Machine learning and A.I. is already in the world of media and online communication all around us. But it’s broken into marketing and advertising too, even if you don’t realize it.

Will A.I. ever replace a human’s expertise when buying ads? Bant Breen, founder of innovative reputation management firm Qnary, discusses this question. And it’s not cut-and-dry. One thing’s for sure. Just because it’s a shiny new object doesn’t mean you should be using it in your marketing efforts.

In fact, many agencies shouldn’t be using these systems at all… at least not until they have the right “infrastructure” set up.

We discuss that in detail, as well as… 

  • The first place to start if you’re interested in A.I. 
  • What machine learning can and can’t do – at least right now
  • The biggest challenge in implementing A.I. solutions
  • “Off the shelf” A.I. you can use now in your marketing efforts
  • And more

Listen now…

Mentioned in this episode:

Transcript

Pre-Recorded Intro: This is Performance Delivered: Insider Secrets for Digital Marketing Success with Steffen Horst and Dave Antil.

Steffen Horst: Welcome to the Performance Delivered: Insider Secrets for Digital Marketing Success Podcast. We talked with marketing and agency executives and learn how they build successful businesses and their personal brand. I’m your host Steffen Horst. Today we’re going to talk about AI and how agencies and media companies can use it to improve several areas of their business. Here to speak with me about the topic is Bant Breen, who’s the founder and chairman of Qnary: an award-winning reputation management solution company for professionals, brands and enterprises. Bant has worked for numerous marketing holding companies and held different leadership roles as publicists, Dentsu, and IPG. He was inducted into the Advertising Hall of Achievement in 2010. Welcome back Bant.

Bant Breen: Glad to be here.

Steffen Horst: Well last time we spoke about the research you had done to identify how agencies use AI, you had mentioned that it looks like there, there is no common definition for what AI is among agencies that there is fear among agency leaders in regards to well, is AI going to take over the world in their way make people redundant and therefore an agency so to speak, and all the operations are run by machines? What I wanted to start off with today is looking at that insight, is that more for bigger agencies or have you seen that for smaller agencies too? And, are smaller agencies more likely to adopt to AI? And if so, why?

AI & Agencies

Bant Breen: That’s a great question. I think that there are a handful of examples of really, really creative things that have been done by very small agencies, some of them actually based out in Los Angeles, that that, you know, kind of almost like take the chatbot idea to the next level. Very kind of highly niche, highly creative areas and those have been done by smaller agencies. I think that the bigger agencies, some of them like, you know, Universal McCann, is actually a good example of it on the media side. I know that actually Horizon did this as well, where they contracted in the case of UM, they contracted IBM to roll out Watson as a supporter for their media, their programmatic media buying or not only programmatic, really almost all media so just decision making in that process. I think the challenge is I don’t know if that’s stuck. I don’t know if they’re still doing that. I don’t know if it was just like a shiny object. There are smaller agencies in the media space. Like, there’s a company called Albert, that operates as kind of an AI-led media agency.

I think the challenge I’ve had with companies like Albert is that they are they have trained Albert really to focus just on Facebook, you know, buying ads on Facebook. Which is great, but you know, that’s not going to shut down. It’s not enough, right? We need a lot more complexity there. I think the only advantage that a small company agency might have is that they’re not encumbered by the massive structures.

But I think that it’s really important to just reiterate that the skill sets needed for AI are quite different from the things that you’ll find in agencies. You know, most agencies don’t have data scientists. Most agencies have nobody that knows how much about Python or any of the related code base that’s been developed for, you know, machine learning code base. So, and as I alluded to, last time, we spoke all the vast majority of talent has been absorbed into a very small group of companies, you know, the Facebooks, the Amazons, the Googles, the Netflix of this world. Mostly on Facebook and Google.

And so I think, you know, we’re seeing we’re starting to see the cutesy stuff. I mean, it, you know, if you use Google as your Gmail system, you’ll see the automated messaging that they offer. You know, those are all really driven by an understanding of, of how you are messaging things as well as how people respond to text that’s similar. So, it is all machine learning. 

So I don’t know if a small agency has an advantage. What one thing I will say is, I don’t feel, you know, that kind of old adage, Steffen, where, you know, we used to say to smaller companies, like “listen, if you use your digital strategy effectively, you can you can play on the same level as a big company” Right? You know you can, you can compete. So, a mom and pop or a startup can compete with the big guys.

I haven’t seen a lot of that yet. In the machine learning space where the mom and pops are doing it better than the bigger guys. And then also the most meaningful systems that are being released right now with AI are things like call centres, chatbots, things that are managing like the user interface and the user experience, optimizing media budgets. So… things that aren’t that sexy, but in fact, actually require a lot of data. Right? And bigger companies have more data.

So, you know, so I think, in fact, actually, I think you’re highlighting a really interesting opportunity and a challenge, how does a small company compete when they don’t have all that data, you know?

Steffen: So I mean, having the data requires to collect it, right? I mean, it has to be stored somewhere and then you need to be able to slice and dice it in order to make sense of it. As you said earlier, and I paraphrase, if you ingest crap, you get crap out of it, you know. Then it’s not going to work well.

So probably it’s a two-part question. What systems can you know, small-midsize agencies use to do data collection? And then do they really need data scientists to get started with AI? Or can they not just start off with using solutions that are out there? For example, you know, there’s a system I think it’s called Node which is backed by Mark Cuban, which is a sales software, an AI-based sales software. You mentioned chatbots and other solutions.

So where do people have to start? What solutions can they use in order to start collecting data and then basically preparing the data to be used for the systems?

AI & Small-Midsize Agencies

Bant: Yeah, you know, it’s a great question. I had a series of conversation with a couple of folks that have been working in the machine learning space for a long time over the, over the last couple of months, and, you know, one of the things that they highlighted was that we’re all in this, like mad rush to, you know, have like the plugin or the app that we can just be like, “Oh, that’s what we need, that’ll make everything work.” 

We’re still kind of at a slightly earlier phase of AI, I think, where we really have to kind of pause for a second. And I would say, ask yourself some fundamental questions, which is: what is it exactly that you’re trying to achieve? Because the best tools… you can get a lot of kind of pre-structured code that is designed to do specific things, right. And so, I mean, you know, there’s a ton of open source, you know, software out there that, that someone can pull and actually start to use to develop. 

But you need to understand like, is it that you’re trying to train something to recognize photographs? Is this what you’re trying to train something to recognize anomalies in buying patterns? You know, you have to be very, very clear on what you’re trying to do. Because AI doesn’t really, you know, while you can kind of give it unstructured data and see what it comes out with… it’s not particularly great at that yet. You know, we’re still trying to mean that you can do that and say, Alright, well look, look for any patterns you know, and see what you find.

The first phases for agencies should really be like, looking at what you do and say like, “couldn’t we do this more efficiently? Could we be doing this at greater scale? And then matching it up?” There’s probably what you know, what you can do is, you know, there’s a ton of stuff from IBM, there’s a ton of stuff from Google, let’s go from Facebook, that you can get, you know, to utilize freely and, and of course, all those guys will take your money to help you.

Steffen: Sure they do. They love your money.

I mean, I think you know, the most simple form is, if you look at Google, for example, the snippets that you can program in that can be uploaded in your Google account in that can handle certain tasks, right. So for example, for us here, at Symphonic Digital, you know, we have one of our guys that looks at processes and looks at can that process be automated, and if so, then he goes about and creates a snippet and upload this into Google. We test it, and if it works and does the job, we cut X amount of time out of the day to day management, which, which helps because then you can focus on, I would call it higher-level things.

So you focus more on strategy on the bigger picture rather than being deep into, you know, into the nitty-gritty little things. Which, you have to pay for if you have a human being sitting there. But you’re right. It doesn’t make sense.

Bant: But yeah, but I mean, I’m kind of saying that fundamental question right up front, which is you need to be able to ask that initial question, right, the initial thing that you want, and then and then machine learning can support you to come up with a better answer for it. 

Steffen: So, once an agency just thinks about it and comes up with you know, these are the areas we believe we can get a software solution that can pick up the slack. How would they go about to collect the data, what systems and do you recommend? Do they have to pay for it? Can they use a solution like Google Analytics to start off with?

eCollecting Data: How Is It Done?

Bant: Sure. I mean, they can use Google Analytics. I mean, there’s, there’s, you know, there’s an argument to say that Google has a set of tools, Amazon has a set of tools, those being you know, probably Amazon being kind of like a huge cloud-based provider, Microsoft as well with their Azure set. You know, those would be the companies that you should look at their sites, look at how they allow you to store data and then allow you to, to actually utilize many of their products for structuring those that data to be utilized effectively within their systems.

It’s a, you know, there’s kind of a, there’s a running battle in one of my in Qnary, where we have, you know, some of our systems are Google, some of our systems are AWS and you there’s a lot of developers that hate that. There are some developers that actually actually think it’s quite important to do that. But they are structured slightly differently.

So it really will depend on kind of what systems you’re building off of that you, you know, the best way to optimize.

Steffen: You mentioned cloud a second ago, I think that cloud storage is, probably was, one of the biggest hurdle for AI solutions to progress faster. But, what you at the end of the day is, you need a place where you store all the data sets. At this point, are the cloud solutions available good enough for the agencies to advanced AI?

Bant: Yeah, at this stage, sure. But, I mean, the need for greater storage is going to increase exponentially and so that’s an ongoing discussion and an ongoing area that is going to require a lot of work.

Because, I think one thing that people don’t remember, you know, machine learning uses a ton of energy and just a ton. And so, you know, it’s not the greatest thing for the environment, believe it or not. So I think it’s going to be one of the big discussions that we’re going to have to have as a society, like how do we do these things more and more efficiently. You’ve seen, you know, big companies try to figure out cheaper ways to keep their storage you know, cool. You know, Google put it on a barge and I know there’s one group that has some as Amazon has several Microsoft has several that are in like Greenland, and you know, it’s it requires… it’s gonna, that will be an increasing need.

So I guess, um, you know, for the primitive stuff that we’re working on today, sure we’re okay. But for the stuff that you read about, and you see in the movies, we’ve got to figure that data storage out still, you know. 

Steffen: You just said the primitive AI stuff. What are the areas currently and that might differ from agency to agency, depending on how they answer the question, you kind of state a few minutes ago, but what are the areas at the moment where agencies should consider AI as a way of cutting out the manual labor and letting the system do the work? Are there specific areas that you identified in your research that the agency already use?

Bant: Yeah, yeah, I don’t have the complete list in front of me but I mean, there’s a lot of stuff that I’m you know, where you have a connection to clearly sort out specific data, right or to update that data.

So it may be trying to structure, almost using a kind of like all the basic stuff that we would have had for like SEOs. You know, think how things are tagged, how you sort information along the along those lines. How you find anomalies in that information. That type of use of AI is going to become more and more prevalent.

We’re starting to see it used very much for you know, things like photographic recognition, facial recognition. Right. So, trying to understand, as I said earlier that a dog is a dog, a cat is a cat, that Steffen is Steffen, you know when they see you on a, on a site of some sort.

There is, I would say, that probably one of the areas that is being developed and probably will develop a lot more is essentially kind of like AI-driven, dynamic creative. So I’m kind of pulling together information about the user in real-time, they can essentially create kind of a, an optimized ad unit of some sort, I find that really interesting or an offer of some sort will become more commonplace. As I said that areas like CRM related areas like even email marketing, call centres, managing how the user interface or that experience that an individual will have in a digital environment, how that can be optimized and how it can learn quickly, is going to be helpful. 

And then on the creative side, you know, you know, Adobe and several others are already developing tools that would allow an individual to render kind of, like an idea of much faster. Right. So like, you know, you know, building and collecting on your vision of what a cat is, you know, in a much more rapid way to so that you could design all the characters for your new production of Cats the musical. So, so I think that you know, we’re going to see it touch a lot of areas very quickly.

I would say that you know, areas that are that I so, you know, one of the things I was looking at a couple of weeks ago was, you know, audio and how, so, you know, AI is quite, quite useful in determining patterns of authorship in music. So it literally can tell you all the songs that were written by John Lennon that were written by The Beatles. Right.

So things like that I think are going to be… so from a research perspective, it’s particularly useful. That being said that the AI created music today is extremely rapid. So they got a long way to go before they’re gonna come out with, you know, the next hit, I think. 

Steffen: Yeah, well, you know, I want to come almost full circle. You know, we started off in the first podcast, talking about your findings and from your research. From your perspective what needs to happen for the marketing executives to lower the barriers to embrace AI technology and start looking at how they can improve processes internally? I mean, with that always goes along being more profitable. Is it the software solution themselves? Do they need more education? What is it? 

Bant: So, you know, there’s been a, you know, there’s a group of Harvard Business School professors actually tried to look at that very question, which was, you know, how should a company actually get started on this stuff? Should they actually create kind of like a separate team or a group that focuses on AI? Should they try to integrate it into kind of the day-to-day?

I think that a lot of it’s going to depend on the specific company and the specific people involved. What I do know is that if you can get an organization aware of aspects, so like, obviously, as a manager, you need to have a certain level of understanding, if you can educate people on the general area of machine learning, and then you can help them kind of get through the planning task of how it could impact their business. I think it you know, it’s dangerous at this stage to jump to the natural human place, which is where all CFOs will go, which is “How can I use this to cut my workforce?”

I think that you know, I’ve seen that in my, my own company where, you know, when I sit down with the finance director, you know, he’s like, yeah, we can reduce, you know, our creators from 500 down to 200. Let’s, you know, I don’t jump to that yet. You know, we’re not I don’t think we’re there. We’re there yet. That shouldn’t be the initial driver. Albeit, I think it will be a factor over time.

Because I actually, here’s an interesting one… You know, there’s a hotel in Japan called the Habbo Hotel, which was this scary hotel, which was a hotel that was only populated with robots. And they did it kind of almost as like a funny, cool, trendy thing like a robot hotel that you could stay at. What they found quickly was, in fact, actually earlier this year, they’ve actually removed 30% of their robots. Because, you know, so even though robots will lose their jobs if they don’t perform. Like they have to learn, they have to learn quickly as well.

So I think that you know, Steffen, I hope to have a better answer for you in the next couple of months on that question on exactly kind of the steps that I think, you know, a marketing department should take, but certainly there is kind of a much deeper general knowledge that needs to happen. And then there needs to be a really thoughtful discussion with the people that are in charge of talent, to think about the types of people that they’re recruiting into the business. 

Because the marketing group of the future does not look like it requires a different type of executive. And it will require people that are translators as well, I mean, data scientists and marketers do not speak the same language.

So there’s going to be a lot of work that’s going to have to be done that that gets those groups, you know… there needs to be, you know, a whisperer. An AI whisper, you know, who can somehow explain things.

Steffen: We talked about humans throughout the entire podcast. But as kind of a last question. Is AI making humans relevant moving forward for agencies and media publishers? Or where is the good medium?

Bant: You know, I would say that probably the hardest areas for neural networks to copy are kind of like the creative or illogical logic of human beings. And so, you know, humans have tremendously interesting brains that allow us to do lots of things that are very hard for machines to do at this stage. 

I would say that if your job is, you know, very, very kind of like cookie-cutter, right like, like if you work at a call centre or a contact centre of some sort. I can’t imagine that in 10 years time, there will be rooms filled with people answering phone calls anymore. I just… I believe that almost all of that will be automated. I actually was… I was in a meeting with Google, I saw some of the stuff that they’ve developed in that area. And it’s just so good already. You know, in five years time, it’s going to be flawless.

And so, it’s going to be able to detect motion, it’s going to be able to detect speech patterns. The system that Google has developed, in fact, actually, you know, when it talks to you it makes human inflections: pauses, ums. You know, you really feel like you’re talking to a human being so I think a lot of those types of roles may fade away. But in terms of kind of a higher level needs for strategic thinking of how do you have your company breakthrough, that’s going to be important.

And then I do think one of the hardest things for humans is that AI machines sometimes you know, that it will have to be there will have to be a human that evaluates the results because you know, AI will come to a conclusion that may be the most efficient but it may not be the most ethical, right? Or it might be the most profitable, but it might be racist, right?

So you have to… so there’s going to literally have to be a lot of buffers and protections, and it’s going to be bumpy. There’s going to be so many mistakes along the way. And I think the thing that probably everyone worries about the most, I think right now I’m seeing a lot of goodwill put forward towards doing things in an ethical way. Last year in October, there was the universal guidelines for artificial intelligence were put together and it was agreed on by, you know, I think most of 70 countries and you know, these are kind of like principles that most people have agreed to, but how we police these things, how we manage these things is going to be tricky, you know, the most dominant nations in AI right now are China and America.

In China, everything is prohibited. So everything is controlled by the state. So AI just allows them to know everything in more detail. In the US, we generally don’t like to regulate things, right. I mean, in the US, if you were to kind of take like a pure capitalistic viewpoint, we don’t like to have the government involved in things, right. And so in some ways, like that European voice in the AI world where, you know, I would say Europeans would regulate business they tend to want to have more government involvement, and they have a better social welfare system because of that as well.

That’s still… that that’s getting a lot of lip service right now. I don’t know how much it will get in terms of action. I mean, it really is an interesting one, like countries like China and the US are investing, like, exponentially more than Europe is right now in AI. So and that I think is going to start to have material effects on the world economy.

Steffen: Yeah. Bant, I mean, it has been great talking to you about the topic of AI.

Bant: Sure.

Steffen: Thanks so much for your time. And if people want to get in touch, want to find out more about you, how can they do that?

Bant: I’m sure you know, you can always find me on LinkedIn or Twitter, but you know the others you can always just email me at my funny name that’s very short but very hard for people to remember somehow. Which is Bant@Qnary.com. So Bant@Qnary.com and I’ll be sure to get back to you. 

Steffen: Wonderful. Well, thanks everyone for listening. If you liked the performance of our podcast, please subscribe to us and leave us a review on iTunes or your favorite podcast application.

If you want to find out more about supporting digital you can visit us at SymphonicDigital.com or follow us on Twitter @SymphonicHQ. Thanks again and see you next time.

 

Share This