A lot of products are much of a muchness. There is little that differentiates them; they work perfectly adequately. That’s why, when a device has that extra something – a factor that you can’t quite put your finger on – it immediately stands out. That point of differentiation is a feature constant across iconic products and almost instantly makes them a classic in design terms, whether it was Sony’s Walkman, the Nest Thermostat, or the iPhone.
That intangible extra is often referred to as a product’s ‘soul’: what helps it transcend being merely an item that performs its function and weaves it into lifestyles, routines, and daily rituals. It’s a product that has human-centred design baked in, character, a clear purpose, and the ability to evolve with its users’ needs. Sometimes that extra bit of magic is in the way you interact with the product, other times it comes from a completely unexpected aspect – for instance, the subtle delay that makes large language models (LLM) chatbots seem like they’re typing rather than immediately synthesising a response.
Those details are the result of painstaking hours of research, exploring the intersection of psychology, human behaviour, technology, and design, along with significant amounts of investment. They emerge from sitting with people in focus groups, noticing what they don’t say as much as what they do, and reading user reactions to the ideas and prototypes put before them. It’s an incredibly time and resource-intensive process that is a stretch for many established companies, and too often almost outwith the means of start-ups and early stage businesses developing new technology.
What is a ‘synthetic’ approach
But a new, AI-enabled approach is beginning to change that. A growing number of brands are on the public record for opting to use ‘synthetic focus groups’ – personas developed by LLMs, using survey data – as a more efficient and cost-effective alternative to testing new ideas, products, and services on conventional consumer panels.
Through this exercise, the AI-generated personas simulate how different types of users think, behave, and respond to a product, service, or idea – in other words, ‘role play’. Like real people would in a traditional focus group, they ‘interact’ with prompts, scenarios, and each other. The value of roleplay is that it tests assumptions early, exposes blind spots in design, and reveals how different user groups might react in the real world.
On top of that, the role play element can be integrated throughout the design journey. Instead of speaking to people at project milestones, the designer can have a constant dialogue with the synthetic persons to validate assumptions at different stages, test different ideas, and talk through different approaches and options as the product begins to take shape.
The advantages of going synthetic
There are some clear advantages to taking this approach. For one, the cost and time commitments are far lower than traditional, in-depth consumer research. Finding the key points from a survey with 10,000 responses takes a lot of legwork, and synthetic focus groups can aggregate the data and distil that down to a handful of personality types that then interact with one another to give you useful feedback.
This is a massive efficiency gain. If AI can handle the baseline functional requirements – the ‘what’ and ‘how’ – and get you 92% of the way to a finalised concept, it clears the noise so designers can focus entirely on the ‘why’. That’s the 8% extra that makes a product more than the sum of its parts – in other words, its soul. And that can be done at every stage of the design process, through that consistent dialogue with the synthetic personas.
But that shouldn’t be mistaken for replacing real users. Instead, synthetic focus groups should be seen as a fast, low‑risk way to explore reactions, refine messaging, and identify friction points throughout a product design process before involving actual participants. That then leaves you able to focus on the individuals, or groups of people, who will give you the insights that truly make a difference.
In practical terms, if you’re a furniture company expanding into China, synthetic users can tell you about the physical constraints, modularity needs, and other fundamental requirements unique to that market. When speaking to real people, you can then dive straight into the deep cultural rituals that are more difficult to pinpoint, exploring where the subtleties of interaction between users and product can be found – after all, technology is at its best when it’s invisible.
The validation challenge for designers
With any new technology, there are inevitably challenges for people to factor into adoption. For instance, in any kind of research there is the ever-present trap of using synthetic focus groups to validate your own assumptions. The worst outcome you can have from consumer research is that your product is perfect – that invariably means something has gone wrong somewhere along the way. If 99 people out of 100 agree with your assumptions, you need to focus on the one that doesn’t.
That is a particularly pertinent issue in AI. The technology’s sycophancy is a known phenomenon, prioritising flattery and approval over accuracy. And that can feed into one of the most common pitfalls in design, where designers are convinced of a particular solution to a challenge. If a brand goes into this process with preconceived notions about what a product should be and an LLM validates those assumptions, they will likely double down without having properly stress-tested the idea.
This is a technical challenge designers need to be aware of and aim to solve. Instead of using AI for validation, the synthetic role play principle should be employed to simulate adversarial personas or contradictory feedback to find the ‘awkward pauses’ and friction where real innovation happens. It could also be used to help amplify any biases in a way that makes it more obvious where there could be unconscious bias at each phase of the development journey.
A quicker route to better products – particularly for start-ups?
For any company developing new products, synthetic focus groups are a huge opportunity. While established brands – such as Coca-Cola and The Times – have embraced this approach so far, the biggest beneficiaries may be smaller, more agile companies. Better access to enterprise-grade data processing could allow them to compete with larger competitors by finding the magic in their product ideas faster and most cost effectively – and that can only be positive for innovation.
When people talk about the productivity gains AI will enable, this is a great example of what that means. Finding a product’s soul is imperative to its success and if we can make that process 90% more efficient, then we could be looking at better products developed more quickly, with enhanced outcomes for the customers that ultimately use them. But it shouldn’t be confused with relying on AI and synthetic users entirely – that comes with its own risk of creating nothing but middle-of-the-road, vanilla products based on nothing more than consensus.
AI is the map; designers’ intuition is the compass. The best products come from dissonance: the awkward pauses in interviews, the contradictory feedback, and sometimes pure gut instinct. Far from being about replacing the human, it’s about getting to the right humans to have the right conversations, and the more time you can spend focused on them, the better the outcome is likely to be for everyone.
