Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
Mind the Model: 013 Millie Marconi
36 Plays13 days ago
Welcome to Episode 013 of Mind the Model: The Modern Marketer's Guide to AI! Your hosts, Emmalee Crellin and Nathan Guerra, sit down with Millie Marconi — a seasoned entrepreneur who spent nine years growing a marketing agency before founding TestFeed, a VC-backed AI platform that's reimagining how businesses access market research.
Millie breaks down the world of "synthetic audiences" — AI-generated panels built from real human data that make consumer insights accessible to businesses that could never afford traditional research. She walks us through how it works, where it gets it right (and wrong), and why she believes over 70% of market research will be synthetic by 2028.
This episode covers:
- Democratising Market Research: How TestFeed delivers on-demand synthetic audiences — putting consumer insights within reach for companies of all sizes, not just the big players.
- Trust & Reliability: How TestFeed uses peer-reviewed "semantic similarity rating" to validate AI panels against real human panels — and course-correct when they diverge.
- Marketing to AI Agents: Millie's vision for a future where AI agents make purchasing decisions — and why marketers will need to learn how to market directly to machines.
📚 Resources & Links
TestFeed: https://testfeed.ai/
Millie on LinkedIn: https://www.linkedin.com/in/millie-maronei-11b22b7a/
Prompt Cowboy: https://www.promptcowboy.com/
🎧 If you liked this episode, follow, rate, and share Mind the Model with your fellow AI-curious friends — it really helps us grow!
💌 Got thoughts or ideas for future episodes? Drop us a message at mindthemodelpod@gmail.com.
Recommended
Transcript
Introduction and Podcast Production Challenges
00:00:10
Speaker
Welcome to Mind the Model. I'm Nathan Guerra. And as always, I'm joined by my co-host Emily Crellin. Em, how's it going? Howdy. It's going well here. Nate, how are you? You know, I'm i'm enjoying the kind of cooler weather as it comes in.
00:00:23
Speaker
Yes, autumn has settled into Sydney. Now, I wanted to give a bit of a treat to our listeners this week. and Like yeah candy or? Sneaky snack. No, a bit of a behind the scenes of how we actually produce the podcast. Because...
00:00:39
Speaker
It is a passion project for us. We have full-time day jobs and parenting um on top of everything else. Oftentimes we get asked the question, how do we have time for this podcast? And just wanted to take you through how I actually formulated the questions that we're going to be asking our guests today. So when I had my first interview, quick 30-minute chat with our guest, Millie, she was so kind to record the chat and her ai you know. ah Friend? for Her AI friend joined the chat as well. And Motion provided really great rubric and outline of what we covered. And there was a certain part that like we were on fire going back and forth during our like pre-interview chat. And it's like, that's exactly what I want to talk about.
00:01:25
Speaker
So I took that little snippet of notes, I put it into Claude, and I said, this plus her website for a test feed gives me an overview of questions I should ask.
00:01:38
Speaker
And it matched all of the framework and outline of the questions that we have asked previous guests, and it boom spat out a template that wasn't 100%, so I wasn't gonna copy paste it over, but it was schmicko, and then immediately put in the outline.
00:01:54
Speaker
And can I just say from that point of doing it then to where we started like massive time saving. Is research irrelevant now? I would say that not necessarily research is irrelevant, but the quick formatting, the understanding of our tone and context of how we shape these podcasts.
00:02:13
Speaker
Yeah, it just, it seemed like such like such a fast win. And I did this all because we're such great planners. I did this at 9 p.m. last night. and So we're we're kicking goals over here.
00:02:26
Speaker
All right, so now the only problem is if this podcast isn't breaking ground, we're going to be like pillared for the fact that we've used AI and not done it ourselves. I think so. And, you know, we've been chatting offline about different aspects of the podcast of how we can identify their repeatable tasks and, you know, quite a time suck. For instance, show notes, we always like to reference during our chat, oh, check it out in the show notes. And we want to make sure we put as much information in there as possible.
Guest Introduction: Millie Marconi
00:02:52
Speaker
It's been fantastic having Gemini Pro kind of audit the Google Drive export of the ah video podcast, giving us a brief summary and then matching that back to a template for our show notes.
00:03:05
Speaker
So, you know, we talk all the time about AI and now we're starting to really utilize it more in our workflows. So i just wanted to, you know, lift the curtain and share some wins that we've been experiencing.
00:03:16
Speaker
You know, it's good that we walk the walk as well as talk the talk and and demonstrate that for people as well. Because, you know, what we want always want to be doing is showing people practical ways how we're using it, as well as how inspiring guests like our guest today is using AI. all um I'll go ahead and introduce our guest today. We have Millie Marconi. Millie is a seasoned entrepreneur who previously spent nine years growing a marketing agency and building technology. Her newest VC-backed venture test feed is an AI platform that generates AI synthetic audiences from real human panel data. It's driven by her own struggles with unsuccessful products and the need for accessible market research. She's now scaling test feed to offer on-demand market research and has built 30,000 plus social following by consistently building in public.
00:04:03
Speaker
Welcome to Mind the Model,
AI Models and Tools in Entrepreneurship
00:04:04
Speaker
Millie. Thank you so much for having me. I'm stoked to be here. And just to start this off, have a couple of quickfire questions to ask you. What's your favorite AI model and what was the last thing that you used it for? So I love Claude Opus 4.6. It's one of the most advanced AI models. It has a 200,000 context window, which they have 1 million in beta. And I love it for actually getting enterprise-ready documents. So basically I could compiled my entire pitch deck. I then compiled a proposal for each potential customer that we speak with so I can send through a branded proposal and deck for each specific client. And it's just phenomenal. I can download it ready to go.
00:04:48
Speaker
And so it's doing the decks for you? Absolutely. And it's phenomenal at also following a brand guide as well. Definitely takes quite a bit of prompting, but once it gets there, it's yeah saved hours and hours and hours. And are you using an agent or are you using just constant prompting? um So i actually am a big advocate of Prompt Cowboy.
00:05:05
Speaker
I use Prompt Cowboy every day. i think it's a fantastic tool. And I actually previously used to use our own product. We had a B2C product and i actually created a persona of a prompting agent. And so I actually would just all day ask our own tool, but now I use um Prompt Cowboy. So speaking of ah prompting, do you have a a particular methodology that you follow every time or like an approach that you use?
00:05:29
Speaker
Yeah, so now I'm much more led by i have all of my different projects within Anthropic. um And so all of them are set up with specific project prompts, which again, I compiled through Prompt Cowboy and through a number of different, I guess, methodologies and um information that I found on the internet around how to get the best output. But yeah, i do think for most people, speed of execution, Prompt Cowboy is a great example. And I do see a future whereby our current B2C product, so our our test speed, it was a Chrome extension for the purpose of populating personas quickly, which we've actually pulled offline. But I can actually see us reskinning that as purely a prompting tool.
Millie's Entrepreneurial Journey
00:06:16
Speaker
Can you give us a bit of a 60 second rundown?
00:06:20
Speaker
know Who are you? What's Test Feed? So I am Millie Marconi. I ran a marketing agency for around six years. I started in entrepreneurship straight out of university when I was living in London. My partner at the time was in marketing and he was paying his interns £11,000. He didn't set the price for the record. um And so we basically looked at post my graduation, I was finishing my degree in London. um That would be what I would expect to earn. so from there, we went straight into our own business. we went to an e-commerce store. Then from there, we evolved into a marketing agency, product studio. In 2023, I raised venture capital. And here I am today.
00:07:00
Speaker
One thing that I wanted to point out is that I did a bit of digging on you and I saw that you actually started a company called Yesterday in HR Tech. It got BC back in from Antler and then you just pivoted entirely. What happened? I jumped into HR Tech after working doing headhunting on the side. So I did headhunting with a colleague for a number of years. It was a phenomenal experience. I got exposure to Australia's best and brightest were hiring C-suite for digital companies.
00:07:26
Speaker
And I saw a lot of inefficiency in that process. So I was a marketer. My background is marketing and my colleague, her background was also marketing. So we approached the problem quite differently. um I then really, i guess, became fixated on the problem and fixated on wanting to provide the technology solution to the inefficient process. So in 2023, I jumped into the Antler Accelerator,
00:07:51
Speaker
um sort of with no expectations of actually raising the capital, but walked out 12 weeks later with a co-founder and the capital. So we were focusing on screening, so candidate screening using AI.
00:08:04
Speaker
And so we built a product. It was enterprise-style product. We had ah the product in market for over 12 months, generated over one hundred k in revenue. And then we pivoted and the pivot was very organic. um It was actually driven by an event that happened that sort of showed us this incredible demand in this new space.
00:08:27
Speaker
um And so it was something
Content Simulation and Synthetic Audiences
00:08:28
Speaker
that we sort of just jumped on and and ran with and haven't stopped ever since. I'm curious, what was the big event? As a founder these days, you really have to be a content creator. It's almost an expectation. It's distribution. It's It makes absolute sense. um From an investment standpoint, it can be a competitive advantage.
00:08:46
Speaker
And I was really trying to become a thought leader in hiring, which was quite stupid because I had no actual formal background in HR. um So I was doing these very, very...
00:08:58
Speaker
cringy LinkedIn posts that were all AI generated and they were very stuffy and corporate. And I was trying to, I guess, make up for my lack of experience. um And one of them actually went viral and it went viral for the wrong reasons. And it really, I guess... Was it a wake-up call? Yeah, it definitely was. It was a post that essentially what I was trying to say and how it was articulated, two different things, and also how it was perceived was very different. So I posted it, it went...
00:09:28
Speaker
absolutely gangbusters. It was gaining so much momentum so quickly. It was actually only up for about an hour, but in that time, and it had done some pretty significant damage. um So that was this, I guess, pivotal moment whereby I really, really lost my confidence and I was so gun-shy posting on LinkedIn.
00:09:46
Speaker
And my partner at the time, who was an AI engineer, actually built me this content simulation engine So he put together this prototype quite quickly that basically allowed me to simulate my posts in a sort of sandpit environment before they went out. So we had personas and we had populated the audience that was representative of who we thought my audience was. And the litmus test for for the tool was actually to rerun that exact same post and to see how accurate the sentiment was. And it was spot on. So it sort of showed me, okay, now I have this sort of secret weapon to know the reception before I post.
00:10:26
Speaker
I actually shared it with a couple of startup friends and it got to the point where they would be sending me their LinkedIn posts every single day, asking me to simulate it and then asking for me to send back the revised version.
00:10:38
Speaker
And so then from there we built in sort of engagement metrics and understanding and trying to simulate how the algorithm would predict the performance of the post. So then I filmed a demo and posted it on LinkedIn and it went crazy.
00:10:51
Speaker
And so it really showed us, wow, there is serious demand for this sort of simulation experience. So then we basically jumped on it overnight and just ran with it and it hasn't stopped ever since. The evolution of the product has gone into sort of a different direction. We started with content simulation, which is high frequency, low value really.
00:11:14
Speaker
So now we're in a much more high frequency, high value um area whereby we're working in synthetic audiences. So the application of to assist with market research.
00:11:26
Speaker
And how does a synthetic audience differ from, I guess, a traditional focus group or a panel? So synthetic audiences are essentially the use of synthetic data. So it's not necessarily a live focus group whereby you have people sitting in a room, you're asking them questions, you're collating the data.
00:11:43
Speaker
you basically are using AI to replicate or reproduce these studies. So it's a phenomenal use case of AI whereby it allows us to really democratize market research, which is previously only available to businesses with very, very deep pockets. And I know from my own experience, having built over nine years, i built more than enough stuff that no one actually wanted because I couldn't get that customer insight. We do include humans in the loop in our process. um We run a side-by-side human panel for our early pilots, which we do that for a number of reasons.
00:12:17
Speaker
um But essentially, it's this very exciting evolution in market research that researchers predict, over 50% of market researchers predict that it will be over 70% of market research by 2028. So over will
Reliability and Growth of Synthetic Markets
00:12:31
Speaker
be synthetic. One thing that you mentioned that I really wanted to highlight here is the democratization of market research. And that's what synthetic audiences can bring to the industry. What exactly do you mean by that?
00:12:43
Speaker
Yeah, so traditional market research, it's one and done. So you do one conversation or one study, you get those results, you work with those results, there you go. The reason why synthetic audiences are really exciting is because it allows us this always-on element. So our focus is very on allowing rapid and iterative results concept testing, surveys, A-B b testing. And so the reason why that has the huge power to really democratise market research is for the first time it's much more feasible, it's much more um resource ah friendly
00:13:25
Speaker
And we have this ability whereby collect all of this data so then we can just continue continually test and iterate with the existing synthetic population. So we build a synthetic population, our clients then have that It's a self-serve product and they can just test as rapidly as they'd like.
00:13:44
Speaker
How does a marketer know that they can trust a synthetic audience to give them good results? Yeah. And that's the most common question I get almost every day. And we've spent over five months in research and development to make sure that we have this sort of reliability. So what we have done is we use a peer reviewed methodology. So it's called a semantic similarity rating and essentially we can contrast both the human panel and then the AI panel. So we can actually see the exact divergence between the two and then we can course correct accordingly. So it does provide this layer of credibility and this layer of trust that means that, okay, we actually have confidence now that this is reflective of real world outcomes. you You noted earlier that the future is leaning towards more synthetic audiences, but do you think there are
00:14:34
Speaker
is a use case for when a marketer should still go and talk to humans. For us, we are offering our early pilot. So we're running the human panel at the start. We show the reliability with this rating system, and then we offer the ability to repanel as well. So that's up to the organization how frequently they want to repanel. But that's a deliberate decision to ensure that we are keeping humans in the loop. We're never diverting too much away from the actual human in the process. And for us, I think it's the right decision because also from an adoption standpoint, we always need to sort of piggyback off existing behaviour.
00:15:12
Speaker
If you just come in and completely change the process, it's really hard to, i guess, get that initial adoption when someone's so familiar with their existing process. So that's also a deliberate decision on our behalf.
00:15:24
Speaker
Practically speaking, i'm a marketer, I'm using TestFeed. How does it work? So essentially it's the same as if you were to run any market research study. So you'd have an understanding of who you specifically want to talk to. So we capture those research requirements. We would either run a human panel or we would ingest data that you have. So if you've run a panel in the past, if you have really, really rich data on your customers, we're also looking at integration pathways at the moment as well. So how we can basically utilize all of this rich data that you already have, perhaps in your CRM.
00:15:58
Speaker
And then from there, we build this synthetic population. So that's basically an overview of the cohort that you want to speak to. And then from there, you can design the study. So we have a research assistant, which is an AI assistant, which helps you populate the questions, and ensure that you're going to get the right outcome or information. you're actually articulating the question correctly.
00:16:20
Speaker
And then from there, you can run the study. So it's similar to how you'd run any other survey, populate the questions and send it to the participants. And so then the AI basically performs the study and then you get the results back accordingly. And then you can see individual respondents as well. So it's very similar to the existing process. It's just obviously it's using ai Okay, so I'm a user. i go, I'm signing up on a test feed and I go, okay, here's a bunch of data.
00:16:50
Speaker
Here are the audiences I think I want to test. That process takes how long? Is that like a back and forth with human beings? Or i mean, could I do it today with, you know, could Emily and I do it right now with, ah for our art Mind the Model podcast? We will be self-serve very shortly. So at the moment, it's there's some services component to the process just because we are early working with our pilots and refining the interface and the user experience, but it will very shortly be completely self-serve. So you can basically upload all of the information you have on who the specific audience is, and then from there, populate the study, run it. It'll show all of the results for you, compile them, and then you can rerun as well.
00:17:34
Speaker
I could run... like one question at a time, like I could just go, hey, quickly, I want to know what people in my audience are going to think of this post, or I could ask them like, hey, my boss um is wondering, you know, like, do people like prefer blue or red?
AI's Role and Challenges in Marketing
00:17:52
Speaker
Yeah, so we have the concept testing, we have the A-B testing. um So you can actually show the audiences individual designs. We want it to be very representative of having these individual personas in the room with you.
00:18:07
Speaker
So we want it to feel like they are actually with you, watching your screen, understanding what you're working on and actually giving you that real time feedback. So that was actually the initial um product that we built was very B2C and it was very in real time screen share could actually interact with you as you work. um So now where we've headed, it's it's more reflective of sort of that traditional, I guess, research studies. Getting up with 10 questions and then bang, you get 10 answers back or, you know, roughly. Yeah. But you could work it either way, I guess is my question. Absolutely. Absolutely. And I really, the goal for me is to create this very iterative, always on interface. So we're not there yet.
00:18:50
Speaker
But that's very much where we're heading and particularly with sort of a mid-market slash SME product. What's your take on, you know, synthetic audiences innately not being human?
00:19:02
Speaker
Is that value diluted by the fact that it is more AI driven? Or do you think that it's just as valuable as full-blown human controlled study groups? The value proposition of a synthetic audience is that it's much more feasible. It's on demand.
00:19:19
Speaker
And when we compare the traditional process, of course that needs to be, people need to be paid for their time and it it costs what it costs. So I think the value of a synthetic audience is that it is on demand and that it is feasible.
00:19:34
Speaker
So i don't I don't see it as sort of in conflict per se. And I also think there's an interesting evolution in the decision maker won't actually necessarily be a human in the future, will actually be an agent.
00:19:48
Speaker
who's shopping on your behalf, who, so we're actually, we'll be marketing to AI in the future as well. which Is the AI generating the research to market towards the AI agents that are handling the purchasing on behalf of humans? Like what, what. There's a lot of loops here. Yeah. Like it just, ae AI speaking to AI, like at what point ah humans don't really seem to be in the loop there. It's so insane, isn't it? It's very meta. Um, Yeah, I think that's going to be a really interesting shift and I think we're already starting to see that again even from email reception. So previously we actually did a version of our content simulation for email and so we actually ran a um Monte Carlo simulation, showed all of the different ways that people would receive and what their engagement was of the email. What we realized is the interesting thing is inboxes now are managed by an agent. I know mine is, mine is managed by Motion. Great recap notes, by the way, Moshin. Thank you for that. Yeah, it's this really interesting shift in how we market. And to be honest, I don't have the answers to that. And I don't necessarily have a clear um roadmap of how that will shape our product as we are selling to agents. But it's definitely something that we're thinking about all the time. What are the limits
00:21:01
Speaker
of synthetic audiences right now. For us, something that we're really focusing on is the temporal awareness to how these personas actually have context of what's happening in the world and how that actually impacts their decisions. So again, we've spent a lot of time in in research and development around that. So are you saying that like a synthetic audience that was trained a year ago today won't be able to answer questions with today's much for answer questions of today? It might struggle with the war right now on that's the exact example going to say. Do you find that, I mean, obviously there's bias that's baked within the training of LLMs. Do you find synthetic audiences have any sort of skewer bias? So that's something, again, we're very, very mindful of and we're very um rapidly testing and looking out for. Humans carry a lot of bias also, and it's definitely a flawed argument, but it's something that, again, we are very, very mindful of and very yes actively testing.
00:21:59
Speaker
Correct me if I'm wrong, but it's almost like what you're building is bias bots. I mean, what you're trying to say is, hey, give us a bunch of data and then we're going to create this persona, this person, this this artificial person that has a point of view. And that point of view is a series of biases because we all, to your point, are kind of, we're human beings with biases.
00:22:21
Speaker
Not saying you should change the name of the company to bias bots, but I mean like that might be- And it's not a negative thing, but like exactly to Nate's point, it is bots that have- Point of view. The unique personalities and point of view and inherent biases, I suppose.
00:22:34
Speaker
Yeah, absolutely. The goal for us is to reflect real human behavior. So for us to get the correlation between the human panel, and the AI panel, that's the goal. And so with that, humans have their own bias. So we're wanting to accurately reflect what a real live human panel looks like.
00:22:53
Speaker
I can't help but think about the applications for focus group testing for politics. That's exactly where I was going. Yeah, yeah. Politics, politics, politics. politics and polls and... It's a really interesting application. It's something I'm very, very excited by because obviously the frequency as well of those sort of applications is where a tool like this provides a lot of availabl value. And also, so say in that example, it's also concept testing. So it's understanding, okay, before we go and...
00:23:25
Speaker
spend X, Y, Z on a specific avenue, we can actually just test the general concept before we actually go and produce anything. So it's also this sort of directional um advice and and and feedback, which as a entrepreneur is incredibly incredibly valuable when you are really just iterating and it's just providing this ability for you to just have some sort of proof of concept before we go and invest time
The Future of Test Feed and AI in Entrepreneurship
00:23:55
Speaker
and energy. Have you been approached and have you done any political panels yet?
00:23:59
Speaker
No, i have taught some outreach to a few individuals, but if you're in that space, I'm very, very, very interested. Just thinking about the future, what what do you think is next for TestFeed?
00:24:11
Speaker
I really feel passionate about synthetic audiences to allow entrepreneurs to rapidly iterate and test. So that for me is where I really want to head. I want to provide this always on application.
00:24:25
Speaker
this agentic functionality. So it's basically on demand. The personas are interacting in real time. um That's really what I foresee as the future and something that I think would I would be really, really proud of.
00:24:38
Speaker
One thing that I wanted to call back on of your past experience, and you've been quite open with the fact that you are a non-technical founder, but in a very deeply technical space, what does that teach you about how marketers as a whole who are mostly non-technical should approach AI. I really think that we have this amazing application and this ability to really upskill on the technical side.
00:25:05
Speaker
I have an amazing technical team, there's no doubt about it. But I think from a marketer's standpoint, we have this opportunity to really not be sort of blocked by challenges that would have previously meant that we had to wait for you know a colleague or... So just lifting the floor. And I think for marketers as well,
00:25:28
Speaker
We just need to be playing. Like I'm a big advocate of just trial and error, play. actually had a friend recently say to me, like, why don't we do once a quarter? We just sit down and we just literally do, we just play with random stuff and then we come and we show each other at the end of the week what we've found, what new processes and just this continual exploration, I think as a marketer, because there's no doubt of it.
00:25:52
Speaker
a lot of marketing roles will be disrupted. We need to be sort of very, very proactive in upskilling and playing and learning and trying new tools. So, Millie, we'd like to do a couple of closing questions each time for our guests. um Rapid fire. Do you consider yourself to be an ah ai optimist, realist or pessimist?
00:26:12
Speaker
Optimist. I feel so excited. Obviously, there's definitely concerning elements to AI, but I think particularly being a female founder and how much work we need to do to bridge the gap between male and female entrepreneurs and funding and what have you, um I think that we have this real ability to overcome those technical challenges. And I think it's fantastic. I think it's a brilliant time to be an entrepreneur and a brilliant time to be a
Reflections on Synthetic Audiences
00:26:38
Speaker
marketer.
00:26:38
Speaker
ah If you can fill in the blank for this sentence, in three years' time, synthetic audiences will be blank for marketing. Mainstream. Well, Millie, thanks for joining us. It was really great to meet you. And I definitely learned a lot about synthetic audiences. And if you need any, you know, beta testing, but we're always happy to throw the mind, the model stuff in there, see what people think. Love it. Play around. Absolutely. I'd love that.
00:27:02
Speaker
Thanks, Millie. Thank you so much.
00:27:09
Speaker
Em, what'd you think? That was such a fascinating conversation about a topic I knew nothing about. Had never played with synthetic audiences, never understood them really. So it was great to a founder in this space speaking to us about it. You do any synthetic audience stuff with your day job?
00:27:26
Speaker
Nope. I just do pure first party data. See, that but that's one of the interesting things. Okay, total tangent to what we just heard, but like synthetic data is a really useful way that people are taking and able to run simulations without having to worry about PII.
00:27:43
Speaker
So I would have thought that that would be, you know, you take some learnings and then synthetic data audiences. A hundred percent. I'm all for that future. ah It's just not currently. Eventually. i'm eventually It's not currently what i'm working with. And it's interesting because the way that she was speaking about The future, like the fact that synthetic audiences are more like cheaper to unlock because they don't require humans to be yeah fed. There's no actual human being there. Tea and cake. um it Just the the cost barrier is gone.
00:28:14
Speaker
And then the scenario planning is just so extensive. You'd have to be a very mindful marketer then to not just go down the rabbit hole and get sucked into the void. I took ah Mark Ritson's at mini MBA and he really values the...
00:28:35
Speaker
um the the focus group, but it's not like a day-to-day thing. It's ah an occasional thing. And now part of it is, again, the cost. So I'd be curious to get, you know, we should get Mark Ritson on and get his perspective on it.
00:28:51
Speaker
During that conversation, though, I just couldn't help but think, ah, like great application for marketers, but also politics. Yes. and You and I agree. Same thing.
00:29:02
Speaker
Yeah, you and i recently attended the Pod Save America live recording in Sydney. Big fan of that podcast and those guys. And just, you know, with the upcoming general election in the US only a few years away, would be very interesting to see if synthetic audiences make their way into focus group and poll testing.
00:29:22
Speaker
Yeah, I was, um yeah, completely agree. I was like, I bet the Democratic National Committee would be interested in something like this. or Australian local politics as well. Or, yeah. Viral citizens. Yeah, exactly. um Or even, um you know, I'm a big NPR listener and um they do focus groups consistently throughout the year. Very interesting to run and see, you know, if they ran, for example, from a news at gathering perspective,
00:29:48
Speaker
their traditional panel plus some synthetic panels and see how they compare it. Like that's the sort of thing where I think people are going to be very comfortable over time as they run two different types of panels together and start to see how effective these things can be.
00:30:03
Speaker
I know as, um, internally at my company and other agencies, they have internal panels, but we're often hamstrung by the amount of actual people that can get in the group. And so to scale that out, you could theoretically bring in synthetic panels, synthetic audiences. And yeah, I personally, I like the reframe of the bias bots.
00:30:25
Speaker
it's um it doesn't It rolls off the tongue, but it doesn't exactly bring the most- Confidence? Yeah, confidence. um But it's, you know, humans are full of bias, but that's also what makes us human. human
Conclusion and Listener Engagement
00:30:38
Speaker
Yeah. definite like would Does that make AI more human then?
00:30:41
Speaker
Oh, dear. It's too early for you to be breaking my brain, Nate. All right. Well, Em, want to wrap us up, I think? We have had such incredible feedback on Mind the Model that actually 95% of our listeners go on to recommend it to another friend. So thank you for this wonderful community. Continue to like, share, and subscribe. And any questions you may have, throw them to mindthemodelpod at gmail.com.
00:31:08
Speaker
And remember, the intelligence might be artificial, but the wins are real. See you next time. Bye.
00:31:17
Speaker
This episode was brought to you by Testchimp. Testchimp is a web developer's best friend. Testchimp uses deep runtime context to automate your testing workflows, generate scenarios, write scripts, and fix bugs all in one place.
00:31:33
Speaker
Stop wasting time and start shipping quality code faster. Bugs suck, but finding bugs doesn't have to. Let's test you.