variant before it's seen by a customer?
- [09:03]
A
Yeah, so the test and learn thing, that's still very much should be part of a, you know, a marketer's toolkit. But what you don't want to do is just test a bunch of bad variants of language. Right. Like it's, it's. You're just finding the best of a bad bunch. I think at the end of the day, you want to be able to start on, you know, put your best foot forward. You want to test a range of diverse but also performant variants because you'd also. You don't know what's. People cannot look at a piece of content and say, that is going to resonate unless this is, you know, Mad Men. And we, you know, you've got this expert marketer that can, you know, that can look at a piece of content and go, that's. That's what the people want.
- [09:47]
B
Right.
- [09:47]
A
You know, it's impossible to, especially when you're thinking about many different channels and touch points. And there's a scale element of this as well. And so you need to be able to have some level of intelligence. And so our model, based off billions of data points, billions of opens and clicks across channels and languages we've trained these models to understand. Okay, this is actually going to resonate with your customers. And we have all the data to back that up and the model can understand that. And so that's our kind of that predictive layer trained on all that data is what is, I suppose, our secret sauce. It can enable you to still do all the things that you're used to doing, like testing and learning and doing things across multiple channels and touch points across the entire customer lifecycle, but just gives you that leg up and enables you to scale that across, you know, as I say, channels, touch points, journeys and regions as well.
- [10:45]
B
Well, yeah, because the, the Open AIs and Geminis and you know, those large language models, I mean they're, they're also predictive, but in a way, right. They're, they're predicting based on a very generic, you know, they're trained on, on many, many things, but they're, they're trying to predict what someone else likely has done in the past, not what's relevant to your audience on that, on a specific channel and so on. And so, so it's a. They're all predictive. But, but you're, what you're talking about with, with Jacquard is very much tailored to a brand and their audiences.
- [11:17]
A
Right, exactly. So, so most LLMs or all of them I suppose to some extent are kind of fine tuned to be answer format, you know, in, in chatgpt. Give it a question. You know, I think you've probably seen when you're in ChatGPT, sometimes you get the, do you prefer this answer or this answer? You know, that's, that's it kind of that reinforcement learning of giving you an answer that you like. We're doing that, but the, call it the, the cookie that we're giving the LLM, this resulted in an open, this resulted in a click. And so we're fine tuning the output to something that ultimately people are going to resonate with, but we're also ensuring that we're not going so far to stretch away from someone's brand guidelines that becomes clickbait because the, you know, I can come up with an email subject line that everyone's going to click on. It's probably going to say something like you've won a million dollars. Right. And so in that sense there is, there is a framework that you have to operate in and if you go too far on one end, you're going to end up with coming out with content that is very clickbaity, gets opens, but at what cost? You might win the, win the battle and lose the war. And so there's a lot of nuance in both adhering to that brand voice, those brand guidelines, but also maximizing performance in a way that affects all your KPIs, not just opens and clicks, but gets more conversions, increases average order value, increases lifetime value, better retention. All these data points come into the mix when we think about what kind of content is going to resonate with customers. Yeah.
- [12:46]
B
So how does this Change the creative and the campaign planning process, then, you know, does it, how does a role like, you know, copywriter, campaign manager change when you know it's, it's not. AI isn't just there to give me 20 to your point. Okay, maybe if we're lucky ideas, but not, not ones that are, that are likely to perform. You know, how does it change when it's, it's also kind of AI is forecasting success as well.
- [13:16]
A
Yeah, I think about it as well a couple ways. One is property writers today can't actually keep up with the amount of content that's necessary. If you think about the number of channels now that's expected of a marketing team, say the number of touch points and the number of variations of content that you need to do the test and learn stuff. I mean, it's impossible for any one copywriter in any one business. There's always more that needs to be done. And so on one end we enable, we enable copywriters to, to do a lot more. But I also think that the role of the copywriter doesn't really change that meaningfully. You know, you could be an expert in construction but not be laying the bricks. You know, like at the end of the day, a copyright's expertise is in brand voice. It is in how the brand articulates itself to its customers. And so in that sense, this just is just another tool for them to use to be that expert and to really execute better as a copywriter. And so in that sense, it's a tool that they can use rather than something that ultimately will ever replace a copywriter. Yeah, I don't think any copywriter needs to be an expert in typing. Right. For example, for this to work. And so in that sense, they're still a very much necessary part of the process, but they're the ones using the machine rather than being the machine, so to speak.
- [14:42]
B
Yeah, And I mean, in that sense, I think that elevates. I mean, certainly AI is going to augment the work that they do, but I think it actually elevates the role of the copywriter as opposed to, as you're saying, replaces the need for one. Right. They're very much integral to the process, but they're just not needed for hands on keyboard quite so much. Right, right.
- [15:03]
A
It removes the menial element of it and adds intelligence to it as well. They're still pulling the levers and twisting the dials. And especially when you get to things like hyper segmentation and personalization, you know, that that is, that is an impossible task. With copywriter, you know, if you've got a thousand segments across a number of channels and touch points, you're not going to write thousands of lines of copy. Yeah, but we can, you know, Jacquard's able to do that. It's able to, to, to generate hundreds of thousands of lines of copy and personalize that to individuals and ensure that that copy is performant. So in that sense it's a force multiplier.
- [15:45]
B
And so I know we've talked a bit about the multi channel aspect and just the sheer scale of this, but in addition to those, there's also depending on what industry you're in as well and what geographies you serve, there's also compliance issues and things like at. From that perspective, you know, how does, how does an AI system not only do the things we've talked about already, you know, scale with brand voice and all that, but also the nuances of things like compliance and even, you know, cultural nuances.
- [16:20]
A
Yeah, that's a great question. There's a couple things. So one is there is some examples of compliance or cultural nuanced challenges. So compliance side. So we've got a number of customers in healthcare, for example, you don't want to make any health claims about something that maybe doesn't have claims around it. On a cultural nuance side, you know, what's interesting is we have a customer where the tone of voice for that really resonated with boomers, for example, was the opposite of that that resonated with Gen Z. And so the sentiments, the urgency, the length of. And you have to realize that picking a particular tone of voice for your brand may, and in many cases will alienate one set of people and also not only between demographics, but countries as well. I think, for example, I mean, you're American, I assume I'm British. We see between even countries with the same language. The same largely applies to French Canadians and people in France who speak French. What we find is urgency has a massive difference. So in the uk, people do not like content that is highly urgent. They want calm, helpful language. They don't want to have something that says sale ends in 24 hours, buy right now. But in America, that really resonates.
- [17:47]
B
Right?
- [17:47]
A
And so understanding that and training a model that understands the nuance of someone's location, understands what kind of sentiment is going to resonate with the fact that they are 24, based in New York with a certain gender versus someone who's in London who's 45. That can be the difference between really being Successful in the brand and not. And on that compliance issue of making a health claim or, or the regulatory, regulatory issue, especially in both pharma or in, in financial services as well. You don't want to rely on chatgpt that can hallucinate, that can not really understand the value of, or understand the product itself. Not only the value of the product, but the product, it can go off the rails and it's also not distributed into the systems where you want to send the content. And so that's where these kind of neuro symbolic guardrails come in, where we both do use generative AI to come up with content but we have all these rule based filters to prevent something from being sent that you wouldn't want to send and also feeding it with the data to ensure that we're getting this cultural nuance both demographically and geographically as well.
- [18:58]
B
Yeah, yeah, yeah, definitely. I think I had somebody on the show recently talking about the difference between like the highest performing UK ads versus US ads and similar, similar to what you're saying as well. And you know, even just the idea of high context or low context.
- [19:15]
A
Yeah, yeah, yeah.
- [19:17]
B
As a whole, that's probably a topic of a whole other show but definitely, definitely a lot of, of of nuances there. When you're even, you know, country to country in Europe, you know, there's differences between levels of context and all those kinds of things.
- [19:30]
A
And it can, it can go, it can, it can not only not help your brand, it can hurt your brand. You know, I remember growing up in the uk, you would sometimes get American ads on TV and people always wouldn't like it if there was an American voice. And I actually, I'm in the US at the moment, I'm based in New York and you do get a lot of ads with British accents on the TV and Americans don't mind it. And so, so there is a, there is a lot of nuance there that can either help your brand or hurt your brand. And understanding that and taking a very data driven approach to it is ultimately necessary because otherwise everyone says send the right thing, you know, send the right message to the right person at the right time. It's like this, this thing that marketers always say, but they never say don't send the wrong message. Right like that, like not sending the wrong message is as important or more important than sending the right message. And so having that cultural nuance and compliance layer ultimately is incredibly important for a brand to achieve. And you can't just achieve that with ChatGPT. It's not going to help you there.
- [20:37]
B
Right? Well, yeah. And so let's talk a little bit about moving beyond what you've termed conventional AI and what that, what that really means. And you know, what, what's the shift in mindset and you know, we've talked about some of this, I realize, but you know, what's the fundamental shift in mindset and technology required for a brand to go from using AI really as just kind of a workhorse to generate content to a much more strategic engine that actually helps them with their competitive and brand differentiation?
- [21:11]
A
Yeah. So conventional AI is an interesting phrase. So I think pre2022 conventional AI was kind of predictive AI. Yeah, this kind of, you know, a lot of businesses were using data to do things like predictive churn or send time optimization or things like that. And that was kind of more conventional AI. I think now conventional AI has become generative AI because that's the AI the Dow everybody knows. And so ironically that that flip has now been that moving beyond generative AI means that you kind of need to lay a predictive AI on top because that's the thing that ultimately shifts from just like I said, a productivity and efficiency tool to a performance tool. So yeah, so I think in that sense that's what we mean. Like if you combine those two concepts, you kind of get the best of both worlds. We have this multi agent architecture, as I say, whether it's cultural nuance compliance, whether it's ensuring that it's resonating with people on a demographic basis, all of these components are taking the input of what resonates and then feeding that back and then optimizing the system. So in that sense, moving beyond conventional AI means moving beyond generative and actually thinking about what is the outcome of this rather than I, I really hate the productivity efficiency argument a lot of the time because everybody says it and it's so hard to quantify.
- [22:34]
B
Yeah.
- [22:34]
A
And ultimately what are marketers trying to do? You know, no marketing department really gets sets out to begin with to saying let's be as efficient as possible. They say let's, let's, let's improve performance, let's get more engagement, let's get more customers, let's increase lifetime value, let's increase our attention. And so taking that outcome driven approach I think is far more valuable than just saying, you know, we can write a sentence instead of you having to do it.
- [23:01]
B
Yeah, yeah, well, and your point is well taken. As far as, you know, it's funny how people forget that there was AI for you know, like three or four decades prior to 2022. Right. So what, I guess everything old is new again. Whatever to.
- [23:16]
A
Exactly.
- [23:17]
B
Because all of those things that, you know, people in AI are, I'm sure, always rolling their eyes that people are talking about the new, you know, the new thing that is AI. But, you know. Yeah, to your other point, you know, AI is synonymous with generative AI, for better or worse, with most people since they, they were using it, because anybody that's used an algorithm since, you know, anyone that's searched on Google or, you know, done anything, has been using AI for decades. But, yeah, I think to help with the conversation, it's, it's helpful to, you know, to talk in those terms of the consumer adoption and prosumer adoption, you could even say of AI, I would say really started with ChatGPT and stuff like that.
- [23:59]
A
Yeah. And I think people, I think people miss the, you know, we get a lot of, we get a question. It's like, why shouldn't I just use ChatGPT to generate my marketing content? And it's a bit like, well, what are you trying to achieve? You know, at the end of the day, are you just trying to not have to write the subject line or write the push notification or the sms, or are you trying to achieve an outcome on a KPI within your, within your business? And we work with Sephora or Gap or United Airlines and we start with, what are you trying to achieve? What are your marketing goals? And we don't really even think about the productivity and efficiency gain there because ultimately that, that may help the end goal, but the end goal and the outcome is ultimately what we're trying to get to.
- [24:40]
B
Yeah. Yeah. So for, for marketing leaders out there, what's a, what's a first step that they could take to again, get, get beyond some of the, the basics and some of that generic output. You know, is it, you know, a technology change or audit? Is it a process change, talent assessment? You know, what, where, where should they likely look first?
- [24:59]
A
Yeah, it's, it's, it's not that hard. I mean, one, I mean, have a, have a brand guidelines, which every brand pretty much does. Have an understanding of what you're trying to achieve. And we work with customers where, you know, we start with one channel. One campaign, could be an abandoned basket campaign, could be a welcome campaign. You know, it isn't this kind of total transformation of your business that requires this. You start with one. And that also allows us to cut the data, it allows us to refine the output. We Have a team of computational linguists that kind of walk you through the process, that can kind of align the AI's output to your needs. And it's super easy just to get up and running. Like AI in itself, given how many people have adopted it is so natural and so easy to use. And so I think people look at this as, oh my God, this is some kind of massive change management process. But at the end of the day you've got all the right pieces. You just need the technology to enable you to do it. And it is very much a core walk run approach.
- [26:01]
B
Yeah. Well, Toby, thanks. Thanks for joining today. Got a couple last questions before we wrap up here. So first, if we were having this interview one year from today, what is one thing that we would definitely be talking about?
- [26:15]
A
That's a great question. I think AI slop is what everyone's talking about at the moment and I think it's only going to get worse. I think there's this whole kind of dead Internet theory where the whole Internet has either already or will become just a series of bots talking to each other and commenting to each other. And I think increasingly that is, that may well become the case. And so in that sense, how to kind of navigate the Internet, navigate your life surrounded by AI content that doesn't resonate, I think that will become more of a challenge and hopefully Jakarta can be a part of that conversation where we can help brands kind of cut through that, that noise and, and, and not sound like AI slop and not be ignored as much as I think it's going to become.
- [27:03]
B
Yeah, yeah. Well, one last question for you before we wrap up. What do you do to stay agile in your role and how do you find a way to do it consistently?
- [27:12]
A
So that's a great question. So which we're trying to do to go to a lot more events and be a bit more human with our approach. I go to conferences and just understand what people want and how people are talking about things. I don't know about you, but I receive a hundred AI generated outreach emails a day telling me to buy something and we've kind of lost this human, right human part of communicating digitally. And so I try and speak to people as much as I can on a one to one basis and in front of, and you know, get real face time. And I think that kind of grounds us in what we build. I think it helps us really make a product that is a good product. I mean good in both. It's good for the world just as much as it's a good product for people to use and so really understanding what people need and making sure that brands don't kind of fall away into this sea of sameness. So getting out there speaking to people and not kind of getting stuck behind a laptop being bombarded with AI slot.
- [28:20]
B
Yeah. Love it. Well, again, I'd like to thank Toby Coulthard, Chief Product and Growth Officer at Jacquard, for joining the show. You can learn more about Tobi and Jacquard by following the links in the show notes. This episode is brought to you by Tech Systems. They're leaders in full stack tech services, talent solutions and helping companies put it all in action. You can learn more@teksystems.com and thanks again for listening to the Agile Brand podcast. If you like the episode hit subscribe and drop a rating so others can find the show too. And if you're interested in consulting, advisory work, or if you need a speaker for your next event, feel free to reach out. Just visit GregKillstrom.com that's G R E G K I H L S t r o m.com the Agile brand is produced by Missing Link, a Latina owned, strategy driven, creatively fueled production co op. From ideation to creation, they craft human connections through intelligent, engaging and informative content. Until next time, stay curious and stay agile.
- [29:30]
A
The Agile brand. Acast powers the world's best podcasts. Here's a show that we recommend.
- [29:47]
B
What makes a leader worth following? What should you really care about in your job? As technology is changing so quickly, is it just gonna be about machines talking to other machines?
- [29:56]
A
I mean, should you quit your job
- [29:57]
B
and start something on your own, what would that take? What does success and risk look like when we're all at the starting gate together? These are the questions we answer each
- [30:06]
A
week on Lead Human with Jack Myers and Tim Spengler.
- [30:09]
B
Join us each week and subscribe at your favorite podcast platform and YouTube. We'll tell stories, we'll hear from some of the best, and we'll try to figure this out together. ACAST helps creators launch, grow and monetize their podcasts everywhere. Acast.com.