Pivoting from pivot tables: Leveraging AI to transform healthcare marketing efficiency and accuracy

[00:01] This is efficiency to me. [00:04] This is speed and accuracy. [00:07] that’s what [00:08] The Solutions when use properly should provide to us but that’s starts. [00:14] with really understanding [00:16] the data sets that are feeding them you have to imagine people are scared. [00:19] But I think as long as we show that. [00:21] We’re here. [00:22] As a support system, we’re here to help we’re here to educate and we’re going through this together. [00:27] I think that. [00:28] Is the biggest bridge that we can all build today? [00:36] Hello and welcome to the MM+M Podcast my name is Jack O’Brien I’m the digital editor at MM+M. Please be joined by two very special guests today hey Jack how are you? My name is Oleg Korenfeld. I am the chief technology officer at CMI Media Group and Compas and I am Ezra Suveyke the chief technology and product officer at Pulsepoint. [01:00] I’m so happy to have you both on- here and we’re talking about a very exciting topic one. That’s certainly got a lot of traction in the industry as of late, which is AI [01:08] everyone’s been talking about everyone’s been using it everyone’s been– trying to evaluate their usage of AI [01:14] And that’s kind of where I want to start the conversation Ezra maybe you can kick us off here. I know that you’re already using AI to automate campaigns. Can you describe that process for me in terms of automation and maybe what are some of the benefits or results that you’ve seen thus far. Yeah so in a couple of different arenas. [01:30] pulse point originally made it to Mark [01:32] within the industry with its. [01:34] capabilities and contextualization [01:36] if you look at where a pulse points got started in the field of AI if [01:40] you go back about eight years ago. [01:43] We develop neural networks in order to support. [01:45] the practice of what’s called contextualization [01:49] that is effectively what our engineering or our data Science team would call the real AI [01:55] now mind you this is also the same team that says. [01:59] Hey Ezra you know you’re just a a marketer. You’re not a real technologist anymore. You go off you go talk to clients and convince them of. [02:07] The next and the greatest things that we should play around with test and build. [02:12] So, let’s take that with a grain of salt. [02:14] the cool thing is [02:16] and how AI has essentially changed. [02:19] the barriers to entry [02:21] Are getting lower? [02:22] You look at the type of work that we get started with in the beginning at pulse point with. [02:28] With the contextualizar. [02:29] And compare it against what we’re doing today. It’s night and [02:32] Day, it’s essentially the equivalent of saying I’m going to build this utility to do this one thing. [02:38] and now [02:39] we’re thinking through the opportunities to do automations within campaign. [02:44] that touches [02:46] five years of historical data [02:48] a good example some of the work that we’re actually doing with cmi. [02:52] We started on a systems based integration. [02:54] Trying to remove the amount of friction. [02:57] in front of a trainer [02:59] when launching a campaign [03:01] so is that okay, let’s [03:02] Have our systems talk to each other. [03:04] let’s build an API [03:06] and when somebody hits that magic check box within. [03:09] Oh leg system back at home. [03:11] That information is now within the Pulse point platform. [03:14] That’s cool and that was very you know very tedious work work. [03:19] how we are evolving is [03:22] oh [03:22] you forgot to put in a field. [03:25] Let’s look at the historical five years of data. [03:28] And ask the lom, what would you put here? [03:31] That’s a good example of [03:32] nice quick wins where you can insert this mega technology that can do a lot more than [03:38] You know some regression analysis to say you know should it be a one or two in this box? [03:43] It’s really interesting to hear you talk about all that background in terms of what you’ve been able to do in terms of making an actual practical application oleg what about for you and your organisation. [03:53] Sure again. We are a media agency. So it often assumed that we may not be that technologically savvy or we spend a lot of effort on on our own internal development which is actually not true at all cmi over the many many years invested a lot in technology and data. [04:10] aggregation and and management and that allowed us over the last few years to take that knowledge to take those learnings and Take That [04:20] insight and because even though a couple of years ago. It may still have been. [04:24] Called just machine learning optimisations, Use using. [04:27] a data analysis to inform some predictive results [04:31] Now it’s ai-driven and what we’ve done over the last few years actually is look at low Hing fruits of what can change our process. [04:40] So what we’ve focused on is Task automation we’ve realised that. [04:45] there’s a lot of tasks that happen throughout the [04:49] life cycle of planning and buying and campaign [04:52] the historically been done by a lot of humans. [04:56] So if you take those processes understand, what’s being done how they’re being done and take all the knowledge information that we have of what work within work that could be used to train. [05:08] a model [05:09] To start doing that automatically so we started with. [05:12] Just basic. [05:14] Are trafficking with taken? [05:15] Thousands of thousands of companies that historically been done by armies of traffickers. [05:19] And automated about 70 to 80% of that process. It’s not fully automated we still need a lot of human touch. [05:26] To make sure that there’s validation that everything is correct. [05:30] but [05:31] a huge percentage of the possibility automated and once that’s been ordered the next logical step to us would be well if we can automate an R end. [05:39] Why can’t we automate it with our [05:42] biggest partners [05:43] This is where Ezra and I got connected and said hey. [05:47] we’re trying to [05:48] take on more of a self-service approach to using pause point platforms. [05:54] In old days that would require a lot more. [05:56] hands on human [05:58] work if we can take percentage of that process in order to meet it. [06:02] the the [06:04] whole [06:05] process of taking things on [06:07] internally becomes easier [06:09] So this is where our partnership kind of co-development kind of kicked off in specifically taking. [06:15] These very manual processes. [06:17] building [06:18] bots on both end of it. [06:20] And then automating it and once the connection is built and automation is made where. [06:25] We press about button again simplistic terms plus a press about button in on our and that sets up companions on our side or medically that pushes information over to pass point they set up campaigns on their sites, so that process is. [06:37] Fully automated then after that. [06:40] Again to azure’s point what can be done next? How can we optimise? How can we now use? [06:46] the historical data [06:48] to [06:49] just move out of the just campaign setup phase to contain optimization phases. [06:53] And I’m curious. Oh like you brought up an interesting point early on when you were just discussing how your organisation is usually utilise AI and you talked about being able to take these manual processes. [07:02] And making them automated I know that in discussing. [07:05] This topic with other leaders the concern ultimately comes to what then is going to be the human role in the AI process a lot of people have talked about having humans basically. There’s the guard rails for this technology and being able to guide this innovation to get the maximum results. [07:18] Not taking everything out of a person’s hands and putting it all into the machine. I’m curious from your perspective. [07:23] How you see the role of humans in this kind of new AI universe and maybe how it’s going to impact the workforce for at large? [07:30] it’s [07:32] it’s interesting because I don’t see. [07:37] That being some kind of a fearful element of jobs going away. [07:41] Which taking very manual processes and we’re educating? [07:45] our teams [07:46] with these new skill sets. [07:49] In the experience that we had so far. [07:52] With basically over the last few years created a lot of AI experts in our company and I don’t mean developers. I mean. [08:00] Traffic is I mean Media planners people who are comfortable using are differential intelligent tools. [08:07] To look at their process. [08:09] redefine it [08:10] create a new workflow. [08:12] Now that’s one element the second element very important element is to understand. [08:17] how that affects [08:19] everybody financially because [08:21] historically [08:22] there’s a [08:23] team right built around a client to support. [08:26] certain tasks along the way [08:29] If we’re now automating the processes. [08:32] It’s not the same team anymore. [08:35] So the conversation is not necessarily. Oh, you have less people does that mean I have to pay you less. [08:40] Well, not necessarily. [08:42] Because now we have to hire more developers to build the buds. [08:46] We need to hire. [08:48] very strategic senior people to [08:50] train the buds and we need to have people who understand how this works. [08:55] To make sure that the bots are doing what they’re supposed to be doing. [08:58] so transitioning [09:00] the skill sets [09:01] and educating [09:03] the teams [09:04] been very successful in our end are adapts team basically turn these to these ad tech strategies to consultants because as we take away these manual responsibilities of them we create time. [09:15] for them to focus in more strategic other things which [09:17] are very beneficial to our clients because [09:20] often the stuff gets put in a historically been put in the back burner now. They have time to think forward things strategically. [09:27] and [09:28] that allows us to go to our clients and rethink the way we financially work with one another because now. [09:34] It’s more of a technological. [09:36] relationship [09:38] With a managed service on top of it. It’s interesting here you bring up kind of those points again. I’ve heard from other leaders where it’s like. It’s not getting rid of. [09:45] You know your full-time employees it’s it’s being able to utilise their skill set and something that’s more. [09:50] Applicable unless busy work I guess you can say as a refer from your standpoint. How’s your organisation fared on the personnel side? [09:57] Yeah, I would say very similar which is of course refreshing and yes. [10:03] Similar in all cases where I’m speaking to to others. [10:06] In our field and in mind you I’m sure every. [10:09] Industry is different right. They have their own they have their own story. [10:13] somebody that [10:14] you know is simply picking something up from here moving it here in a spreadsheet. [10:20] That ultimately doesn’t have the technical wear with all too evolved from there. [10:24] That’s not who we’re talking about I think we’re talking about. [10:27] Those within the spaces that we participate in that. [10:31] Have a higher understanding of process and technology and learning. [10:36] And you know having these people within your organisation. [10:40] Allows you to very easily say the following. [10:43] Well, we’re going to focus on Educating team. [10:45] that is the [10:47] single largest [10:48] initiative that we have around AI [10:51] because what we want to do is take. [10:53] Every individual that is capable. [10:55] and say [10:56] you are now an inventor. [10:58] go off and do R&D [11:00] go off and tell me what you need don’t ask me if your job is in jeopardy. [11:05] Make yourself obsolete. [11:07] But also show me the path to how you’re going to be upskilled. What do you need from me in order to do that? [11:13] It’s it’s really exciting time. [11:16] Alright, I wasn’t here when. [11:17] You know the first computer was put on somebody’s desk. [11:20] But it’s kind of the same thing it’s I you have to imagine people are scared. [11:24] But I think as long as we show that. [11:27] We’re here. [11:28] As a support system, we’re here to help or here to educate. [11:31] And we’re going through this together. [11:32] I think that. [11:34] Is the biggest bridge that we can all build today? [11:37] Just to build with azure is talking about just experienced. We’ve seen an r and where. [11:43] After presenting the use cases that we’ve been able to do with automation of certain tasks the feedback that we got from around the company saying well. What about this task? What about that? Task could people who are leading in the ones who? [11:56] All want to learn and they are helping us to better build that end to end solution now at every stage where we questioning. [12:04] Whether the process needs to remain manual or can be automated. [12:08] absolutely and [12:10] that’s an great analogy with computers. I’ve been using the analogy with. [12:14] Emails when email is introduced and even today like you don’t need to know exactly how the email that you’re right ends up across. [12:21] somewhere on the other side of the world [12:24] to take full advantage of what technology can give you. [12:27] I feel AI [12:29] can be that next. [12:30] evolution of that [12:32] kind of [12:33] efficiency [12:35] I want to say here in date myself, but I’ve heard a lot of leaders talk about the fact that oh like when the worldwide web and when the internet really came online and kind of revolutionized. How people were using the personal computer it’s something akin to that. I can’t really relate to that because I was. [12:49] Far too young at that point but I can see how there are those parallels from people that were around at that time to be able to discuss it one thing though that I have her a lot of parallels, too is like if we had this conversation about a decade ago. [13:00] About like what the big it innovation of the time was it was big data. Everybody was like we need to have data. We need to have large quantities of data. [13:08] Day-to-day data and then it was like oh, but we also need to have high quality data. They can’t just be data for data’s sake and obviously these programs and protocols and technology is run off of that data. So there is that need to if you’re going to feed mediocrity you’re going to get mediocrity, but if you feed it. [13:24] Something that’s actually meaningful you’re going to get meaningful results. I’m curious from your perspective how the data quality goes into the side of the conversation. Oh like maybe you can start us off there and after I feel free to hop in absolutely everything starts and ends with the quality of data that feeds the algorithm. [13:39] the crap and crap out [13:41] analogy is perfect here. [13:43] you can take [13:45] the worst possible data sets [13:48] feed it in the best possible– algorithms. [13:50] And you will get wrong answers. [13:52] the examples of all the chair GPT oopsies that we’ve seen [13:57] a year and a half ago [13:58] Is that it’s been scraping the bottom of the internet? [14:02] to get the insight to get the for the answers that would was providing of course a lot of [14:08] that was not [14:10] good data. [14:11] so I [14:12] tend to think about it this way we spent. [14:15] over a decade [14:17] really understanding [14:19] the importance of quality of data [14:21] In our work. [14:22] we spend [14:24] years [14:25] find not just defining what kind of partners do we need to work with butt? [14:29] Understanding how they source data. [14:32] how they [14:33] clean it how how did they refresh it? [14:36] So now that we have this new set of exciting tools. We should not lose our heads. [14:42] and assume [14:43] that we can just trust these things. [14:46] To figure it out. [14:47] No [14:48] We need to keep that level of scrutiny over the data sets that are used to feed this algorithms. [14:54] to get [14:56] the results that we want because in the end of the day again. This is efficiency to me. [15:00] This is speed and accuracy. [15:03] that’s what [15:04] the Solutions when you use properly [15:06] should provide to us. [15:08] But that’s starts. [15:09] with really understanding [15:12] the data says that are feeding them. [15:14] What do you think asthma? [15:16] Oh man, we we hammer we hammer this point home constantly. [15:24] Garbage in garbage out. Let me give a real world use case. [15:28] That we started going through. [15:31] postman goes back [15:32] many many years we’ll just say let’s even just go back 2012. [15:36] The company is changed who had who it is. [15:39] several times over with [15:41] mergers acquisitions and then finally [15:44] the last change going back six years ago where [15:47] We verticalized our company within. [15:50] health care marketing [15:51] we need to produce a user guide for our application. [15:55] Which is product aware? [15:57] so that [15:58] people can interact with it as they’re using the software and say hey. [16:01] How would you configure this box? What are my options? How what’s your what’s your recommendation? [16:06] Now if we simply said ok. [16:08] upload [16:10] Or the confluence wiki. [16:13] What would we get? [16:14] we would get [16:15] a lot of really interesting materials [16:19] that look right. [16:20] Until that one moment. [16:22] Where I don’t know the response that you’re given to how do you configure a checkbox? [16:27] Is now based on 15 years? [16:29] old information [16:31] so you have to [16:33] do the heart of work first? [16:35] And that is essentially the lesson that we that we learned. [16:39] Around these lol lamps. [16:41] and [16:42] you just gotta go. [16:43] Rewrite your materials Curate your materials scrub your materials. [16:47] Ultimately you’ll be better off. [16:49] Because the beneficiary that’s using these tools in order to actually produce the next level of answers or Q&A is going to have the right answers. [17:00] We’re not going to end up in a situation where somebody. [17:02] Says you know what? [17:04] I don’t trust this thing anymore. [17:07] and that is the worst thing that can happen right this Revolution [17:10] around AI [17:11] is happening because [17:13] we want it to write we’re trusting in these machines to essentially do our bidding. [17:19] But let’s give it a chance. [17:20] Let’s give it the right information. [17:22] And Ezra I’ve really appreciate you being on the show and being able to offer these insights in terms of how we’ve seen. [17:28] AI role out & you know obviously the the opportunities along with the mistakes. [17:33] Wanted to leave our audience with some of something. They can get excited about forward looking if we were to go you know in the near term three to five years with this technology. What is your expectation in terms of what it could bring? [17:43] To the industry, and how do we get there and as or maybe you can start us off and then oh like you can hop in. [17:48] yeah [17:49] I see this. [17:51] crazy horse race in my head [17:53] right horses are lined up. [17:56] They they start the race. [17:58] And it’s literally like every step. [18:00] The next course of the line is now in the lead. [18:03] And I see this happening for the next two years. [18:06] things like [18:07] The model doesn’t matter anymore, we’re going to hear. [18:10] Alright everybody is going to have a model that does something. [18:13] And it’s going to be good enough. [18:15] And good enough. [18:16] Let’s not interpret good enough. [18:18] Today from good enough a year from now. [18:20] With this technology. [18:23] What I see. [18:25] being solved within this this space is [18:28] around data [18:30] portability [18:31] the biggest struggle we have is [18:34] getting our data to the right place. [18:36] we have [18:37] petabytes and petabytes and petabytes of information [18:40] that we are using. [18:42] in order to [18:43] ask questions and get responses to [18:46] how do I get this from my data center? [18:49] into open AI [18:51] how do I get it to Google how do I get it to meta? How do I get it to? [18:55] you know [18:55] whatever it is the next company maybe you’re going to start an AI company. [18:59] To do prediction next week. [19:02] This is the fundamental problem. There’s this massive barrier to entry. [19:07] And at some point in time. I think we’re going to wake up and again. [19:11] The models are all going to be good enough. [19:13] And we’re going to be fighting with. [19:16] Who’s the easiest to use? [19:19] That’ll become the barrier tangerine tree. [19:21] And then after that it will become. [19:24] I need specialisation. [19:27] We are a market an ad tech company in the marketing space. [19:32] Working with cmi. I need something just for that. [19:37] Where is that evolution? [19:39] So I think we are. [19:41] You know whether you call it an inning in a baseball game. We’re not even in the stadium yet. [19:46] I’ve heard second inning I heard late third inning I say we’re not even in the stadium. [19:51] Yet, I like yesterday. [19:53] You know two days ago. We went to bed? [19:55] And then on what may 14. [19:57] We watched a company. [19:59] having a conversation with their phone telling jokes speaking Italian [20:03] out of open AI [20:04] that was crazy. [20:06] What’s going to happen tomorrow? [20:09] I would add to that. [20:11] point [20:13] just pragmatically speaking. [20:15] the billions of those being vested right now [20:18] Haven’t found a way to. [20:20] Justify themselves yet. [20:22] This all these investments. [20:24] Don’t have a model how to how they’re going to make money yet. [20:28] At some point. [20:31] People who are investing will need to show results. [20:33] And then we’ll really see. [20:35] how [20:36] things will change [20:37] at the really kind of [20:39] macro level [20:40] the billions of dollars being invested into openai Google Facebook [20:45] none of this is making any money yet. [20:47] But billions of dollars are being spent at some point. [20:49] they’ll need to [20:51] show [20:52] Returns [20:53] that will change everything. [20:55] Because that will change. [20:57] What people willing to pay for? [20:59] And what is going to be sustainable? [21:01] to maintain [21:03] these technologies because they’re very expensive. [21:06] So that’s kind of again macro. [21:08] statement [21:09] on more of our short term [21:11] opportunities level [21:13] next step to me again is continued digging into. [21:18] Tesco automations of different processes that will really move the needle [21:22] in a modern agency [21:24] again, I represent an agency. [21:25] I need to look at what will help my clients? [21:29] the next big thing [21:30] we see is media planning. [21:33] Media planning historically has been driven by a tonne of data inputs. [21:38] To decide which channels make the most sense for the different campaigns. [21:41] Like I said we have. [21:44] decades of data [21:45] Of what works what doesn’t work across many different channels all that data? [21:49] Will be feeding. [21:51] our decision engines [21:52] to recommend Media plans [21:54] to our buyers [21:56] and I Believe by the end. [21:58] Of this year. [21:59] Our Agency will build that where. [22:02] Already in in design stages. I think we will have an ai-driven. [22:07] Algorithm that will be recommending Media plans. [22:11] to our [22:12] Media teams [22:13] like you both said, it’s it seems like we’ve only scratched not even the surface of this if we’re going to use the iceberg metaphor there is just so much that’s going to be out there to actually uncover and [22:23] And live out the potential I appreciate you both being able to detail. [22:27] The before and present but also looking into the future of what this all going to mean for the industry, so I appreciate you both being on the show and being able to offer those insights. Thank you. Thank you very much.