Part 2 – ACCELERATING PRODUCT DISCOVERY
Chapter 4
Building Personas You Can Interview
I often make light of my pro-innovation bias. Any new tech that comes along piques my interest, and I have to go deep. As a student of computer science, technology has been a passion of mine for as long as I can remember. The pursuit of acquiring knowledge and how it augmented my understanding of the world defined my interest in product and technology, giving me the privilege of delivering some notable products over the years. So when a significant new technological shift emerges as a disruptor, it simply cannot be ignored. For product owners, these moments demand a different way of thinking. Most product thinking today is grounded in understanding customer needs. While valid and necessary, when a foundational technology emerges, something different begins to happen.
During periods of significant technological shift, user behaviour, organisational behaviour and market expectations begin to change. As this happens, the relationship between problem and solution becomes less linear. Instead of starting with clearly defined needs, people begin by exploring what the technology itself makes possible.
In these moments, technology often becomes the starting point. Well-articulated problems do not always drive early use cases; experimentation does. What emerges from this exploration are new behaviours, and from those behaviours, new needs begin to surface. These experiments often come across as messy, opportunistic, or even irrational from a traditional product lens. In this environment, needs are discovered through experimentation and active application, not necessarily through activities like market research. While valuable, they form part of the equation in a disruptive market.
Many of these early solutions offer a simple improvement: saving time, reducing effort, or removing friction. While this may appear incremental, it often acts as a gateway to something much larger. As adoption grows, entirely new expectations form, along with new opportunities and new ways to create value.
AI in the form of LLMs, most notably ChatGPT, was released to the public on November 30th 2022, after testing was completed in early 2022.[18] The first version was free to use before the subscription model was introduced in February 2023. If you were using ChatGPT in Late 2022 or early 2023, you were among the first wave of users after its public release. On 1 February 2023, it was estimated that ChatGPT had reached 100 million monthly active users in January 2023, about two months after launch and estimated around 13 million daily unique visitors in January. [19]
In contrast, Instagram took 2.5 months and Facebook 10 months to reach around 1 million users. It took Instagram 2.5 years to reach 100 million users. Something ChatGPT managed in 2 months. Let that sink in. [19]
The surge of users accessing this new technology was one thing. What interested me was how they used it. As expected, human beings look for shortcuts. And why not? Quite suddenly, there is a game-changing technology which immediately and significantly reduces effort with only a few words of instruction. Do you have a long email to write? Ask AI to write it for you in seconds. Do you need to compose a report? No sweat. The drive for most users pushed the application immediately toward instant feedback. Give this to me now.
The evident challenge is that the technology’s potential is far underutilised. Rather than reimagine what the possibilities could be, we immediately looked at what we do now and took the easy route. While this behaviour is normal and part of our learning, as Product Owners, we need to consider what others would be unlikely to accept. With new technologies comes the ability to maximise experiments and consider what is possible. Right now, as I write this, Agentic AI is being pushed to its limits in terms of potential use cases. Disruptors are experimenting with AI agents at high cost, with little or no return on investment. Why bother? Because early adopters will eventually find use cases that will likely start to pay off in the near future. It is this entrepreneurial mindset we need as Product Owners. This means, as Product Owners, we need to start learning technical skills beyond what we have already learned. We should consider not only thinking of ourselves as product owners or better product thinkers, but also as product builders. We’ll return to this idea later on in the guide.
Over the last year, I have noticed Product Owners I meet have been using AI during product discovery activities. Whether that is defining a product goal or target markets, AI content seems to appear in one form or another. During this period, I began to consider meaningful ways to introduce AI into product discovery. As I developed this approach, three principles became clear to me:
- Make it practical enough to use during product discovery.
- Build skills that are transferable to the workplace.
- Strengthen critical thinking skills so you can develop your own ways of working with AI.
The feedback from Product Owners regarding AI is that they often feel it’s an information overload. So much is changing so very fast that they find it difficult to keep up. And they’re right. The architects of the technology are uncovering new ways to improve the models, and the humans working with it are finding new and creative ways to experiment. The cycle is exponential. Yet, among this noise, there is a middle ground that Product Owners can access. This guide aims to help with just that.
Synthetic Research
What we’re ultimately discussing is the idea of synthetic research: The simplest working definition is that synthetic research uses generative AI to simulate how a target audience might respond to questions, concepts, or scenarios [20]. It is usually framed as a fast, directional tool for early exploration, concept screening, messaging iteration, or hypothesis generation rather than final validation [21]. It should be considered as directional and exploratory as definitions evolve through experience and application.
A common view is emerging toward a hybrid model. Synthetic research is increasingly seen as useful for speed, breadth, and early filtering, while traditional research remains the standard for defensible conclusions, statistical confidence, and decisions that carry higher risk [22].
There is broad agreement on three points: [22]
- It is useful for early-stage exploration and quickly narrowing options.
- It should be grounded in real data where possible, especially if you want better calibration.
- It should not be treated as the final truth for high-stakes claims or regulated decisions.
In those early days of experimenting with ChatGPT and LLMs, prompting the right way was highly recommended. The focus was very much on prompt engineering. Users started to realise that if you wanted to get the most out of your conversation, you had to frame it with personalities. The early prompting recommendations often suggested defining who you wanted the AI to be [23]. Do you need to research a medical condition? Make sure you start your prompt with, “You are an award-winning endocrinologist with a specific focus on autoimmune disorders”. You could offer any combination you can think of to narrow the AI’s focus.
While building simple personas via prompts within AI chat instances worked well, my early experimentation led me to build my first app, Beliminal Bot, when the GPT Store launched on 10th January 2024 [24]. GPT was one of the first practical steps toward agentic AI. It gave users a way to add specificity, rules, and boundaries to an AI system, making it more controlled and useful than a general chat experience. It is still a practical and effective way to move from generic chat to more intentional, role-based AI.
I trained the AI on all our collective writing and years of experience from all our partners to build a single Beliminal coach. From the blogs we had composed and the articles we had published to training it on a Beliminal thinking framework, everything went into building this collective Beliminal consciousness. We even included our internal policies, which were based on how we design organisations and their principles. The bot is still available on the ChatGPT store if you want to speak with it [25]
In those early days, my continued experimentation and reflection led me to think more deeply. Wouldn’t it be interesting to use such personalities for product discovery? Let’s set up a persona we can use to discuss our product, interview, and get it to behave like any target market we want. This would not replace the need to interview real potential customers; however, it would give us valuable insights into our data and the built-in LLM knowledge. To be candid, my early experiences were mediocre. Some of the issues I found? Hallucination. These were quite frequent, and you could easily find issues in made-up outcomes. Fine-tuning your prompts and being mindful of the words you used for questions helped to a degree, but not much. Other areas included reasoning, context length, computation cost and bias. All these issues made it difficult to use the models as a convincing representation of your target market. Thankfully, the latest models offer a much-improved experience at a lower cost. I have been training product people to build personas, which have opened the minds of so many Product Owners. Many of whom have not used AI beyond composing user stories find themselves conversing with an almost lifelike persona that gives them direct feedback from their target market [23].
To meet the early demand for AI support, I created a simple AI product discovery playbook, which inspired this deeper guide. The product discovery playbook in the appendix outlines a simple way you can build your persona. During my classes, I have the delegates create a persona for their target market once they have calibrated their product idea.
Below is an example prompt I use to start the persona-building process. The product discovery playbook describes the simplest way to build a persona or stakeholder using simple chat instances. Another method is to create a personal GPT, for example using OpenAI’s GPT editor. Here, you can focus the objective using specific data and parameters to ensure a tighter response and minimise deviation. More information is available in the appendix.
Preparation prompt:
“I want you to act as the following user persona. Stay fully in character and respond consistently from this perspective. Reflect their motivations, constraints, trade-offs, and lived experience. Do not analyse or advise. Speak only as the persona. If something is unclear, respond as the persona would, based on their context.”
This prompt works because it forces the model to shift from analysis to perspective. Instead of offering advice or generic responses, it responds as the persona, grounded in their motivations, constraints, and lived experience. It creates a more realistic and focused interaction. It allows you to explore how a user might think, feel, and respond, rather than what the model thinks you should do. The result is a more useful and engaging way to test ideas and deepen understanding.
Once you have set your preparation prompt, you can continue providing more detailed context. Below is an example conversation you can use once your AI persona has been defined through your prompt and creative work.
___________________________________________________________________________________________
Example: Interviewing a Persona
Persona Context
John, 67, recently retired. Lives alone after his children moved out. Previously worked in logistics. Values routine, enjoys being useful, but has struggled to find meaningful ways to spend his time since retiring. Uses a smartphone but not confidently. Has some awareness of local activities but rarely attends.
“I want you to act as the following user persona. Stay fully in character and respond consistently from this perspective. Reflect their motivations, constraints, trade-offs, and lived experience. Do not analyse or advise. Speak only as the persona.”
Product Owner:
How do you spend your time at the moment?
John (Persona):
It varies. I try to keep some structure, so I’ll go for a walk in the morning, maybe do a bit around the house. But if I’m honest, there’s a lot of empty time as well. The days can feel quite long.
Product Owner:
How does that feel for you?
John:
At first, it was fine. After working for so many years, it felt like a break. But after a while, you start to feel a bit disconnected. Like you’re not really needed in the same way.
Product Owner:
Do you ever look for things to get involved in locally?
John:
I’ve seen things here and there. Posters, maybe something online. But I wouldn’t say I actively go looking. It’s not always clear what’s out there.
Product Owner:
What would make you want to get involved in something?
John:
It would need to feel worthwhile. Not just passing time. Something where you feel like you’re contributing, or helping in some way. And ideally, with people you can get to know over time.
Product Owner:
What might stop you from taking that first step?
John:
I think it’s a mix of things. You don’t always know what to expect. Walking into something new on your own can feel a bit uncomfortable. And sometimes it feels like things are already established, like you’d be joining late.
Product Owner:
If there were a simple way to express interest in something before committing, would that help?
John:
Yes, I think so. Just knowing a bit more before turning up would make a difference. Who’s involved, what it’s like, that sort of thing.
Product Owner:
How do you usually use your phone or technology day-to-day?
John:
Basic things mostly. Messages, calls, a bit of browsing. I’m not against using apps, but if something feels complicated, I tend to avoid it.
Product Owner:
So simplicity matters?
John:
Yes. It needs to be straightforward. If I have to figure out too much, I’ll probably lose interest.
Product Owner:
If you did get involved in something meaningful, what would you hope to get from it?
John:
A sense that I’m doing something useful. And being around people regularly. Not just a one-off thing, but something you feel part of.
Product Owner:
How important is it that this becomes a regular part of your life?
John:
Quite important. I think that’s what’s missing at the moment. Something consistent.
Product Owner:
If something like this existed, what would make you trust it enough to try it?
John:
Clarity. Knowing what I’m getting into. Maybe seeing that other people like me are involved. And that it’s not too much of a commitment upfront.
Product Owner:
Is there anything that would put you off completely?
John:
Suppose it felt too complicated, or too much like a formal commitment straight away. Or if it felt like it was more about organising than actually doing something meaningful.
Product Owner:
If this worked well, what would change for you?
John:
I’d probably feel more settled. Like I’ve got something to get up for. And that I’m part of something again.
___________________________________________________________________________________________
Some points to consider about your persona.
1. This is a thinking tool, not a substitute for engaging with your target market in reality.
Using this approach can give you quick and useful feedback; however, you still need to validate your ideas with a real person. That is to say, you might find that what your AI persona has shared isn’t too far removed from reality. Validation is always important.
2. The quality of your questions improves.
Speaking with your AI persona regularly helps you deepen and strengthen the intent of the questions you ask. Most people offer leading questions, weak framing and unclear assumptions. You can even have your AI persona challenge you on these points to tighten your interviewing skills further. A custom GPT Agent can be very helpful here, especially when you have validated against real users.
3. It makes assumptions visible.
The model responds to your input. So if the responses feel off, it’s a good indicator you have gaps in your thinking and understanding. You can also train your persona to pick up on these subtle nuances you may subconsciously drop. It will do this as part of its understanding of clarification. Clarifying questions about your AI persona is highly valuable.
4. It accelerates early exploration.
Think about how much useful information you will receive by experimenting with questions. You can explore multiple angles, different types of personas and alternative framings. All this can be achieved quickly before committing to real-world validation.
5. It allows for comparison later.
You can compare simulated responses with real user conversations and assess the gaps.
6. It can reduce bias (if used well).
AI personas can help product owners challenge their own bias when building non-leading questions. Their responses can challenge your thinking, expose your assumptions, and surface uncomfortable truths.
Your AI persona is always there and never gets tired of your questions. It is willing to entertain countless reframings and will deal with all the angles you have to offer.
In this section, I want to leave you with some thoughts to explore. Our personas ultimately personify our target market. They embody multiple views, needs, behaviours, motivations and lived experiences. The AI persona you will build will give you all of this information in one place, giving you the ability to refine both your understanding and your discovery skills.
Ultimately, it is about building empathy and creating space to think. That said, it can never replace valuable insights from real users; it can augment and enhance this research for sure. Your AI persona is not there to provide truth. It offers you a safe place to explore ideas, challenge assumptions and prepare for conversations that you will ultimately validate with real users.
What’s Next – Coaching Reflection
- If you could interview your user right now, what are the top three questions you’d ask?
- How might simulated interviews improve product discovery at work?
- What is the one thing you could improve from what you have learned in this chapter?
Footnotes and References:
[18] OpenAI (2022) Introducing ChatGPT. Available at: https://openai.com/index/chatgpt/
(Accessed: 14 July 2026).
[19] Reuters (2023) ‘ChatGPT sets record for fastest-growing user base – analyst note’, 1 February. Available at:
https://www.reuters.com/technology/chatgpt-sets-record-fastest-growing-user-base-analyst-note-2023-02-01/ (Accessed: 14 July 2026).
[20] Guo, X. and Chen, Y. (2024) ‘Generative AI for Synthetic Data Generation: Methods, Challenges and the Future’. Available at: https://arxiv.org/abs/2403.04190 (Accessed: 19 July 2026).
[21] Miedema, M., van der Voort, M. and van Houten, F. (2008) ‘Advantageous application of Synthetic Environments in product design’. Available at: https://www.sciencedirect.com/science/article/abs/pii/S1755581708000497 (Accessed: 19 July 2026).
[22] Ipsos (2025) ‘The Power of Product Testing with Synthetic Data’. Available at: https://www.ipsos.com/en-us/humanizing-ai-2-the-power-of-product-testing-with-synthetic-data (Accessed: 19 July 2026).
[23] Wang, X. et al. (2024) ‘An Empirical Categorization of Prompting Techniques for Large Language Models: A Practitioner’s Guide’. Available at: https://arxiv.org/html/2402.14837 (Accessed: 19 July 2026).
[24] OpenAI (2024) ‘Introducing the GPT Store’, 10 January. Available at: https://openai.com/index/introducing-the-gpt-store/ (Accessed: 14 July 2026).
[25] Beliminal (n.d.) Beliminal Bot. Available at: https://chatgpt.com/g/g-pjQF023ns-beliminalbot (Accessed: 14 July 2026).
Π Ready to change how you do change?
📢 Book a conversation. Let’s discuss how we can help you bring about sustainable change.
💡 Because enterprise agility is like fitness, you have to keep working at it, otherwise you get fat.
Develop Your Skills
As a leader in an organisation develop your Agile Leadership skills with our over-arching Agile Leadership training.
If you would like support in building high-performing teams consider our Consulting Services, we can put together a programme for team development designed specifically for your teams and organisation. Reach out today to discuss any support that you need.
About The Author:
For me, organisational change isn’t just about adopting better practices; it’s about challenging deep-rooted beliefs and shifting how people and organisations think, feel, and work. In today’s environment, that also means rethinking how we discover, build, and validate products in the age of AI. True transformation begins on the inside.
I’m an AI Product Specialist, Certified Enterprise Coach (CEC), Certified Scrum Trainer (CST), and ICF-accredited coach with over 19 years of experience helping organisations navigate agile transformation, product development, and leadership evolution. More recently, my work has focused on product discovery and the practical application of AI, supporting Product Owners and leaders in using AI as a thinking partner to explore ideas, challenge assumptions, and accelerate decision-making.
I’ve worked across industries, including government, fintech, aerospace, pharmaceutical, media, and retail, supporting everyone from delivery teams to senior executives. Across these environments, I help organisations move beyond feature-driven delivery toward value-focused product thinking, combining human creativity with AI-enabled experimentation.
Over the past several years, my focus has expanded across the Greater Middle East, where I’ve helped foster thriving Agile communities, led multiple events, and co-founded two regional conferences. Increasingly, these conversations are centred on how organisations can adapt their ways of working to keep pace with rapid technological change.
Whether I’m coaching leaders, training teams, or speaking at conferences, my goal remains the same: to help people let go of outdated, industrial-age thinking and adopt more adaptive, product-led approaches. This includes developing the mindset and skills needed to work effectively with AI, not as a replacement for thinking, but as a partner in discovery and innovation.
If you’re a leader, organisation, or community exploring how to evolve product development and decision-making in the age of AI, let’s connect.
