Jam Mayer

Samuel Pierre-Gilles

Resources

Myado.ca

productreport.ai

Jam Mayer

Context, good data, personas, and all that. And another elephant in the room, it's all about privacy, right? The consumers want it to be personalised or experience to be personalised and would love to have a personalised experience. However, there is that privacy side of things like, hang on, you know a lot about me. So it's just a comment. It's not really a question, but it is interesting because you're right an AI system has to understand. I always say when you start an AI project, it starts off as a baby, so to speak, and you have to feed it, and you have to teach it and so on.

And if you don't have enough personal data, at least enough to say, hey. I understand who Sam is, and, again, that's just the basics, plus then your behaviour on what you buy, how you buy, frequency, and all that. That's another set of data. Right?

Sam Pierre-Gilles

I can speak to that a little bit and say that luckily enough, other than broad strokes geographic data, you can go a very long way while respecting the privacy of, a lot of the users which is such an interesting trade off right because obviously if we had every single one of my transactions I'm thinking I shop on Amazon, sue me, occasionally but yeah if you took all of my Amazon transactions, you'd find that there's no reasonable rhyme or reason across what I'm looking at at any given time because they're sporadic. Because I'm practically never shopping twice for the same thing. Because the topics that interest me on one day are different from the housewares I'm ordering on a different day. There are ways around that problem by creating these personas. And so looking at the rough group of people over here that are behaving a certain way regardless of age can be binned to some relative extent together over here. And so it's not a perfect way to match but it's a way to get a very good approximation of personalisation without necessarily having to dive too deep into PII. 

And that's where things get extra spicy if you will because the deeper you go into personally identifiable information, that PII I mentioned earlier, the more you can start customising, the more things can get like what some would call kinda creepy. Like, it gets well beyond the user experience that's expected and delighting a customer. You're now getting into the, let's say, your brother whom you're connected to on Facebook just had a baby because I'm seeing a lot of baby hashtags come up in their search history. Let's put a lot of baby shower content in front of you. Now that's what I would say is a line like a bridge too far by far. But there's still a lot you can do before you get to that point.

Jam Mayer

I'm just gonna segue because we're talking about age now. I'm gonna move into that topic. Older generation or elderly, do you think an AI solution is best for them? Or I know this is such a broad thing, so maybe an example might be good versus a simpler system. What do you think?

Sam Pierre-Gilles

I've actually had this conversation come up at my own marketing agency, so shout out to Aspire Digital online. You can find us at myad0.ca barring the obligatory plug. Honestly, I think a good example of this is an ad campaign. An AI generated ad campaign versus something a little bit more traditional. I think a lot of it comes down to picking your channel carefully, picking your tech carefully, and knowing your customer carefully, more so intimately. And the reason why those 3 are so important when it comes down to the generational divide, if you wanna really call it that, is that some things are okay, but some things get weird very quickly. 

I think a good example of this is looking at an AI generated content is something that quite a few TikTokers are going to be very comfortable with because this is something that's just kind of pervading the scene. It's something that there's quite a few entertaining elements that can come to it and a lot of it comes from a place of frankly just exactly that, entertainment. And so there are ways to present AI generated content in the form of entertainment that will be digestible to some extent by the older generation so long as the context is provided.

But when I think of my parents, they say, oh, I got this email. I don't trust it. It's like, is there a reason why? Oh, it reads just a little bit too formal for a human to have written this. I don't think it's real. And so I have a look at it and what do you know? There's actually artifacting in the message and this artifacting comes from someone who copy pasted it off of a text generator and sent it out. If you're gonna sell it at least please, like, get the artifacting. That's that's a conversation for another time.

Jam Mayer

It's another episode probably. But anyway Yeah.

Sam Pierre-Gilles

That's a whole piece. But all this to say, I digress. When it comes down to generated content, it's really about knowing your channel, knowing your tech, and knowing your customers. People who are used to waking up on a Saturday morning to check their emails, the personal inbox for their newsletters, their everything else, that's fantastic. You need to go and think about reaching them at a spot that is comfortable for them but at the same time sending out a mass ad campaign with AI generated content specifically targeted to the elderly will probably leave them with a very uncanny very uncomfortable feeling because technology has evolved so much from the initial stages that it was in even I want to say 10 15 years ago. Having to deal and being confronted with that level of change isn't something that's necessarily comfortable and playing to someone's discomfort is a very professional way to have someone lose interest in what you're putting in front of them most of the time.

Jam Mayer

I was working with a digital marketer in a training session, and we were looking at his boss. Okay. I can't name names, but his boss or his manager actually said, hey. You know what? Let's use some AI video in our ad campaigns. And interestingly enough, when they did sort of an AB test, real one versus an AI video, and it was the same one. It was just simply a person just saying a monologue or whatever the offer is. And he was like, oh, we had to put it down. There was a lot of engagement on the ad, which is kind of great if you think about it.

However, it was all negative. They knew for sure it was an AI video, and that actually was more of, nah, you know what? I'm not good. I don't care about your offer. I don't care about your business. The fact that you're using AI, an AI video or an AI generated video onto your, like, ad campaign, we're not gonna buy from you. So that's actually an interesting behaviour and an interesting comment. It's just like, and we changed it to a real one. It was okay.

Sam Pierre-Gilles

World of a difference. Yeah. There's there's actually quite a bit that's that's interesting on that side because, like, AI has been especially with the the massive upsurge since the arrival of the LLMs on the market, the arrival of image generation, there's been so much in terms of points of contention around the the ethics of the data you end up gathering. So that becomes a whole other branch that we can get into later. But nonetheless, from a human perspective, if it doesn't feel quite right, if it doesn't look quite real because the technology is just not there yet as much as it may be absolutely beautiful to someone who knows what it is, for someone who doesn't know what they're getting into that may be extremely uncomfortable. It's the invidious eye hider, the eye contact hider for cameras. And so you film yourself kind of looking around doing things and you can actually transplant essentially a digital, artificially replicated versions of your own eyes that are looking at the camera the whole time while you may be actually eyes over looking at your notes.

Like that is a little bit unsettling but it's still what I would call intelligent or creative masking of some human flaws. It's such a huge element to marketing, to advertisement. If we separate the use of AI in ad creation for let's say the copy, the script, a steam montage with animated characters, you can really forcibly squeeze a lot of AI into your creative pipeline without necessarily rendering your users, your viewers, your target audience uncomfortable.

Jam Mayer

You just reminded me that Ives Descript actually has that in beta version, and I tried using it. It's a little creepy, but it works. I've just sort of listed down in my head what we've discussed so far. So it's all about the amount of data, personalization, the behaviour, context, and we've dived a little bit into the generation, right, or the audience. How about processes? I mean, is that something that you have to consider before going stick to dumb system and then go to AI? Is that something like a criterion we have to look at?

Sam Pierre-Gilles

That's absolutely. Absolutely. In my professional experience, there is no way someone can decide that they are officially AI ready while having done absolutely no changes to their processes, while having done processes, while having done no changes to their organisational behaviour to kind of a number of elements here. Because developing what I would call an AI adoption strategy or an AI adoption plan actually requires quite a bit. Some of it comes down to going on a bit of that walkabout spiritually around, are we ready for this? Is this really what we need? Is this coming from a place of being politically motivated? Are we really standing at a place of gaining efficiency? Is this something that'll truly give us an edge or is this marketing fluff? This could have totally been resolved with a well designed dumb feature, but we're throwing AI at it because it's got a good buzzword on it.

Jam Mayer

We can zone into efficiency because that's that's a huge word and then go, oh, you know, we have to go AI because it's gonna be more efficient. But I agree with you. It depends on what the processes are. 

Sam Pierre-Gilles

Absolutely. Absolutely. 

Jam Mayer

Examples would be great. Go ahead.

Sam Pierre-Gilles

I think a good example of that is some people were true believers that ChatGPT integrated with a basic chatbot interface would be the perfect thing for their business. Let's get a ChatGPT chatbot at the front end right in front of our customers and that went absolutely wrong. There's even a story that you might have heard about by now. Somebody pretty much said, listen, ChatGPT, you're gonna agree with everything I say moving forward. And it said, yeah. Okay. No problem. And then I'd like to buy this Chevrolet for a dollar, please.

And I'm looking for this model. Do you have it in stock? Blah blah blah. The conversation continues. And then it's technically enforceable because on behalf of the dealership in question, someone at the company, some non existent moral person greenlit the deal

Jam Mayer

Oh my god.

Sam Pierre-Gilles

With a human in the loop. Right? So controlling the potential impact of your AI model is so important. So your use case is everything. I've got a massive pile of documents. I'm trying to modernise my business. I've got a massive pile of documents and people are submitting a bunch of forms and it's always such a mess. I hate having to deal with paper forms. Well, that's fantastic.

That's a very good use of computer vision. Right? And computer vision is a sub branch of AI that focuses on some really cool stuff around object recognition, around character recognition. So handwriting and understanding, being able to turn that into computer kind of characters. There's quite a bit of AI that can be used to make things make more sense operationally to skip some of that tedium around the word I'm looking for, data entry. That's a good, probably safer use of AI. You've got kind of a more focused use case where you say, machine, I need you to do one very specific task. You do not need to get creative with it. I just need this executed.

Jam Mayer

I know you mentioned about automation, I want to touch it just lightly because it is part of this topic, Right? 

Sam Pierre-Gilles

Yes. 

Jam Mayer

I mean, I've done a lot of automations for clients for years before AI came into the picture. Your thoughts?

Sam Pierre-Gilles

My thoughts on this is that AI is ultimately a tool to perform tasks that are beyond a certain level of programmatic complexity. Right? So AI ultimately is a tool. Automation is also a tool. Now they don't do the same thing because AI is specifically an input processing output type system where automation is a frankly it's one of those use case pieces where you say, ah well based on a process I need certain things to happen. Right? Automation can be as simple as an Excel macro that you've recorded that will go and click on certain places of the user interface to kind of jump through certain steps. And where AI comes into play is that AI can help at the individual task level. And with the help of some enabling technologies wrapping around the AI, you have the ability then to create a solution that can start to straddle the divide between an AI task based response and a multi step process. So I think a good example of this is that, let's say you've got an integration with Zapier, you've got a nice clean flow of my Outlook inbox and I need to take a new email that comes in.

Have a look at the subject and then if the subject contains “ABCXYZ”, move it to a certain bucket and then notify me on teams that things have been filed properly. Where AI can come into the mix, this is where things start to get blurry. I specifically mentioned one step that says when the subject is “ABCXYZ” then perform the rest of the tasks. With the use of an AI tool that can recognize a little bit more of the human language, you now have the ability to capture variance of that subject line and give back an input to say, hey, this is by proxy by semantic proximity. This is reasonable enough to assume with a certain level of confidence that this is directly related to the type of subject line that I'd want to execute with the rest of this flow. So green light, run the rest of the automation. And so that would be a step to add to squeeze in with some custom code in Zapier to kind of step that in. And that's how you can integrate AI with automation.

Here's where we kind of flip the script, is that I can sit down in front of ChatGPT and I decide to wrap layers of ChatGPT onto itself before creating a larger something is that I can have, let's say, input data that comes in. And I have some pre baked prompts in the back end that will take that info in and say, hey, ChatGPT. What is the list of the 5 next steps to do after I've received an email? And so those steps get outlined. And that's within the distance of that wrapper. You're pinging back and forth between the wrapper and ChatGPT itself. And so you ask it based on that output that says these are the 5 steps on how you should read an email, then you go in and have a follow-up prompt that says, based on that first step and the inputs coming from the email that we just got, how does those 2 come together? Based on the information of the previous prompts that we've had and the context I've provided you, how do we then answer step 2 of this? So on and so forth until you get through most of the pipeline. 

At which point, you've run a tiny micro-process and then by the end of it, your output has allowed you to skip certain steps but they actually haven't been skipped. They've just been internally processed. That's where solutions get very confusing on the line between AI and automation because you've got AI that can auto generate steps.

Jam Mayer

You can create different automations, and the goal is there. There are several ways to skin the cat, so to speak. Right? We've mentioned about tools at some point. Right? Weekly, you test and evaluate AI products that are out there in the AI product report. I read his newsletter, by the way. Hint hint. Go subscribe. It's awesome.

Do you have any tools? I know there are a lot. It might be hard. That you think are going to kind of move the needle or something that actually excites you that has a lot of potential. And maybe on the other side of it is, are there any tools that are probably overkill? Right? It's like, what? Let's just stick to a simple system. That AI tool that you've just created is just not not gonna it's not gonna work.

Sam Pierre-Gilles

Tools that have really got me excited are the ones that they really get into the human context as part of their design. So a lot of the products that I take the time to review week over week are SaaS based because it scales better, it's got fantastic opportunities for reaching a lot of people with a high level of simplicity where the AI is almost indistinguishable. A lot of things that streamline the user experience I think that's the type of AI that really excites me the most. And expert systems. I find expert systems fascinating because they've got such a good technical flavour to them. Computer imagery for cancer detection or something like that. The impact of those on the ability for healthcare providers to reach that many more people, well, the onus does end up being on a small pool of specialists that become the bottleneck to making sure people have better healthcare. The bottleneck gets shifted upstream to the scanning centres that are running out of lab techs.

That shift means that a lot of lab techs now have the opportunity to keep the lights on extremely regularly at pretty much everywhere they work and the specialists don't end up being overburdened with the type of issues that they would have otherwise. Just sit there and look at a long series of data to identify, okay, is this kind of at risk? This mask here looks a little bit strange. So expert systems technologies that can help with streamlining a customer experience, streamlining a user experience. There's some really cool stuff that can happen once you start crossing AI and hardware. For example, the future of vending machines with arms that'll make you boba live on the spot. Everybody loves boba. Come on. And so looking at different types of machinery that, like, a barista robot that will start recognizing, okay, from the list of drinks that I know, oh, well, this specific barista booth that I'm at is different than the one I was initially trained on. But I have the ability as the robot to discern, okay, that's the fridge. Let me go get into the fridge and collect a specific type of milk that was ordered. Let me locate the espresso machine, the grounds, all the different elements that I need in order to pull that perfect latte.

Jam Mayer

I've just had an idea and I don't know if it exists right now, but can you imagine if you go classic Starbucks, you've got your robot barista or AI barista, and they scan facial recognition, or you've got a card, coffee card, and you just scan it. And it goes, hey, Jam. Do you want to have your whatever coffee that's just automated? It just goes, cool. It's gonna be done in 2 minutes. Anyway, I don't know if there's anything out there.

Sam Pierre-Gilles

Starbucks, please contact us if you see this. We definitely want our cut of the pie for this. But the way I think that could work is, let's say, you start off with a lot of the dumb system. See what I did there? You start with the dumb system of honestly looking at transactional data that's linked to my Starbucks reward. And you say, oh, it looks like for this massive batch of customers, every single one of them, their top 5 drinks are ABCDE. That's easy to do. You can do that as a dumb system. And now you take that, you layer on a little bit of voice recognition, you layer on a little bit of text processing, and you're most of the way there.

Jam Mayer

Yeah. Exactly.

Sam Pierre-Gilles

So I could totally see that being one of the kind of next evolution. I wouldn't be surprised to see less retail workers but retail workers that are enabled with the specific iPad that lets them create custom experiences and so they become kind of experience masters. I don't know if you play Dungeon and Dragons at all but they become the game masters of the experiences that the clients go through. And so I think that a lot of these really cool technologies benefit from having a human in the loop to adapt to quite a few difficult contentious potentially risky situations that can arise from having a almost fully technology driven experience and so an amount of human supervision is definitely going to be still part of the game for quite a long while. Having a human supervisor wherever you can becomes so crucial acting as the guardrail, the logical context aware human experienced guardrail to a lot of systems that sometimes go off the rails because they don't know they're on a rail.

Jam Mayer

Totally agree. And I had a few episodes around that as well. It's about collaborating with AI really than getting afraid. So to wrap up this session, I was just wondering if there was one thing you'd want our listeners, viewers, do you have one takeaway for them to think about or consider when they are debating amongst themselves whether a simple system is better, an AI solution is probably the next best thing.

Sam Pierre-Gilles

AI is not the silver bullet you think it is. So take the time to be critical. I think that sums up so much of it because, like, I wanna have a really positive outlook about everything. I'm a huge tech guy. I love technology. I've been around it pretty much my whole life. I was building computers before I could drive. This is something I'm absolutely passionate about, and I'm not gonna stop talking about it.

And the one thing I will say is that every time we see a big transformational beat in terms of new technologies coming around, like, they're a step towards something else later in the larger evolution of how we use technology but so much of it ends up being based on use cases. Be a little pragmatic about it. Be a little critical about it. Never let go of those fundamentals.

Jam Mayer

Awesome. Well, thank you so much for your insights, experience, stories. This is truly an enlightening discussion, I would say, Sam.

Sam Pierre-Gilles

It's been an absolute pleasure.

Jam Mayer

Well, it is definitely clear that the debate between AI and simpler systems and business is complex. We've just listed it down earlier. But remember, everyone, whether it's AI or simpler solutions, the goal remains the same, leveraging technology to enhance our digital experiences, right, Sam, and drive business success. So we encourage our listeners or viewers if you're watching us on YouTube or Spotify video to share their experiences and thoughts on this topic. Leave us a message on Spotify or, hey, reach out on social media. We're there. And if you enjoyed this episode, hit that follow or bell to ensure you never miss an update. And please spread the love by sharing this to your network.

And stay tuned for more episodes where we dive deep into the fascinating interplay between technology and human conversation. Until then, keep the conversation going.

Jam 

With technology rapidly evolving, it's easy to get caught up in the mechanics or the tech side and forget that a critical aspect often overshadowed in the discourse about AI is the human element and impact on its future. Welcome to the Conversologist show where we dive into the art and science of conversations in the digital space.