The World Communications Forum Association in Davos (WCFA) hosted a PR masterclass on AI and communications led by Purposeful Relations co-founder and PR Futurist Stuart Bruce, and Aaron Kwittken, founder/CEO of PRophet and CEO of the Comms Tech Unit at Stagwell Marketing Cloud.
AI generated transcript of PR masterclass on AI and communications
This an AI generated transcript created using Happy Scribe. Check against video as it is automated and has not been checked or corrected.
[00:00:06.100] – Aaron Kwitten
All right. Welcome, everybody. My name is Aaron Kwittkne. I am a founder and CEO of a company called profit. I’ll get into that in a little bit. I also run the comms tech cloud for Stagwell. I’m here from One World Trade Center in New York City. It’s morning for me. I woke up very early for you guys. That’s how much I care about the topic and about our community. The haze you see is from the wildfires in Quebec. They’ve been so generous to send their smoke our way. I’m joined by Sir, I’m going to just knight you now, Stuart, Sir Stuart Bruce, who joins us from Leeds in the UK. Pr futurist, co-founder of Purposeful Relations, and also founder and CEO of Stuart Bruce Associates. We purposely titled this webinar, Communications engineering, almost everything you need to know about artificial intelligence and communications. The Almost is very important because we could probably talk about this for hours, and we’re still early days. It is moving very quickly. I’m speaking on behalf of both myself and Stuart in that you are, I think, fortunate in that you have two practitioners that have multi decades of experience, not just in PR, but also have a great passion for all things, technology, AI and comms.
[00:01:30.860] – Aaron Kwitten
We’re going to cover a couple of things today. What exactly is AI? How is it being used, both generative and predictive? I’m going to cover off on that. What is a prompt engineer? What are prompts? Legal and ethical considerations. And ultimately, and probably most importantly or consequently, what type of impact will AI have on our field today and in the future in multiple aspects. We’re going to break this down into two different or three sections, if you will. We have about an hour. The first is Stuart, we agreed, is going to handle all the really, really difficult topic questions. More about the consequences, how it’s being used, things we need to consider. I’m happy to chime in on that as well. I’m going to go through an explainer, breaking it down for some of you. You probably know this, for others you might not. So I’m going to break down what is generative AI, what is predictive AI, how does it work, large language models, things like that in very simple terms. Then we’re going to open it up to questions. Everybody here has the ability to unmute themselves. When we get to questions, you can feel free to put it into the chat.
[00:02:41.450] – Aaron Kwitten
We’ll monitor that. But you can also just raise your hand, use the emoji to raise your hand. We’ll try to get to you. Having done many, many, many of these over the last couple of years, actually, I can tell you that we’re going to run out of time. We’re going to have more questions than we do have time. That’s perfectly fine. Maybe we’ll do a part two one day. I’m not sure. For those of you who are not part of the World Communications Forum Association, it is technically headquartered in Davos, Switzerland. It’s an international organization, and the goal is to promote and enhance the standards of comms and public relations globally. It’s a platform for all of us professionals, practitioners, academics, industry leaders to basically do what we’re doing today, exchange ideas, discuss latest trends, challenges, and opportunities in the field. Stewart, when he presents his few slides at the end, will also address a few opportunities and a few programs that the World Communications Forum Association has upcoming. I will now pass it over to my friend Stuart to take it away. Then I’ll come back and torture you with some of my slides, and then we can have, hopefully, an incredible open discussion about this really cool topic.
[00:03:57.530] – Stuart Bruce
Brilliant. Thank you for that amazing introduction from New York. I’m looking forward to being there in a few weeks’ time. What can we actually do in an hour? Because this topic is so big, we’re only going to skim the surface. What I was going to kick off on was just looking at some of the top level issues that we should be aware of. Then Aaron can dive into some of the detail of practically what it all means. But actually, for most of the time, it would be great if we could have a proper conversation and people could jump in there with their questions because there’s a lot of information out there on AI and AI and communications. So we don’t necessarily want to cover things that you already know. We want to try to cover some of the new aspects. So I’m just going to quickly share my screen now and run through some quick slides. So first up, I just want to do a little bit of an intro about who I am. I’ve been doing comms and PR for 30 years now. One of the things that I’ve done throughout this period is I’ve always really been into what’s changing, what’s changing in our industry.
[00:05:25.210] – Stuart Bruce
I’m old enough to remember when email first came in, when I first started in PR, we were dependent on fax machines. But I’ve always been a believer in embrace new technology and do as much as we can. Now, today, I run a couple of companies. One is Stuart Bruce Associates, and we specialize in crisis communications and reputation management. But the one that’s most relevant for today is Purposeful Relations. What we do is we work with inhouse PR teams and PR agencies to help them to innovate and improve. We identify the new technologies that they should be using. We tell them which are the hype that they should avoid, and we make sure that their team has the right culture to be able to embrace and benefit from it. So that’s who I am. But what I’m going to kick off is this quick…
[00:06:19.570] – Aaron Kwitten
[00:06:20.180] – Stuart Bruce
Do we need to know about AI and PR comms? Well, I think the first point that I want to make is that it’s not new. Since an event last year when chat GPT came out, everybody’s been getting very excited about this. And it’s been at the top of everybody’s agenda. But actually, if you look at when Aaron’s company started, he was around way before that. So there’s been AI in PR for a long time. Back in 2015, I was talking with companies like Automated Insight, who’ve got a tool called Wordsmith that does generative copywriting. It could take a data set and write stories from it for things like sports results or election results. So it’s not new. We’ve had a long time to begin to think about what some of the implications of AI and comms are. And then organisations like the Charter Institute of Public Relations, it was publishing papers on AI and comms back in 2018. It’s been looking at the ethical implications of it. It’s been looking at some of the practical implications of the tools that we can use. That would be the first point I would make. It’s not new.
[00:07:31.830] – Stuart Bruce
So somebody’s telling you that it is, that’s an instant sign that they don’t really know this space. Second point is it impacts everything. So the bit that’s getting everybody’s attention is generative AI and looking at tools that do things like chat GPT that can write copy for you, can answer questions. But that’s only part of it. Some of the other uses of AI that I’ve seen or have been using are things like using AI to meta tag videos. I did a project in the EU where they had hundreds of hours of video footage of European leaders attending events and cabinet ministers, et cetera. They could use AI to identify who those leaders were, some of the things that were in those videos, tag it all. So for the team that came to edit it, instead of having to plow through hundreds of hours, they could go straight to the bits they needed to use. And that’s been around. That’s not particularly new. Whether we’re talking media relations, whether we’re talking crisis communications, stakeholder mapping, all of these things are being impacted by AI. So that means it’s a huge opportunity for us. There are so many ways that we can embrace it.
[00:08:54.370] – Stuart Bruce
And the flip side of that being is that just because we’ve got all these amazing tools, it doesn’t necessarily mean that they don’t need humans anymore. One of the reports that the CIPR published was called humans Still Needed. I actually had a call last week with Jean Valin and Professor Anne Gregory, who were the two main authors over it. We’re just kicking off a project now where we’re going to be producing a second version of it. I can spot one or two names on this call that will be getting a call from one of us because we’re going to be doing some qualitative interviews to get people’s views on where all this is going. That report is going to be published later in the year. So it’s an opportunity, but there are challenges. One of the things that worries me a lot is new entrants. I started off by saying that when my first job, I was doing things like cutting out press cuttings, I was sending faxes, all fairly mundane stuff. But actually, I learned a hell of a lot, pardon my language, about public relations and communications and what journalists want, what clients want, what politicians want, just by doing some of those mundane jobs, if those all disappear, how do we provide those roots into the industry for new people?
[00:10:15.830] – Stuart Bruce
That’s something that we need to tackle. I think a lot of the threads of AI about losing jobs, I think are being overblown, but that is one that concerns me personally. So I’d be interested to know what other people think on that. Then, of course, AI has risks. That’s one of the things I wanted to focus my introduction on now about what some of those risks are and how we can mitigate them. I call it the ABC of risks. It’s accuracy, bias, copyright, danger, and ethics. Accuracy. I’m not going to delve into all of these in-depth, but accuracy is about AI can hallucinate. Just because AI is chat GPT, it can make stuff up. Now, does that mean we shouldn’t use it? Of course it doesn’t. It just means that we check it. There’s this wonderful phrase that’s used in US journalism school, but if your mother says she loves you, check it out. Well, if AI tells you something, check it out. It doesn’t mean you can’t use it. Bias. Now, this is a real issue because a lot of these models, they were trained on material that’s out there. If the material out there is biased, if we don’t have enough female CEOs, if we don’t have the right reprisation of race and gender, that’s all going to have an impact.
[00:11:42.350] – Stuart Bruce
And it was trained on Western models. A lot of this comes from the US and Western Europe, so Africa, Asia isn’t necessarily represented as well as it should be. So that’s another place where bias can come in. But we need to be aware of these things. Another part of accuracy and bias is there’s another element of AI. It’s when we start to use it on our own data. So that’s another really powerful use. If we’re using AI to analyse our own corporate data, it’s up to us to make sure that that data is representative and to do things to eliminate that bias element. Copyright is something else to be aware of. And there’s two issues to this. There’s the people who are saying, Oh, it’s been scraping all this content. It’s using copyright material. This is terrible. Yeah, I can understand that argument. The flip side of that argument is who hasn’t been to an art gallery and been inspired by the pictures they’ve seen? Who hasn’t seen the work of an amazing photographer and been inspired by that? Who hasn’t read an author and been inspired to write in their style or to maybe use some of their plot devices?
[00:12:54.980] – Stuart Bruce
That’s not stealing their content. It’s inspiration. So there’s a question of how much of AI is scraping content, how much of it is inspiration by machine on a massive scale. The other side of copyright is who owns what’s being created. Is that going to be owned by you as the prop, putting in the prompt? Or is it going to be owned by the AI engine? So if we’re going to be using these things for work and professional purposes, we just need to think some of these issues through. The danger, well, it appears hardly a day goes by without people talking about the danger. The latest one we’ve had is it’s going to lead to us ceasing to exist. It’s going to be bigger than the pandemic. I don’t really believe that. Of course, there’s risks there, but we just need to mitigate them and to be aware of them. The one I wanted to just touch on now is ethics. Because a lot of these issues, if we’ve got a strong ethical framework in place, we’ve got a way of dealing with it. So one of the things I always encourage my clients to start off with is actually just think about how they’re going to use it, what rules are they going to put in place themselves.
[00:14:06.710] – Stuart Bruce
And one of the tools that I use for doing this is something called the Alethia Framework. Rolls Royce has been a client of mine, but it was advising them on small modular actors, mini nuclear power stations. It wasn’t on this project, so I didn’t have any involvement in this. But Rolls Royce have been using AI for decades to do things like assess the safety of aircraft engines. That’s life critical stuff. So getting the ethics of AI right is critical to them. So they’ve taken all of that knowledge and experience and created this framework that can be applied to anything. It’s been applied to the health sector, it’s been applied to the music industry. You can take this framework and use it yourself to think about how can you use AI in the most ethical way. And it’s got three main aspects to it. One is looking at what the social impact of what you’re doing is. The second one is the one that we’ve spent a little bit of time talking about is how accurate is it? How much can we trust it? And then the final one is governance, putting your own rules in place, but also looking at what government is doing, the US, the UK, the EU, they’re all progressing with regulation, but we’ve got to make sure that we’re guiding politicians and legislators to make sure that it’s the right legislation.
[00:15:33.720] – Stuart Bruce
Okay, so what’s next? Think big, start small, act fast is my maximum AI. So you’ve got to start doing things now, even if it’s proactively deciding what you’re not doing. This space is moving so quickly, you’ve got to be touching it and understanding it. An experiment. If you’re not using it professionally, then create a personal chat GPT account. Actually, you shouldn’t be using some of these tools using your corporate accounts. The best way to play with them and experiment is personally. And don’t put your confidential corporate information in there unless you know how it’s being used. We’ve had Samsung, that was one of the examples we saw where people were engineers were putting things in there. We’ve covered what AI’s limitations are. Fx, we’ve covered. So if you’re going to use AI, first thing to do is what do you want to do with it? How can you align it to your businesses or organisations goals? But as PR and comms people, we actually also have a second responsibility. It’s not just about how we use it. We’ve actually also got an onus to advise our leadership teams on the implications of AI, on what the organizational implications are going to be, what the reputational, what the relationship implications are going to be, because we’re going to have colleagues who are plowing ahead, experimenting and doing things without really understanding some of these issues.
[00:17:06.600] – Stuart Bruce
So we’ve got to educate and inform those colleagues throughout the organization. And I think one of the ways to do that is set up a group, a company wide group of some sort that is going to tackle this in the round. So not just from a comms perspective, but actually from an organization wide perspective. And then make sure that group knows what it’s doing, that it’s got some expert counsel and advice there. I’m going to leave my next couple of slides and come to them at the very end. I’m going to hand back to Aaron now to do his bit.
[00:17:47.440] – Aaron Kwitten
Thanks, Stuart. That was awesome. That was amazing. As always, you never ever disappoint. I’m going to share my screen. The first slide you’ll see, which is fun, is Stuart and I both like to use pictures of ourselves when we talk about ourselves. Just to prove the fact that we were speaking somewhere else. So we are expert speakers. So there’s mine. This is actually from, let’s see, South by Southwest at an Axios event where we launched our generative AI feature called tailor. Just real quick, I mentioned this at the start. Like Stewart, I’ve been doing this for more than three decades. I feel it every morning when I wake up. I feel it in my bones. Now I feel it in my brain. I am an agency guy. I like to describe myself as a recovering agency person. I’ve managed large agencies over my career. I started my own almost 20 years ago, sold it to a company called MDC Partners, which is now Stagwell. I’m a podcaster. I contribute to The Drum. I wrote for Forbes for 10 years, which is part of the impetus for starting PRophet almost four years ago. And I’ll talk about that in just a second.
[00:18:53.930] – Aaron Kwitten
Because when I did create the MVP or the minimum viable product for profit, people thought I was insane to leave my day job because I’m like, It’s all about AI and comms. The world is changing. Comms tech is going to be our friend. It’s going to be very important. It’s going to be consequential. And then chat GPT came along, people understood it. But also they also chat. Gpt, I think, obfuscates really the potency of what we’re talking about. And like Stewart, I’ve got multiple jobs. I’m an immediate past president of PRSA in New York. I was on the PR Council as a trustee in the US. Many of you are probably familiar with ICCO. I was Head of the Americas for ICCO. I’m also on the Board of the School of Journalism and Public Affairs for GW University, my Alma mater. I run this thing called the CommsTech Cloud at Stagwell. And we have two main flagship products. Both are AI driven. One is for earned medium and the other is called qualified and it is an influencer and creator platform. I’m going to go through a little bit of an explainer because I think it’s helpful.
[00:20:01.780] – Aaron Kwitten
Let me actually go into slide show view. Sorry about that. All right, that’s better. So real quick, there are really two ways to break down AIs or relates to comms right now. Putting aside some of the image work, right? One is generative and one is predictive. The generative part is basically anything. It can generate data, audio, code like GitHub. It’ll help create code for you, images, texts, simulations, 3D objects, videos. That Rolls Royce case study is super cool. I’d love to learn more about that. And then there’s predictive. I don’t want us, and Stewart mentioned this, I don’t want us to get caught up in the generative. That is cool. It’s interesting. It’s scary. It’s a little sexy. Fine. That’s the generative stuff. But I think the real future for AI when it comes to PR is predictive. Why do I say that? I say that because predictive actually is going to help us see around corners. It’s what our clients, what our higher ups want. And what predictive AI is doing is we’re moving away from Boolean search, typing into a search bar using keywords. And now we’re moving into this space called Vectoring, Machine Learning, Semantic Search, which means it’s contextual, to now estimate the proximity of interest and intent based on past behavior.
[00:21:20.520] – Aaron Kwitten
What do I mean? When I say the proximity of interest and intent based on past behavior is the reason why I created profit. If I could look at what reporters have written on the past, I should be able to predict whether or not they’re going to be interested in a certain topic in the future. That’s what predictive AI is. When I think about the consequence of AI and comms in our world, I think about this notion of a communications engineer. That is not a skill set, it’s a mindset. It means how do we tether together predictive and generative? I think it’s Roy Perez asked… I’m sorry if I mispronounced your name, asked in the chat earlier, This thing is going to automate, what does it do to our relationships? I think the way to combat or to address that challenge is in the middle of predict and generate is personalization. So reporters should not and should never get the same pitch over and over again. But guess what? That’s what people are doing without AI, because there’s a complacency that’s been built into our business based on these large analogue media database companies where people spam the shit.
[00:22:27.430] – Aaron Kwitten
By the way, I do curse a lot. I’m sorry. I’m just an ugly New Yorker. Spam the shit out of reporters. And then we live in the sea of despair and rejection, expecting reporters to react to us and be like, Oh, yeah, I want to write about that. It doesn’t happen. Ai can actually help us personalize which we should be doing, pitches to reporters. This communications engineer sits at the intersection of art and science. They manage, disperse narrative and content using data because ultimately, because unfortunately, our business has been based mostly on experience and instinct. We don’t have the data that our marketing siblings usually have. And the communications engineer is not going to focus on workflow and CRM and automation because we’re in a business of articulation, not automation. They’re going to have this agile mindset. It’s going to be a flywheel of text text. We’re going to predict journalists and influence and create our interest and sentiment. We’re going to generate and regenerate resonant content. We’re going to look ahead to see what trends there are. We should be able to surface encounter mis and disinformation, even if it is accelerated by AI because that is human behavior.
[00:23:35.780] – Aaron Kwitten
And ultimately, we should be able to use CommsTech, protect brand integrity and safety. So much of our business has been focused on how did I do? Trying to figure out attribution models and monitoring. Not enough has been focused on how do I do better? And that’s where I think AI can play a huge role. So I always like to quote the CEO of Microsoft because why not? And I think this is what Stuart was saying earlier when he talked about how humans are still needed. This is not man versus machine. This is about human with machine. This is about making sure that we all have this communications engineering mindset so we can be more productive and performative. It’s the three P’s. I’m sorry, I know it’s a little silly. And I have a very strong opinion that if we embrace this, we slay today or we sing tomorrow. I really believe that. And I’m sorry for being so binary about this. And I do recognize and appreciate all of the risks, the ABCs that Stuart had mentioned before. I agree with every one of them wholeheartedly. But ultimately, as an industry, we’ve spent decades now, we spent all of our hours helping others tell their digital transformation, their innovation journeys, yet we don’t look at ourselves.
[00:24:49.310] – Aaron Kwitten
We don’t talk and ask, Okay, how is it that we can perform better? The greatest accomplishments or achievements in our business has been email and social media. That’s pretty much it in the 30 years that I’ve been around. That is insane. So how do we use AI and comms tech to replace instinct with intelligence, move from analogue to digital? I don’t believe in media databases. Ai will help us find media targets. It’s no longer about media monitoring or surveillance, it’s about safety. How can we also reduce our time? Can we save time? And let’s get out of this workflow solutions mindset. We shouldn’t be on just one platform trying to automate and take a CRM, which is a commoditized approach which does erode relationships, to building relationships as reporters. Sorry. Okay. People often ask, Well, what book should I read? How do I get smart about this? This book is not about AI, but it was very influential for me back in 2018, 2019 when I started PRophet. Gill Lepore is a very prolific writer for the New Yorker. She wrote this book called If Then, and she goes way back to the 50s. I mean, if you want to go to the touring test, we could do that as well.
[00:26:03.920] – Aaron Kwitten
But AI, to Stuart’s point earlier, has been around for a very long time. She specifically focuses in on this thing called the People Machine. That’s what it was called by a company that helped Kennedy actually win the election in 1960 in the States. And that was the first time large data sets and data science was used originally for CPG and for advertising companies and then put into the political realm. There it is. That’s the people machine. That was some of the beginning of the commercialization of AI as it relates to finding which consumers would be interested in what types of topics and trends. So why is this so important? I don’t think it’s important. I think it’s consequential. One, this is going to fundamentally change the way we work. It may even change the agency business model. Our agency business model is a triangle. All the junior staff are at the bottom. Will it flip it? Can we flip it? But at the same time, to Stuart’s point earlier, we learn from the mundane, right? I am concerned that my kids do not know how to read a map because they grew up with GPS.
[00:27:09.150] – Aaron Kwitten
I think this is going to dramatically change the way agencies and brands produce, deliver, and measure their work. And think about it, by 2030, the AI market is going to be a two trillion dollar industry. I think that number is actually low, but it’s absolutely incredible. So a couple of quick things, just some explainers, because when we talk about AI, a lot of people think it’s artificial. It’s not really artificial. It’s sometimes just augmented. But what are we talking about when we talk about text to text generation? Large language model. You probably hear this all the time. That’s what chat GPT is. It’s trained on huge amounts of data, and they’re basically generating new phrases or sentences to make educated guesses about those topics with similar meanings. That’s what text to text generation is. Some examples, right? It could be, again, it could be a diffusion model. And there’s a couple of examples here. When we talk about neural nets, you see that on the left, that means it’s learning from itself. It’s a neural network. It’s trying to basically replicate how our brains work. On the right, there are diffusion models, right? And that’s more about images.
[00:28:18.810] – Aaron Kwitten
Specifically, we’ve heard about MidJourney, DALLI-E, Stable Diffusion. Stuart’s point earlier, it’s really fun to play with these things. These things are toys right now. They will be tools. And when you do experiment, experiment with tools, tools that have SLAs, that have agreements, that have walled gardens. If you want to generate something and you’re worried about copyright and you want to generate an image, I would tell you, go to Adobe because they have a walled garden. Adobe guarantees that all of the images that they are using to help you generate an image have been signed off on. They’re compliant with copyright law. That’s incredibly important and it also comforts me. Just a couple of examples here. These are all images generated by AI. I’m sure some of you on this call have played around with this stuff. It’s pretty cool. These are all generated by MidoJourney. We actually have some photo credits here. I guess the question is, is it really theirs? I believe it is. That’s also for another debate. The other thing I’m really excited about and terrified is this notion of text to music, text to 3D, text to speech. These are all examples of companies.
[00:29:31.110] – Aaron Kwitten
Google can do text to music. Luma AI does text to 3D. Replica can do text to speech. Text to video, Meta does that. It’s pretty incredible. One thing of Stuart didn’t mention that absolutely terrifies me is synthetic media. So basically there’s a lot of video and audio of me and Stewart out there. And it’s not hard for any of you on this call, probably in under a day, maybe even under an hour, to create a fake video or fake audio of us saying abhorrent things, which ultimately… Thank you for the Kennedy sign there, Stewart, I see that. Which ultimately could lead us to be canceled. So that terrifies me. And I think that that’s only going to get accelerated. And we can talk about that in the MIST disinformation realm later. Many people, you’re thinking about this notion, everybody talks about this thing called the transformer model. This was the beginning of the more mainstream, wider adoption of things like ChatGPT. Actually invented by Google researchers, basically, it could do things like understand protein, write software code, make online recommendations. They are trying to learn context and relationships. It’s all about relationships. It doesn’t require a lot of data.
[00:30:50.140] – Aaron Kwitten
Most of the large language models that I was talking about before are built on these transformers. And that’s what’s happening when you type in a word or a phrase and expect it to then spit something else out. Here’s an example of it. The disclaimer here is this is incredibly oversimplified. You have this huge corpus of data. You’ve got the large language model that’s training on the data, trying to find context. Then this is where you come in as a human. You’re prompting and it gives you a response. And to Stuart’s point earlier, if not always correct, that’s okay, because you have to think about this loop as you’re either in the loop, you’re on top of the loop managing it, or you’re out of it. We should never be out of the loop. That’s when AI is making decisions for us without any human intervention. That is fucking scary. That should never, ever happen. We always need to be in or on the loop to manage it. Some of the world’s greatest AI driven companies like DataMinr that helps to assess ground truth, they have these folks called domain experts who have PhDs in data science.
[00:31:57.570] – Aaron Kwitten
They make sure that the AI is correct so they’re able to validate and verify the data that’s being put through and then spit out of the system. So how can they work better? Basically, you have this large corpus of data. Again, you’ve got the training. And then this is what I talk about you being in the loop, reinforcement learning with human feedback. This is where we come in. When you want to take a press release and punch it up into a pitch or write a crisis statement, you don’t just let it go out into the world. You need to edit it. So when I built profit, I thought about it in two ways. One is how do I surface which media are going to be interested in my pitch by uploading a pitch? And also how do I make my pitch more interesting? T hat’s the generative stuff. Ultimately, my goal on the generative side and everybody’s goal should be can AI get me 50 %, 60 % way there? In the same way that when I hire someone for a first draft or a second draft, I know it’s not going to be perfect.
[00:32:57.080] – Aaron Kwitten
I’m always going to come in over the top. But you need to create this loop. There needs to be a feedback model, and that’s how large language models will get better. So what happens behind the scenes? What does it do with my prompt? It looks at the prompt, it converts it to tokens. This is what happens. I’ll go back. They’re the words, there are the tokens. We can’t understand that. They’re basically computing words. They’re transitioning words into tokens or numbers and then finding, basically embedding, what is the context of what you’re putting in and trying to relate it. That’s the semantic stuff I was talking about before. And it’s running in seconds or nanoseconds through layers of this neural network to produce this. And that’s where this transformer is coming into play. It’s trying to learn what your intent is. And it generates thousands and tens of thousands of different probabilities into the next tokens. And then it gives you basically what you asked it for. This all happens in a matter of seconds. The reason why I wanted to share this is because it is mind numbing, mind blowing to me that this is happening in seconds.
[00:34:07.530] – Aaron Kwitten
But that’s what’s happening with Microsoft Azure, with Open AI, which is what we sit on with Google B ard and these other large language models. It’s literally making up a response based on mathematical probabilities. It’s going to get better. What we’re seeing right now in some of the hallucinations, I don’t know. When I first asked ChatGPT months ago who I was, it says I was a CEO of Webber Sandwick. I’m like, That’s awesome. I was never the CEO of Webber Sandwick. But it’s going to get better because that’s basically it’s learning. And especially if humans are part of the loop, it’s learning on itself. It’s doing it one word at a time. There’s no database, there’s no search. They actually don’t know anything. You know everything. We’re the experts here, and you have to keep that in mind. We have agency, we know everything. This is just a tool. And then what we end up with is something that looks like intelligence and comprehension, maybe even sentience. It’s not quite there yet, but that is what it’s aiming to do. I think of generative and predictive as your collaborator, your teacher, and an assistant. And I don’t mean that in a demeaning way.
[00:35:15.390] – Aaron Kwitten
If I go back to the slide earlier when I quote the CEO of Microsoft, this is human plus machine. That’s how I view this. It’s this collaboration loop. You’re a prompt, AI does things, it responds. You have to refine it. That is on you. And if you don’t do that, then that’s an epic failure on your part in the same way that you wouldn’t edit someone who’s on your staff’s work. How does it change the way we work? Well, look for time consuming tasks, building media lists, for example. Currently, the process for building media lists is hugely flawed, and then the process for actually reaching out to media has become very complacent. Look for tasks you dislike. It doesn’t mean that you shouldn’t do them. It just means that maybe you could use AI to help you get past that and even save some of your staff and some of the attrition that you have in your own staff. Look for moments when you need to learn or understand something. This is Stuart’s example earlier about walking into an art gallery. For me, sometimes just playing around is almost like a personal one on one brainstorm.
[00:36:19.930] – Aaron Kwitten
And I think AI can really help with that. And that really speaks point four, which is inspiration, ideas and feedback. I’m not going to play this whole thing. I just want to give you just a quick example of what I mean. So this is the profit platform. I took a press release that is about 12… Actually, it’s 982 words long. You can see it down the lower left hand corner. And what I’m going to do here is I am basically going to punch it up into a much tighter pitch. It is abhorrent, well, it’s probably too strong of a word. It’s not great when PR people just send press releases and spam journalists without any personalization or context or anything. So what you can do here, and this is a great example of AIs, you can pick a tone. Again, this is using a large language model, and then you could, in about, I don’t know, on average, about 26 seconds, not only create or generate a pitch that might take you a couple of hours, but you can also generate social posts. Now, I’m actually doing this in real time. It’s a video, but I’m doing it in real time, so I didn’t cut the video to show you this in two seconds.
[00:37:30.080] – Aaron Kwitten
The idea here is, and this is I think how PR is going to be aided, the idea here is that how do you take something, any content, it could be even 30 words of a prompt, make it really interesting, and then use that to then predict what reporters would be most interested in that pitch. And there it is. There’s the original content on the left, and then there’s the generated content on the right, and then there’s social posts that we created. That was very, very quick. Is it perfect? No. Should it be perfect? I hope not because I don’t want to be replaced. This is again, where you come in over the top as the assistant to make it better. You can copy and paste it into your social channels and then edit if you want to. And then you can pick the generated content that you think is most interesting and then you can play pitch it, basically. So you’re pitching in the Cloud. We’re going to be testing our pitches in the Cloud going forward before we test our pitches in real life for reporters. Reporters don’t want more less interesting, less resonant emails in their inbox and pitches.
[00:38:26.940] – Aaron Kwitten
If you want to create a relationship with a reporter, stop spamming them. Actually find out which reporter is going to be most interested in your pitch. And you do that by actually evaluating it. And then you’re able to… This is natural language processing at work, right? We’re able to figure out what categories are most likely, or beats, if you will, are most likely relevant to this content. And it can do it very quickly. And if you don’t see a category that you think it would be a health story or finance story, you can add that category. And then you are literally pitching in the cloud, and that is the future of PR. I really believe this is it. Is it perfect? No. Do you still need a human to come in over the top and to assess what reporters are going to be most interested or the right reporters in this pitch? Absolutely. You could pitch podcasts. You’ve got different categories here. I’m not going to have it go through the whole way, except I just want to show you one quick thing, which is, ultimately, you can personalize pitches. And you can regenerate and create a personalized pitch like that to a reporter.
[00:39:29.030] – Aaron Kwitten
And you can have pitch points on top. Tell me why. Why is it important? It’s like a math teacher saying, show me your work. This is going to be the future. You’ll be able to generate 40 pitches to 40 different journalists that are personalized in under five minutes. This is the future, and I don’t think it replaces us. I think it just makes us more performative and predictive. We did a state of CommsTech report. If you’re interested in it, you can see the URL below or you can just email me or link in with me. We identified six major trends for this year. One of them that Stuart touched on is this notion of bias. And it’s not just bias in AI, but I actually think about it in the reverse. There’s a lot of bias when it comes to influencers and content creators. At least for sure, I know in the States and in Canada, black and brown influencers are paid 35 % less than their white counterparts. I believe that AI can create a level of transparency in a business that’s very opaque when it comes to these contracts between brands and influencers so that everybody achieves pay equity at scale.
[00:40:32.620] – Aaron Kwitten
So that’s just an example of potentially AI for good. And if you want to keep up on some fun AI news, I think I have you on this, Stuart, we have this thing called Sunday Brunch. Why is it Sunday? Why is it Brunch? We wanted to do something different. We wanted to spam you on a weekend. Basically, it’s a very short free newsletter that covers all things AI. My favorite section is the Talk nerdy to Me section, which gives you some great examples, both positive and negative, around how AI is being applied in society and also in our field. I will stop there. I know it was a lot. I talked fast. I tried to keep my cursing to a minimum, and I think we can open it up for questions. We have about 16 minutes to do that.
[00:41:21.260] – Stuart Bruce
Brilliant. Just that was excellent, Aaron. Thanks for that. While people are thinking of questions, you just prompted one or two thoughts. First up, I’m really glad you swore because I’m from West Cumbria and we use swearing as a form of punctuation, so I’m more relaxed about that now. This whole thing about performing better, I think it’s absolutely critical. I’ve been banging on about this for years about there’s too much obsession in PR with proving our value and looking at what’s already happened. The real value, a lot of the work I do is measurement and evaluation. But the key part of that is learning and the incremental improvement. And just think, if you’re going to go in and talk to a C suite or a CEO, what are they going to put most value on them? You’re telling them you spent their money and it was okay, you got some results, or helping them to predict the future, helping them to make better business decisions. It’s obviously the latter. And that’s where AI can help us to do that. And this amazing tools.
[00:42:23.410] – Aaron Kwitten
Can I add one more thing? I’m glad you mentioned that because I often say, having been in this business so long and served clients for so long, people think they’re way more interesting than they really are. And then the ad guys are able to pay their way into creating interest. We have to earn our way into that. And ultimately, we are stuck with the hard conversation, the client saying, actually, that story doesn’t really have a whole lot of substance behind it. And then they take offense to it and then we get fired or whatever, right? Or there’s tension. And I think that being able to have data to prove out or at least create a more thoughtful discussion around interest is something that we’ve never had before. Otherwise, it’s just our charm, our good looks, maybe, our instinct, our experience in trying to convince somebody that that is a good or not a good story to pitch.
[00:43:24.290] – Stuart Bruce
Absolutely. The other thought you sparked is that thing about PR not really innovating. I remember speaking, I don’t know whether Tom Murphy is on this call, but 10 plus years ago, I spoke with Tom Murphy, who used to be a senior Microsoft PR guy. He’s now a company called Global Head of Comms for Veeam. I remember Tom saying there that PR doesn’t need to worry about big data because it hasn’t even mastered small data yet. That was 10 years ago and the same would be true. We just don’t embrace these things. I think we’ve got to. That whole thing about when you start to go on to deep fakes and digital humans and what you can do with audio and video, hey, that’s a whole topic for another master class because there’s so much we could cover in that in terms of the tools that are out there, how we can use them, how we can defend ourselves against them. There’s just so much. Hopefully, we’ve got some good questions on the way.
[00:44:27.990] – Aaron Kwitten
I can take that if you want. Simon asked, In terms of pitching the Cloud, is it essentially aggregating journalists’ coverage then figuring out what will push their buttons? Yes, pretty much. It’s aggregating coverage over the last few months from high authority and high authority outlets and verified journalists. And what we’re doing is we’re associating what they’ve written on topics and adjacent topics to then predict whether or not to be interested in this. And we weight it and we’re constantly working at data scientists to reweight it because sometimes we get right, something we get it right, sometimes we get it wrong, sometimes it’s somewhere in the middle. And Roy, I think that’s how you say your name, Roy Perez. Yes, for the most part, we’ve seen a lot of uplift. We’ve also seen some expansion. I think one of the challenges as PR people we have is we know our friendlies, we know the endemics. We always have clients or higher ups in organization that say, Well, I want to be in fast company. I want to be in The Guardian. Whatever it is, they want to stretch. They have stretch goals. I think that ultimately that’s where these tools can, like profit, can actually help.
[00:45:36.880] – Aaron Kwitten
How can you meet those stretch goals?
[00:45:41.200] – Stuart Bruce
I’m just looking at some of the other questions in there. Sophie’s asking about advice to senior leaders and how to take disinformation, misinformation seriously. I think there’s a couple of things. Well, you can just look at some of the examples out there of what’s happened and the fact that it’s actually happened to companies. Just use one example. Back in 2018, there was a deep fake audio tool used by a… It was an energy company. The CEO of the UK Energy Company thought they were talking to the German CEO. It was actually a digitally generated voice telling them to transfer money to an account, which the CEO did because he was so convinced he was talking to his CEO. That was in 2018. Give them examples like that. But the other thing you can do, and I’ve done this back in the old days where we’ve mocked up, I remember one Crisis Coms client where the VP of Europe was really alarmed by a decision that this US CEO had taken, and nothing he said would convince this CEO to change their mind. I came up with the idea of a mocking up a front page of the Wall Street Journal with the CEO’s picture on it and the headline and the story.
[00:47:00.730] – Stuart Bruce
She changed her mind, but you saw it there. But you could do the same thing with some of these deep fake tools. Generate their voice, show them potentially what can happen out there.
[00:47:12.440] – Aaron Kwitten
Yeah, it’s interesting. In the US, pharmaceutical companies, I’m sure they do this as well. I’m just using an example. They talk about how some of the pharmaceuticals are used off label. There’s an off label mentality where you can use some of these tools. We use PRophet recently, or agency did recently in a crisis training situation where we had tailor, the generative component of profit, write stories based on different crisis scenarios. And by the way, talk about large language models, we sit on top of Open AI. Open AI has a little bit of a bias towards the consumer and is a little anti corporate, interestingly. So the stories that Taylor produced were really difficult, very hard on the company, unforgiving. But it was great for a crisis simulation because you want to overengineer crisis simulations with clients. So there’s a lot of different uses with AI and also in preparation for crises and issues. And then also on misinformation disinformation, there’s a company out there, there’s two of them that come to mind. One of them is PeakMetrics. We’ve done a lot of work with them. They’re an LA based company. They are able to not only trace the beginnings of very early on, give you a warning sign, as something might be brewing online around your brand or mis or disinformation.
[00:48:41.000] – Aaron Kwitten
They could also trace its roots and see how it’s spread so you can then create countermeasures. The other company is run by a mutual friend of ours, Ant Cousins or it was, who’s now with Cision, who runs AI for Cision. He’s the head of AI strategy and it’s called Factmata. They do something very similar and I’ve always been impressed at them. Cision very smartly snap them up. There’s a bunch of other companies too, I’m sure, out there, but those are the two that come to mind.
[00:49:07.820] – Stuart Bruce
Yeah, I’ll throw a third one into that mix because those are two of the ones I would have mentioned. The third one is NewsWhip, the Irish company that does really good predictive. It’ll take a story and it will show whether or not it thinks it’s going to go viral and whether or not you should take notice of it.
[00:49:29.410] – Aaron Kwitten
NewsWhip is amazing, absolutely amazing. They just raised some money as well. And all of us have been at this for a number of years, just FYI. We’re not new. We have a lot of scar tissue here, which I think is warming in some ways because we know where the landmines are, at least we know where the known landmines are, and we’ve experienced it as well. But I forgot about NewsWhip, that’s another really great one.
[00:49:55.210] – Stuart Bruce
Okay, so we’ve got a question here from Rejoice about organisations wanting to cut cost and trying to show that humans are still relevant. I think Aaron went through a lot of that. But this thing about this has been around for a long time. So one of the clients I worked with a few years ago, we actually used the wordsmith tool because they had a couple of people whose job it was a government organization, and their job was to take the how money was being spent and take it from a database and write stories for hundreds of local media outlets about what value this government organization was giving in that locality. Well, we just took that data set and we got wordsmith to write those releases. And you know what? The quality was really good because unlike a human, the computer doesn’t get bored. So actually it was churning up quite good quality ones. Whereas with a human, they get bored and they just start doing a quick cut and paste job. But the question about value is that unlocks these people’s time to do more strategic stuff, the active stuff that was adding value to the organization.
[00:51:10.360] – Stuart Bruce
So I think basically you’ve got to show that that’s what’s going to happen, that we’re going to replace some of these mundane tasks, or not necessarily replace, but augment them. But then the value of the human is that they can do these higher level tasks that really add value to the organization.
[00:51:29.270] – Aaron Kwitten
Yeah, and I think I touched on this as well. Look, the churn rate of people in agencies, in particular in the states, is about 25% on average. So agencies in the US have to replace 25 % of their staff every year in addition to trying to grow and replace clients. And people quit their jobs for a couple of reasons. One is they might hate their boss. Two, they might think the culture is toxic. Three, they might have chosen the wrong career. Or four, which is most of the time, they just don’t like what they’re doing and they have asked to do something different because they’re not using the brains in the way that they’ve been trained to or been promised at uni. So they go to these universities, at least in the States, you pay hundreds of thousands of dollars for this. It’s ridiculous. And then now you’re doing very administrative work. And yes, you can learn from some of that mundane work, but there’s probably an intersection between learning and late and making it a little less mundane or accelerating that. So I do think that comms tech tools, if deployed correctly, will also help retain top talent.
[00:52:36.560] – Aaron Kwitten
And while it’s not going to replace people, it’s definitely going to change our roles in a way that I think is probably better and more appreciative and a creative to what we’re trying to do for our clients. And I don’t think that’s a bad thing at all. We’re going to have to learn how to put a really good prompt in, which is not that different than starting a sentence or creating things on your own. You have to think about what is the prompt that’s going to get the best reaction. Mido asks a very interesting question. So I demonstrated a tool that can target to surface journalists. Are there other tools out there that can help use AI to help target other audiences. There’s a bunch out in the paid and advertising space as well. I would also look at companies like Morning Consult, Harris, which is a sibling of ours. They’re going to have something called Ask Harris, which basically you can put in a prompt and it’ll give you survey and trending and polling data and who might be interested in something almost instantly, which is going to be pretty cool using their data set, their large language model.
[00:53:45.080] – Stuart Bruce
Just adding to that, it was interesting, Aaron mentioned Morning Consult, as one of our clients uses them for message testing. They’ll put questions up, they’ll put themes out into the market and different messages and understand which ones are going to resonate. And it’s really good. But one of the things that we’ve been doing as a company is exploring how we can do something similar. We’re in the process of working with a market research company now where we can do the same thing at quite a low cost. We can put a question out there into the field, recruiting a sample of people via social media, a representative sample, and within 24 hours actually have those results back. So we’re using it to hone the campaign before we’ve actually started it. And I think some of those uses out there are just really exciting.
[00:54:44.600] – Aaron Kwitten
Absolutely, there’s no doubt. Nobody asked this question, but in a few minutes that we have, I just wanted to pose it something else because this is a concern of mine as it relates to… It’s more of an existential concern. So we all recall when social media came out, the ad agencies tried to fight us for that. I remember I was pretty bloodied by it, but we won because ultimately social media is at its best, as we all know, and it’s authentic and organic. And it’s really more of a content exercise with creative, which we both have. We won. I do have a concern that some of the larger consultancies and ad agencies think that AI can further displace or commoditize what we do. And we need to be very careful to make sure that we do not seed our ground to these folks because they themselves are under threat for lots of different reasons, some of them related to technology. When companies or industries are under threat, they basically go after and they try to invade other territories, which in this case would be PR. I just wanted to put that out there that this is another reason why we need to stay at the cutting edge and on top of AI and comms.
[00:56:03.120] – Aaron Kwitten
If for no other reason to make sure that we’re able to protect what we do and make sure that the ad agencies and the consultancy don’t try to commoditize what we do because they would in a heartbeat and just roll it all in because I’ve been part of and I’m still part of a large holding company and I see it happen all the time.
[00:56:21.870] – Stuart Bruce
Brilliant. Well, I’m just looking at the time and I think we probably need to start to bring this to a close. Obviously, we’re both available on all the socials, Twitter, LinkedIn, so feel free to connect with us. I’ve just got a couple of slides to finish off with. I’m going to start sharing my screen again. First up, I will just… That weekly newsletter that profit publishes, I’ll just endorse, but yeah, I’m a subscriber. It’s my Sunday afternoon with a cup of coffee. I always found at least one nugget in there of something that I’ve missed or something that’s new or something I find interesting. So definitely subscribe. I’ll also give a plug for my own newsletter. So we publish every fortnight on a Monday. And it’s basically to round up of some of the most interesting research stories that we’ve spotted. It might be new technology. So I think in the last issue, there was a.
[00:57:24.320] – Aaron Kwitten
Story about… Right now, look, I just did it. I just did the QR code.
[00:57:28.440] – Stuart Bruce
Brilliant. Thank you very much. Well, I think you’ll spot a story about you in that last one. I think we’ve got something in there about your latest type with peak metrics. So that’s the type of stuff that we cover in there. So subscribe. The next thing I just wanted to touch on is tomorrow there’s going to be a new research report published in PRovoke Media. This is something that we’ve been working on over the last few months. It’s a big global survey of about 350 communications people from around the world. PRophet, very kindly, will one of our partners in doing this research. It’s going to be published tomorrow in PRovoke. We’ll also be announcing it in the newsletter. So take a look because there’s going to be some really great, interesting findings in there. The final point I wanted to touch on was just a little bit about World Communication Forum Association. Normally, we have Maxim our President to do this bit, but I’m going to give it a go. I’ve been involved in the precursor to the WCFA since about 2013, 2012, something like that. I’ve been to some of the original events in Davos, which we are hoping to get up and revive and do again.
[00:58:53.680] – Stuart Bruce
Once we finalise the details, I would really urge you to go to them. But for now, we’ve got members in more than 40 countries. Vietnam, the US, Australia, Malaysia, Ukraine, the UK, we’re represented everywhere. We do things like these master classes, but we also have… We’re going to try to get the in in-person events up and running again. One of the things you can do at the moment is we have the Global Davos Communication Awards. I chaired the first jury for the first Davos Communication Awards. They’re open at the moment. Get your entries in. Maxim tells me that we’ve made a decision that if you enter three awards, then you get free membership of the WCFA for a year. So enter free, get your free membership. Then the next thing that’s going to be launched shortly is a PR Agency Tracker. So if you’re looking for that, if you’re inhouse and you’re looking for that global support, or maybe you’re an agency and you need help for one of your clients in a different market, the depth and breadth of our membership means that’s something you’re going to be able to take advantage of. So we would look forward to you joining.
[01:00:09.500] – Stuart Bruce
I think all that remains is for us both to say thank you for participating and please stay in touch.
[01:00:17.090] – Aaron Kwitten
Yes, thank you, everybody. And thank you to my incredible co host. This has been great.