Ep 13: PersonaAI - How to Turn Pretty Persona PDFs into Synthetic Customers

Transcript

Transcript

In this episode of Splice of Life Science Marketing, Jasmine Gruia-Gray interviews Matt Wilkinson about PersonaAI - a method for converting static buyer persona documents into interactive AI-powered synthetic customers. The conversation covers the limitations of traditional persona approaches, the technical process behind building synthetic customers, and practical applications across marketing, sales, and product development.

Introduction to PersonaAI

Speaker: Jasmine Gruia-Gray [0:04]

Hello, Matt. Hi, Jasmine. How you doing today? Good. Thanks. How are you? I'm good. Thank you. I'm good. Well, welcome to A Splice of Life Science Marketing, everybody. In today's episode, we're going to do something slightly different. I'm going to interview Matt on a synthetic AI process. It happens to be a process that he's developed and that he calls PersonaAI. So ready, Matt?

Speaker: Matt Wilkinson [0:36]

Absolutely. Okay.

Why Matt Created Synthetic Customers

Speaker: Jasmine Gruia-Gray [0:40]

So you've built buyer personas for years using traditional methods - workshops, interviews, desk research. I think we've all been there. What made you decide to turn those static documents, some of which I have to admit are in my drawer, into interactive AI assistants? And when did you realise that this was something other companies needed?

Speaker: Matt Wilkinson [1:03]

It dawned on me pretty quickly after the launch of ChatGPT 3.5 and I think it was then, when four came out, custom GPTs became a thing. And one of my friends had essentially created a group of different developers to talk to, and then somebody else had created a different group of consultants, and I went, we can do this with personas. So while people had put in some very basic groups to kind of help guide them, I thought, well, why don't I use these detailed persona documents that we created?

And so I created a persona group, or a buying group, and very quickly found that with some careful prompting in the back of that, I could get them to act as synthetic customers, and that was great. And I could get them to very quickly test messaging and test strategy that I'd already created. And I could get them to argue amongst themselves, which is really fascinating to watch. The AI essentially arguing with itself, the different personalities within an AI persona group arguing amongst themselves.

And so that was kind of the first step. But the thing that I think really changed the way that I looked at this was when I realised I could use it to create content, and then when I realised that we could actually disrupt the process of generating the persona in the first place. And that really came from the launch of the reasoning models and the deep research capabilities. And when that happened, I started realising that we can replace a lot of that manual work with a lot of AI work. And that was really when the big shift happened.

What is a Synthetic Customer?

Speaker: Jasmine Gruia-Gray [2:37]

That's fantastic. So maybe we can take a pause here for a second and just describe what you mean by a synthetic customer.

Speaker: Matt Wilkinson [2:48]

Essentially, what we do is we put those flat, boring personas into the AI, and it's essentially creating a RAG or a retrieval augmented generative AI system. And with some clever prompting, we get that information about our customers to behave as if they are a synthetic version of our persona.

And with that, as long as we've generated information related to the questions we're asking them - what's important to them, what are their buying needs, what stands in their way of achieving a certain job, what jobs do they have to be done - all of those great questions that I know we talk about a lot. If we've covered all of those within our persona, we can then use that and interrogate the AI, and we can use that to help guide brainstorming sessions, come up with strategies, come up with messaging, come up with creating content.

And it breaks my heart when I know that personas I'd created for you in the past never got looked at again after being used in the project. So that process made me realise that we can turn these documents that hold loads of great information into something really, really useful.

The Problem with Traditional Persona Workshops

Speaker: Jasmine Gruia-Gray [3:59]

While I had admitted that my personas, whether you helped develop them, or I worked with another group, ended up in the drawer, PersonaAI has never ended up in the drawer for me. So yeah, I agree with all of what you've said. So moving on to my next question, most small life science companies can't afford six-figure persona research programmes that go into excruciating depth on the personalities and so on of personas, but they can't afford to guess at buyer motivations either. How does PersonaAI solve that tension for teams with limited budgets and even less time?

Speaker: Matt Wilkinson [4:45]

So if I go back to the standard process, and it maybe helps, we can then look at how I've tried to disrupt that process, to make great results more achievable. So the standard process very often involves getting a big group of people together on a workshop, whether that's in person or remote. And typically you'd be asking, you start off trying to identify an ideal customer profile, and then within those ICPs, who's involved in the buying group, what kind of job roles are involved in that buying group, and what role do they take in the buying process.

And so in the room, you typically want to have your sales and your marketing and your customer services people. And even if you take the external costs out of that equation, the costs for those people's time is immense. Now there's huge value still to be had for getting everybody involved on agreeing who are we selling to. And so that strategic alignment is absolutely vital.

But the next step always seemed incredibly vacuous to me. People would end up in a workshop trying to answer questions, and very often they'd end up looking at a LinkedIn profile of somebody that they sold to and basing their responses off kind of, oh, we know a little bit about this. Here's a LinkedIn profile. Let's just type some answers out into a question, and then they present their persona. And very often, after that process, you'd go away and take that and the organisation supporting that effort would go away and try and embellish it and make it nicer, and then you would create a pretty PDF and there's a static document that would then get used to help inform the rest of the project, whether that be branding, inbound marketing, product development, whatever it is.

For any of those projects, understanding who your customer is is really important. But that kind of falls down a little bit when people are having to make the psychological jump from going, this is what a document says, therefore this is what I need to do. And I think that's why they get forgotten.

So that's kind of the old process, and that's why I think those personas have limited value, not because of the information process, but because actually people still have to do too much thinking to make that jump. And AI is brilliant at helping make those correlations and those connections.

So the process that I've been through still involves some of that upfront agreement. It really does help to know who you're trying to sell to, who's involved in that. And I think that strategic alignment is really important. But then the process that I developed involved getting as many LinkedIn profiles of the type of person you're trying to sell to, whether they're customers, prospects, or just people that we think we want to sell to.

And then, using a long-winded prompt and 40-odd questions, I try and get the AI to answer, using one of the reasoning models, deep thinking, deep research, and going away and researching that. There's a lot of acronyms for what AI stands for, whether it's artificial intelligence or whether it's just average information. And I love the one that says it's average information, because here we want average information, but we want specific average information. We want average information about a subset of people.

So we get the AI to study a group of LinkedIn profiles and create an average of them. And what's great is that it doesn't just look at here's this career history. It looks at everything they've posted. But we can then take that, and I will get maybe a persona document with 10-point font, just to give a scale, that can be six pages long as an Excel sheet. So these things can be huge. You can have loads and loads of detail, and that detail will be absolutely useless for a human to use, because we can't cope with it. But the AI loves that context. The AI absolutely thrives with that context.

So we can do that, but we can go even further. If we have transcripts of conversations we've had with customers, we can add those in. We can embellish them further. So there's a process that you can do to really get a great starting point. And I think those personas work better than most I would have created previously. But then you can embellish them with real-life information, real-life thoughts, feelings, quotes, and really get a deep, detailed perspective of who your customers are. And then once you've understood who your customer is and generated one of these, all we have to do is put that into the AI itself, and all of a sudden we've got that synthetic customer that we can then interact with through the chat window. And that's really where the process kind of changes.

Pressure-Testing a Keynote with AI

Speaker: Jasmine Gruia-Gray [9:18]

So I'd like to pull on that thread of how we can interact with the AI. I love the story you had told me just before you were giving your keynote or preparing for your keynote in Washington, DC this summer, where you had used PersonaAI or Atlas to interview you to pressure-test your message. What did that conversation reveal that traditional speech prep wouldn't have?

Speaker: Matt Wilkinson [9:49]

I was able to use - I'd created personas for who I thought was going to be in the room and who my targets in the room might be. And so that was the first piece. Clearly, I'm trying to market to those people anyway. So that was helpful because that had a dual purpose there.

Then what I was able to do was to upload my slides, provide a transcript of my talk as well. I practised my talk to myself, and then I was able to ask the AI, based on who you are, what questions are you going to ask me? Where are the gaps in what I'm saying, where don't I make sense? So rather than this being a generic "these are the holes," it's actually "these are the holes that are going to be important to people I want to sell to, the people I want to influence, the people I want to connect with."

And so that's really where that difference happens. I got far harder questions out of the AI than I actually did in the room. Now, did it ask me every question from the people in the room? Absolutely not. But it had two roles. One, was it able to ask me some really interesting questions that I could then respond to, and I could answer those just by pressing the microphone on and actually speaking back to it, get the transcript and then put that in. So that was helpful.

Point number two, that then turned into a great blog post anyway. So that was helpful. That's a really fun way to create some content. But the third thing I think it really did was it highlighted that there were things that I needed to make sure that I said to bring people on the journey that I wanted to take them on. And it gave me the confidence that when people were asking questions that I probably already covered about 80% of them, which was true, and that meant that the questions you get asked are maybe on something a bit different, but it just made me feel a lot more prepared.

Speaker: Jasmine Gruia-Gray [11:30]

Yeah, for me, it's the same thing. It's about this feeling of being prepared. I mean, you're never going to know all the potential questions that somebody might ask you, but the AI is a pretty good approximation, as long as you've trained it accurately and got a good sense of who your target audience is, whether you're preparing for a speech or whether you're a salesperson and you're preparing for a visit to a new client, for example.

AI-Powered Sales Role Play

Speaker: Matt Wilkinson [12:05]

That brings me beautifully onto the talk I'm going to be giving next week in Boston at SAMPS. And the talk I'm giving actually shows - walks people through the process of how you can use AI both to research and then prepare for those conversations. So while I don't go into a lot of detail about creating personas as such, it does talk a lot about the personality side of it, and then uses tools where you can actually create those personas in conversational AI programmes, where you can create personas and then actually pitch to a buying group and get Q&As from a buying group, and where you can actually engage - they will then talk back to you, and you have a conversation with them, and they then grade you against key metrics.

Now, of course, for me, I'm practising, I'm creating the whole thing, so I'm saying I need to hit these milestones. So you might think that I would, if I've created this as a scenario, actually do pretty well, but actually, my first attempt at that, I only got scored 50%, and we know that if you're a salesperson and you're only hitting 50%, that's a fail.

And so it was really interesting to see that even though I'd created all this information and I'd been through a process to try to read it, actually what I really needed to do was to use the AI to help me learn and digest it in ways that I would absorb it better, so that I could then take that into another role play and be able to score higher.

Static PDFs vs Interactive Synthetic Customers

Speaker: Jasmine Gruia-Gray [13:35]

Yeah, there are so many places we can take this role-playing idea, and maybe that's an idea for another podcast. But another part of this whole concept with PersonaAI is that it is intended to be a synthetic customer, and not back to the old static PDF that gets shoved into a drawer. Walk us through what actually happens when someone uses PersonaAI versus pulling up a traditional persona. What's the difference in practice, not just in theory, and maybe we can use a marketing example this time instead of what we just talked about in sales.

Speaker: Matt Wilkinson [14:18]

Sure. So let's start off by imagining that we're looking at a web page. I've got a new product that I'm coming out with, and I need to make sure that it appeals to my persona. Well, I pull out my persona on one document, and I've got my web page on the other, and I've got me in the middle, and what I'm having to do is I'm having to try and read the persona document, put myself in the shoes of that person, and then look at that web page through that lens of that persona and see, how am I actually answering those questions? And I have to keep bringing myself out of myself into the shoes of the persona to do that.

Now, if I'm editing something on a web page, that's maybe not so bad, because it's a constant comparison. If I'm then creating something from scratch, let's just say I'm writing for a specific persona a blog post, or I'm writing that page in the first place, I'm constantly having to shift who I am, to try and be in the persona. And that's a really difficult leap for most people to do, to put themselves in the shoes of that person, and remove their own perspective, their own bias from that.

AI is really good at doing that. And so if you're using it with AI, rather than you trying to do all of that manual mental gymnastics - you know, who am I in? What role am I playing on this piece here? And am I looking at brand tone? Am I looking at what the persona wants? All of that stuff. Actually, using PersonaAI, and a more complex version where you go through a number of stage gates, what I've called Atlas on the website - what we're able to do is really use the AI as "I want to achieve this, this, this, and this." You give it some specific instructions as to what you want it to do, and it puts it through the filter - the persona, and then maybe the brand style guide, whatever else, maybe there's some rules we need to adhere to, and it will jump through those hoops for us.

So the outputs, they're not just generated by the AI, but what they are is actually generated by the AI for that persona. And so we're getting something that's really close to where we need to be. Is it going to be perfect? No, but it's going to be a heck of a lot quicker to go from A to B and then edit and then do that double-check than it is to go through and try and do that mental gymnastics for yourself.

Product Management and User Requirements

Speaker: Jasmine Gruia-Gray [16:33]

The other use case that I think is often missed is using personas and PersonaAI in the product management role, especially when you're trying to put together user requirements. You've got lots of voice-of-customer input that you then have to synthesise and describe as user requirements to hand into R&D in order to generate the functional requirements. How do you see PersonaAI helping in that way?

Speaker: Matt Wilkinson [17:07]

Well, I think it helps in a really similar way, in the fact that - and I think in some ways, with the voice of customer, and you're looking at user requirements, it's even more important. When we're collecting voice of customer, we'll hear quotes, we'll have certain bits of information that maybe resonate with us personally, but maybe become overweighted in the way that we look at things. So using the AI to synthesise all of that information, maybe thematically, and analyse it before we go into the persona - that's really important.

I've been doing some work with friends at the University of Leicester and some others who are actually going through and using that thematic analysis approach. It allows you to analyse data in ways that most organisations just don't have the time or the bandwidth for. To give you some perspective, what probably took about eight or more human days to complete in terms of thematic analysis by going painstakingly through and coding each of these statements, and making sure that you're putting those into the right models, you can accomplish that in about 15 minutes with AI.

Even more, what you can do is you can actually go away and ask what sort of questions people are asking. What are people saying online about these things as well? So you're not just getting voice of customer from the interviews and maybe your CRM and those sources that you've got internally, but you can also look at things online. What are people saying? What sentiment are they about these problems? So you can actually enrich your data from what people are talking about online as well.

So you can get a lot of that information, you can then build that into essentially your synthetic customer that you're going to then be able to interrogate through the life of the product. And of course, then if you're getting voice of customer and feedback at all, which I would strongly recommend - by the way, I don't suggest that we ever use these tools to replace human input, but rather to represent that human input and to keep us on track.

So if we've used it to get through the first phase gate of a development project, and then we go to the customer and say, here's what we've got, let's get some feedback on this. Once we then update our persona with that extra information, we're able to carry that forward, and it becomes like a living record of how we really resolve customer needs and really make sure that we're meeting their jobs to be done at every stage of our development process.

Speaker: Jasmine Gruia-Gray [19:34]

And then there's the opportunity to have a dialogue with the AI and ask it questions. Something simple, like, should I state the user requirement this way or this other way? Which is more accurate based on the voice of the customer? And then once the R&D team has had a chance to put the functional requirements together, you can input that data as well to augment the PersonaAI and ask, what's missing? It sounds like a simple question, but this is now your personal assistant and your helper to make sure that at this very important and expensive phase in product development that you're getting it right.

Speaker: Matt Wilkinson [20:22]

And I think the other beauty of this is that some people will want to see personas, they'll want to see requirements. They want to see different things out of that document. And the translation from one document to another may be imperfect, and then the interpretation of that may be imperfect. We don't have to do as much of that if we actually base it off of a synthetic knowledge base, a synthetic customer. Because even though we still need those documents signed off as part of a rigorous product development process, we're able to go back to that persona, that synthetic customer, and ask it questions. Am I interpreting this wrong? And that becomes really important and really powerful.

The PersonaAI Offer

Speaker: Jasmine Gruia-Gray [20:59]

So you're offering fully built PersonaAIs for $1,000 delivered in 48 hours. That's a radically fast and inexpensive offer compared to traditional approaches. Some people out there might be sceptical and say, what's the catch? And what does someone actually get at the end of these 48 hours that's really of value?

Speaker: Matt Wilkinson [21:31]

Well, the first thing you don't get is a pretty PDF. You do get a PDF, but that PDF is actually of the output - the data that goes in the back end. We want to make sure that that's validated. Typically, that might be somewhere between four and six pages long, so people get the chance to check that there's something real in the back end.

And then the other thing that they actually get is, depending on the systems that they use, we can either build it inside their Copilot environment, or custom GPT so they can access - and anybody with a paid ChatGPT account can access - or can even build these things as skills or projects in Claude. So we can use these in multiple places.

What they get is actually the real-life version of - well, a synthetic version of their customers that they can then engage with. That's what they get. If there are specifics that they want to get, the inputs are pretty simple. There's a series of questions that I would need to ask somebody. I would definitely need to make sure that I've got a series of LinkedIn profiles that they provide me. But if there's also VOC that they want to give me, other bits of information, they just have to provide that. And we use the AI to convert all of that into a really rich design.

Speaker: Jasmine Gruia-Gray [22:44]

And then I think the key is that you then need to use it for a myriad of different opportunities like we've already discussed. Help me review my current website. How well am I articulating the positioning, the value proposition? Are your needs as the buying persona being addressed on this web page? Again, lots and lots of questions you can keep asking it, and it will provide you with suggestions on how to improve.

Speaker: Matt Wilkinson [23:20]

Yeah, absolutely. I think that's the key - that we can then engage with them in such a nice way that it just makes life easier and quicker, and it makes these personas that so many of us go through those processes for - it just makes us able to engage with them in a way that's actually helpful rather than a hindrance. And I think one of the reasons why so many personas just end up rusting away on a hard drive somewhere is that it's hard work keeping them up to date, and it's actually hard work using them. So if we can make that easier, it just makes life easier for everybody. And it increases the value tenfold or more.

Overcoming the "Too Busy" Objection

Speaker: Jasmine Gruia-Gray [24:01]

So here's a scenario. A lean marketing team might think I barely have time to execute. Why would I spend time talking to an AI about my buyers? How do you respond when someone says they're too busy to use the tool that could save them time and, more importantly, make them more efficient?

Speaker: Matt Wilkinson [24:23]

Well, you can take a horse to water, but you can't make it drink. So that's the first thing. But it's so easy when you show somebody an example of what these can do. So I can very quickly build a persona that just knocks the socks off of what most people would be able to build themselves using the process that I've developed. And then by being able to build that into the AI, quickly show how, even with no extra training, we can turn pieces of content that are technically sound but maybe not focused on our persona, into something that is focused on our persona. So we can really quickly show that. We can quickly say, hey, look at this on the website and go here. And that's really easy to do.

Account-Based Marketing Applications

Speaker: Jasmine Gruia-Gray [25:09]

So we talked about how PersonaAI can be helpful for website reviews. We talked about it in terms of how it can help salespeople prepare for a pitch and prepare for objection handling. We talked about it in the context of product management and the use case in user requirements and functional requirements. And you now just mentioned helping to review marketing materials and brochures. Any other examples? Maybe some examples on how it can be used in your personal life?

Speaker: Matt Wilkinson [25:48]

Yeah, well, I can definitely talk about the personal life, but I think that the other great example that we haven't touched on today is account-based marketing. And I gave a webinar for SAMPS - some of this feels like it's a running advert for SAMPS today, so thank you, SAMPS. It's great to sponsor and be a member. But yeah, gave an account-based, AI-empowered ABM webinar, and showed how we can then use these tools - at least synthetic customers - to take generic marketing that's focused at just one of our buying group and cascade that out into multiple messages.

And how we can then go even further, if we really want to target individuals, using personality AI tools like Humantic AI, to then further understand within our persona - actually, what are the groups that we're looking to target? The personalities of the different people that might be in there. So we can do a whole range of things there that allows us to go from very generic marketing through to something that's feeling very personal and really hitting on the jobs that those people need to complete, the jobs to be done, without being kind of, hey, we've seen you visit X, Y and Z, and you've gone to this trade show and you published this, without feeling like we're being stalkerish. So that's one of the things I think is really powerful about this. We can personalise without being creepy.

In personal life, I have to say, I don't trust the AIs too much with my personal conversations. I still, as much as I love working with AI, still have some boundaries with it. But I think that you can get advice as to what's going to appeal to them, what's going to appeal to your persona, and so that really becomes the key thing. And again, talking about giving talks, getting up in front of people, being able to engage with your persona can give you a huge amount of confidence to be able to do that.

Building Around Feelings and Needs That Endure

Speaker: Jasmine Gruia-Gray [27:40]

Yeah, couldn't agree more. Maybe we can end with this. I just finished reading Mark Schaefer's latest book on how AI changes your customers. If our audience hasn't had a chance to read it, I strongly encourage it. You can see lots of dog-eared pages here and lots of ideas in the margins. But I wanted to read one sentence because I think it really encapsulates what PersonaAI is about: "In an age of AI, don't build your brand around features that are constantly evolving. Build it around feelings and needs that endure."

Speaker: Matt Wilkinson [28:22]

I think that's very, very powerful, and that whole book is amazing. And I think it's absolutely right. These personas should not replace the customer. But if we can use them to capture customer needs, customer feelings, and then cascade that through the process, we can really use them to focus on what's important to them. And I think that's how we make a difference, certainly in this first stage of AI.

Speaker: Jasmine Gruia-Gray [28:48]

Completely agree. This has been a fabulous discussion. I don't even think it was an interview. So thank you for your time, Matt, and thank you for creating PersonaAI and Atlas, which, by the way, more information is available on the strivenn.com website. And want to thank all our listeners for listening to another episode of Splice of Life Science Marketing.

Speaker: Matt Wilkinson [29:14]

Thank you, Jasmine. And thank you to everyone listening. All right. Bye for now.

Q&A

I only have three personas from an old project. Can I still use PersonaAI?

Absolutely. Start with what you have. Upload those existing persona documents into a custom GPT or Claude project. Even basic personas become more useful when you can ask them questions directly. The AI will work with whatever context you provide - and you can enrich it later with LinkedIn profiles, call transcripts, or survey data as you gather them.

How do I get started if I don't have budget for the full service?

Build a basic version yourself. Gather 10-15 LinkedIn profiles of your ideal customers. Use a reasoning model like ChatGPT or Claude with deep research to analyse them and generate a detailed persona document. Then upload that document as context for a new chat. You'll have a working synthetic customer within an afternoon.

What's the first thing I should test with my synthetic customer?

Your homepage headline. Ask your PersonaAI: "Based on who you are, does this headline speak to your biggest challenge? What would make you stay or leave?" You'll get specific, persona-grounded feedback in seconds. It's a quick win that demonstrates the value before tackling bigger projects.

Can I use this to prep for a single important sales call?

Yes - and you should. Upload what you know about the prospect: their LinkedIn profile, company information, any previous correspondence. Ask the AI to roleplay as them and challenge your pitch. Request the three hardest questions they might ask. You'll walk into that meeting far more prepared than winging it alone.

How do I keep my synthetic customer from going stale?

Feed it regularly. After every customer call, win-loss interview, or trade show conversation, add relevant quotes or insights to your persona document. Even a monthly update with fresh voice-of-customer data keeps your synthetic customer sharp and prevents the drift that makes traditional personas obsolete.