RJ (PMN): Hi this is RJ Lewis here for another episode of Pharma Marketing Networks Podcast. I have with me today Tom Ollerton from Automated Creative which is doing some really fascinating things when it comes to applying artificial intelligence to the creative development process in measuring process. Welcome Tom.
Tom (AC): Thanks so much for having me RJ this is really brilliant to be invited on a podcast looking forward for a chat.
RJ (PMN): I’m glad you could make it. So tell me a little bit about the company and you’re offering and how you got started in this space?
Tom (AC): Yeah sure. So about 4 years ago I used to work for a creative agency and my job there was Innovation Director and it was my job to see have like one foot in the future basically and have that kind of opinion all kinds of cool things about the future of marketing and I came across examples of creative artificial intelligence so IA’s look good, write music, create video things like deepfakes and create images as well as run media campaigns and run with search campaigns and I just thought makes [unintelligible 00:01:24] as of all these technologies where at some point they would all become far better and there a new type of advertising service would be born with AI in the middle of it so I started and then try and create that and where we’ve got to it just oh literally 3 years later almost of the day is a advertising service that combines human insight with machine creativity that drives the performance of social display ads but at the same time creates first party data about why brands ads work so with performance business and where insights business combined as well I mean we work across all kinds of different sectors whether that is for companies like GSK or Reckitt Benckiser in this space as well as having work for Adidas, Diageo, Pernod Ricard, BOSE, Samsung and so on yeah a lot of experience in trying to find the future of advertising with AI in the sense.
RJ (PMN): That’s a great customer base in across multiple industries which is exciting cause it gives you a lot of views into what everyone else is doing. Drill into a little bit for me the experience in healthcare in particular.
Tom (AC): Yeah so we started working with Reckitt Benckiser out of Singapore I supported their APAC office pretty [unintelligible 00:02:49] in the business I went out there I met the team and we’ve been working on a brand called Enfamil which is known as slightly different things in another market Enfababy in Mexico for example and that is a infant nutrition product of which these several sub products but what the guys wanted us to know was to really use AI to help understand what would be the emotional and written triggers that actually drove interest and performance for the brand because the category like many categories actually tend to stick to their norms so whether you’re in the two phase category you tend you’re gonna see lots of images of ice cream and you know people in pain and so on but actually the more brands follow best practice the more they become similar to each other and the more similar brands are to each other the least that they stand out so that all became to us and said look, can you help us understand the different triggers that get people interested and signed up into their CRM program so we did work across Thailand, Vietnam, Capital of Mexico, US [unintelligible 00:04:01] and we tested hundreds and thousands of ads generated using our technology into which would be emotional drivers of performance and this being some absolutely fascinating insights that we’ve choose of and what market will able to prove statistically that targeting an audience of mums that it was actually pictures of dad and baby that drove past performance the mums and then in North America we’ve run a test where we tested hundreds of ads with a French Canadian speaking audience and the English speaking audience and over series of weeks while we strategically tested these different ads in scale we’re able to see the French Canadians really want its inspiration about being a mum whereas the English speakers in the same market the same demographic [unintelligible 00:04:50] rewards so by we’re basically the orthodoxy of advertising by saying, trying to come up with one idea that’s right for everyone is very very difficult happens very rarely it’s very expensive where I was using this automation technology you can in real time test different visuals, different messaging, different language, different formats, different places, different channels over over over series of weeks and an ongoing typically reproduce like huge huge leaps and performance but also fascinating insights into why people actually click on ads.
RJ (PMN): Interesting. Dive into the product a little bit more for me so it sounds like you are a platform the behind-the-scenes is the AI that’s kind of creating the magic of this kind of almost infinite a b testing to measure performance but there’s also a front end piece of this is actually acting like a creative agency would in assembling the various parts of a creative?
Tom (AC): Yeah so we have a two products one is a self-service to what we were searching audience generate the ads from scratch. Build a strategy for optimizing those ads in over series of weeks we deploy those ads in partnership with a major agency I talked to my support performance and deliver insights but we also have a self-service tool where in-house teams can plug directly into our technology, upload their own assets, create ads it’s scale and create that fist party creative intelligence to optimize their own ads and learn what it is an audience wants to know about so we do a lot of things that are very similar to an agency and we also do things that are very similar to stand alone platform but we are neither so we deliberately sit in an comfortable place between an agency and tech platform so we have a lot of the skills and sensibilities of what you’d expect from an agency but actually delivering at using this Innovative technology that it’s fairly unique.
RJ (PMN): One of the things I’m thinking about as you’re describing the kind of multiple permutations that could evolve and I’ve seen this in some of the examples you say to be for about like AI creating its own musical composers which have gotten really quite good and and fascinating. I’m thinking about the all these permutations and how a Pharma company or really any highly regulated company banking, finance how their legal and regulatory team would react to try and to approve something like this and I’ll give you a quick antidote this is going way back it was very cutting-edge program we did back in 1996 where we created ads on a proprietary online service that were customized with three dimensions so it was targeted to Physicians and we knew who the positions where based on a log in so if you were in London let’s say, it would say, Doctor Ollerton so personalized by last name the London area Geographic is on alert status for the flu tapping into a database to measure the flu in real time in your vicinity. So I remember trying to take that through medical legal review and I felt like the teams head was gonna explode the concept of literally potentially millions of different permutations of an ad coming up which in that case it was only three variables so they got it but I can only imagine the challenges you’re facing when you’re trying to get some of these things approved. Can you tell me a little bit about the challenges you faced?
Tom (AC): Yeah and we’ve worked in regulated space is pretty much from day one I’m sitting in the alcohol tax in the UK and Germany in the US and the US Embassy working with a bunch of brands in which is folio. So Theraflu, Advil, Biotene and ChapStick. We handle so on RB yes there is massively type regulatory approval although we generate hundreds if not thousand ads divided we do it strategically and sequentially where is it when you talk about it is pretty impressive 96 essentially a dynamic ad right so depending on what feeds come in like location or the flue data coming out if the one of the data fee produces an ad but that’s not we do that process of dynamic Advertising seems that the ad was right in the first place where is what we don’t do that. So our approach is we start of by using research and typically that is using social listening research. So we will listen to a millions of conversations about brands and it’s category to understand what we call the conversation landscape which kind a maps out the whole range of all of the different themes of conversations in that subsides and then what we do use our technology to automatically generate ads answer every single one of those conversations and the the data comes back from all of those ads and starts to give us an understanding of what makes this audience take on what is the DNA of the best performing ads. So then what we do is dig deeper into it which of those things right are important and keep on asking questions of the audience saying what is it about this thing what is it about this topic was about this subject is making you click we do that on a strategic to from a visual and language style approach so it’s very unusual to do that but we find it works pretty well for the Pharma companies that we do work with.
RJ (PMN): Fascinating stuff. So Pharma loves to learn from other Industries and your client base seems to span quite a few different Industries who do you look at today and do you say these guys are this particular industry this particular client is using AI very well even if it’s outside of Pharma that might Inspire Pharma marketers to think a little bit differently about their marketing?
Tom (AC): Yes so without wanting to name names to a specifically we have a lot of experience on the alcohol sector and one brand that we worked with on a regular basis is super up for testing different type of hypothesis using technology and we did this piece of work around gifting. What makes someone gift this certain brands by ecom. So we tested like a whole range of different occasions and reasons for gifting. So an occasion might be Christmas or birthdays or back to university or but a reason would be something like to say to say thank you or to you plan some other kind of message and it was really the best performing message in the UK certainly was to say sorry. So people will buying this alcoholic products to say sorry like the brand have no idea about it so we have another thing which is like your only get data back on the things that you test. So you mentioned earlier on the chat about AB testing which is personally just don’t bother because you were already testing two variables on one of those ads may perform because you had already attracted model on it or a funny line in it but you don’t actually statistically learn anything were as our approach to copy would be two things we test at a massive scale we test a lot of different themes and messages but we also test language styles a lot of this is new for the podcast but when it comes to advertising generally speaking a struck line or a key phrase It’s got message which is our product is the best program to take but it’s also delivered in a language style it’s going to be humorous or passive or direct or confident or cocky or whatever what we do is separate out those two things so we test message and retest we also test language style. So one of our friends recently tested 800 different brand claims but then once we knew what the best performers were we dug into what we call language binaries. So we were testing rational versus the motive we were testing humorous versus non humorous testing statement versus questions. So we’ll have a dashboard in all of this data gets beautifully pulled together. So at the end of the campaign we could say look this ad absolutely smashed it and this theme was the best theme in terms of visual and written perspective but we cannot fuse our report back on the type of language that the audience response to and this isn’t a focus group this isn’t just 20 people sat in a room sharing claim behavior. What this is essentially a focus group of Millions and Millions of people in real time.
RJ (PMN): Fascinating. So when I think about AI one of the challenges is you need a lot of data to go in you know, A/B testing definitely is not the right mindset but but it it kind of is if you think about A/B you know going up into you know it infinite levels because that’s ultimately what’s happening is for testing lots of different permutations of something so the concept of A/B still is there but it it’s it’s at much larger scale. But you need a lot of data input to make that statistically valid and kind of proved it out and do you what is your expectations for a client when you say when they embark with you and they say hey we’re going to do a media campaign is there a certain size of spend that they need to to do this at an effective level to really get to a good result and it is there a minimum kind of threshold you know what it but it’s just too small it doesn’t make sense to do this kind of wide spread testing at that kind of spend level.
Tom (AC): Yes so many interesting things in this. So the first thing to cover up before we get into investment spend and so on is that we haven’t just scale A/B testing, so it’s not like A/B plus C plus D plus E it’s not multi vary testing it’s what we call Strategic content optimization where is it you come up with a hypothesis about the audience then you create a ton of ads that test the hypothesis and then based on what comes back from my first hypothesis then you dig further into. So the initial hypothesis might be we think that women will perform better than men in these ads so we create a ton of ads. Lots of different instances with men and women then those ad perform or don’t perform and you be able to get a score of which is better say for example it’s men that perform better and then there’s a huge question mark because what kind of men? So is it men on bikes men on cars men outside using the product men using the product men not looking at the camera and then you have to create another hypothesis which is men indoors will perform better than men outdoors and the you run a bunch of ads like that. That’s not real like examples example that just very simplistic but that isn’t A/B testing that is strategic content optimization while you’re applying strategic thought and their human analysis to what a large dataset is telling you about what the audience wants to click on. So coming back to your point about investment so we’re a professional service either the search tool or the manage tool has a fee associated into that and if that fee dwarfs the media budget then it’s never gonna stack up and so but that said yes you need a lot of data but if you’re limiting the things you’re testing then the number of impressions that you need in order to the inside can come down drastically so you can do a lot with a little but yes if you are your marketing budget is 500 pounds a month then it’s just not gonna work cause you’re not gonna get the scale but it really depends on the size of the market depends on what platform you are on and so on and so forth so there’s no like this cost x that answer doesn’t exist. I apologize
RJ (PMN): That’s very helpful it introduces an interesting theme a lot of people worried about AI and the effect is going to have on jobs but basically what I just heard you say is there’s a definitive roll here for humans in the strategy side of this cause it’s humans ultimately directing what is our hypothesis that we’re going to test your strategically. So what is the role you see for humans in AI marketing?
Tom (AC): We set out to reinvent advertising using AI. So in the back of our minds we thought yeah you know we create this new thing and brought out humans involved anymore, that idea bit the dust pretty early on and we eventually realize that we’re not gonna replace any job so just going to create new ones because it’s a new skill set that is required for the old advertising skills that was you do some research you come up with an insight you come up with an idea you produce that idea and then you use various different technologies and platforms to distribute that idea whereas AI enables you to do is to test each one of those stages at a huge scale so at the research stage you can test far more things cut lines visuals the language styles and so on at the creative idea stage you know you can you can test far more things using automation technology and in terms of getting insight that that kind of multiply massively. So we really think that the if the intersection is human insight and machine creativity which is the really exciting part because machines are really good at some things like lots of small easy tasks out of massive scale and speed and humans a really good it really slow subjective things that take a lot of time and if you try to if you try to get a human to make money like a hundred ads that would take so long as we got a machine to try to understand the new once of a brands create brief it’s going to struggle is so what we trying to do is give the humans the stuff that humans are good at against the machines the things that their really good at and the intersection of those two skills as a compounding impact on each other and going back to the RB example before like RB came back to us and said that the insight of that market that found men and babies more attractive advert that women. So that would have never happened using our just traditional ad tech suppliers because it didn’t have the strategic element it didn’t have the human element there. So there’s a new I never said a new role but certainly a new skillset that if you are making thousands of ads. What should your strategy be?
RJ (PMN): Interesting We are going to take a brief break and then we’ll come back and we’re going to talk a little bit about the scary side potentially of AI when we come back with Tom all Orton from automated creative. We’ll be back in a moment.
RJ (PMN): Okay. We’re back. So it Tom I want I want to drill down a little bit more on this concept of jobs what are the things that that we hear a lot in our industry is that the Physicians jobs will actually probably be taken away faster by AI than the nurses jobs because the Physicians thinking of every binary way that they when they go through a differential diagnosis if this then this and they kind of go down this decision tree to do a diagnosis and they’re kind of a human computers if you will end in making those diagnosis. Where is nurses have a unique skill in empathy and bedside manner which is harder to replicate in AI. As you think about the job loss in the whole ecosystem of advertising and marketing who at the agency in terms of current role should be worried about AI because they’re their role might be displaced and who at the agency is actually going to be the rising star of tomorrow because their skills are even more in demand?
Tom (AC): So I did an interesting stat from Scott Galloway who does the Prof G podcast.
RJ (PMN): Oh. I love him.
Tom (AC): Yeah. yeah and his, his quote from the last couple of weeks something like Google and Facebook either gained or shared the entire value of the entire ad industry that isn’t them everyday so that is the best articulation of you know where the jobs are really and the WPP of this world and the other advertising networks that are fighting over the small change fundamentally. Where is all that, all that spend is obviously moving across some of those platforms. And so I think the jobs that are in doubt are creatives who are gonna carry on wanting to win and produce the big hero pieces of work. So yes there will be always a place for that luxury cinema advert or that TV ad that interrupt. But brands need more than that. They need scalable creative ideas that can flex and change depending on who’s looking at it and what the audience wants are those big batch of those 60 second ads. You know there will be budgets for that but those budgets will get smaller and smaller but when you see the rise if AI driven platforms I like TikTok really over taken the world in a very short amount of time. The ad units that are going to be needed that the type of creativity required will really struggle if we was just thinking in terms of a 60-second TV spots that need to be thinking about how they’re going to use technology to produce a lot of ideas that change very quickly base on what the user wants to see not what the brand wants to say.
RJ (PMN): Interesting. Scott Galloway’s from my alma mater, MYU I did my MBA there. when you think about his view of the world how to have a Google and Facebook are eating the world is that what when you do your application of AI and you go out into the channels where you’re buying are your predominantly buying on Google and Facebook to do this analysis or what are your what are your platforms that your clients go out and execute on in order to run your AI software?
Tom (AC): Yeah we work across most of the major platforms in the last 3 years but yet there’s a chunk of work that is Facebook and Google.
RJ (PMN): Yeah. And you’re really just testing the creative so it doesn’t matter that those are kind of Walled Garden Platforms in terms of having other access to other data.
Tom (AC): No. It does not, that walled garden we report all our data essentially so we can report back to a client’s which of these emotional triggers working at audience level that a market level of the region and a global level. So working with us as a brand is able to see right what is this brand doing at a global level, what are the emotional triggers that are driving this macro level. But actually if you wanna know what’s going on with 25 to 30 years old audience in France we can tell you what can I call you should be using what kind of emotional language what kind of visual triggers that are working for those guys.
RJ (PMN): I want to take the conversation up a level and talk broadly about AI for just a moment cause you’re deep in there so you have an interesting view at any time there’s a new and Innovative technology there’s always a little bit of concern around it and I don’t think I think that’s especially true and I haven’t seen a higher than in an area like AI, you know everything from The Terminator movie to AI taking over the world and defeating humans even really bright Minds like in Elon Musk having kind of negative used not negative on AI generally but cautionary that we really need to put some guardrails in place before it might get out hand. His comments about you know that being a high probability that we’re all already living in a computer simulation and basically life isn’t real I think are pretty fascinating. What are your, where do you come down on kind of AI generally as a tool and that the need for potential guard rails and I’ll put that in one more, one more kind of piece of data for you to think about their is what’s happening right in social media for example where conspiracy theories and such are really taking on a life of their own at a scale that we’ve never really seen before. So clearly the power of the platforms are very powerful and there’s probably limited guardrails today what are your concerns if any with respect to AI and on the flip side, what are the opportunities long term?
Tom (AC): What a question for the three minutes left. So yes obviously the channels have incredible power on weather at in the consumer space or the political space or in the land of fiction and conspiracies and so on and it I think it’s interesting that they aren’t regulated is in the same way the other media channels are regulated I’m appalled that regulation issue is this the scale I mean I heard Mark Zuckerberg station tonight they spend as much money on trying to account count as a fraud fake news than Twitter turn over in a year. So weather that’s true or not, can you hire people that quick, who knows, some of my thoughts about that, yeah we gonna look back on this period and think wow that was wild. and it’s something that it will need to be regulated and we support that and yeah we have no interest in doing anything that isn’t entirely ethical. So I think there is obviously a massive opportunity with AI fundamentally it’s just the technology in the same way that like cars can be dangerous or planes or whatever it can be used for ill and we are fully aware that advertising has a big impact on the way that society sees the world. Advertising pays for the internet in a lot of senses, It pays for the Facebook’s and Google’s of this world and the impact advertising has on people will change the view of other people, you know we actively encourage our brands to make sure that when their creating ads, especially if the ads at scale that we do, they are representative of the audience seeing those ads. And I think we as a supplier to the advertising industry it’s important that we make sure that the impact that we’re having on that society is a good one. It’s very complicated and difficult. Yeah it’s a huge opportunity with AI but if an article got the terminator at the top of it I suggest that you don’t read it. Hopefully those days are gone and a sensible and a nuance conversations can be have
RJ (PMN): Yeah. I think you’re right we’re gonna look back on this period of time and say that was wild. And I think a lot of people are in this period of time saying it’s pretty wild. So last question for you Tom you’ve got a lot of Pharma marketers listening to this and you can almost visualize the light bulb going off in some of their heads right now. What kind of Pharma marketer do today to get started in AI with respect to their marketing efforts?
Tom (AC): If you got Pharma marketers having the light bulb moment call me. I will talk you through my experience of it which is working with a bunch of GSK brands, working across RB as well as the brands I mentioned at the start of the show manatees yes I love it I get up every morning super excited about reinventing advertising with AI. I’m a working Pharma so you can shoot my ear off about that all day long.
RJ (PMN): Fantastic. Alright thank you very much everyone and thank you for listening this has been RJ Lewis and Pharma marketing network was Tom Ollerton from Automated creative I hope you enjoyed it and we’ll see you next time.
Listen to this podcast, Ep. 012 – Tom Ollerton here: https://www.pharma-mkting.com/the-pharma-marketing-podcast/
