Article by ICG member John Habershon
Personas have been the quietly powerful output of qualitative research – often created, presented, then set aside as merely interesting additions. I believe AI will transform personas from being a nice-to-have deliverable into what they’ve always promised: the central organising principle for understanding brand value and the roadmap for targeted marketing.
The process of creating personas – identifying patterns and clustering participants into meaningful groups – is a key creative analytical skill of qualitative researchers. But it is also a core strength of AI.
How do we exploit the power of the new AI tool whilst maintaining our qualitative skillset?
The Traditional Art of Persona Creation
Creating personas has always been one of the most satisfying aspects of qualitative research. It’s where analysis transforms into recognition – that moment when a client sees a set of consumers reflected in a persona you’ve uncovered. When each persona sparks genuine recognition from customer-facing people in sales and marketing you know you’ve captured something real.
The process exemplifies what we achieve through rigorous creative thinking and human empathy.
But the traditional process has always had limitations: time constraints, restricted sample sizes, and the sheer cognitive load of pattern-matching across dozens of interviews.
Case Study 1: Plant-Based Meat – Human vs AI
A couple of years ago I conducted research testing six advertising images and product propositions for plant-based meat.
Based on interviews with consumers with families, I created five personas by hand:
‘Trad’
‘Reducer’
‘Trialist’
‘Foodie’
‘Veggie’
Recently I returned to the data, uploaded the 30 transcripts to Claude AI and asked it to create five personas.
Here’s what emerged:
‘Traditional Sceptics’
‘Pragmatic Reducers’
‘Convenience Adopters’
‘Ethical Pioneers’
‘Curious Experimenters’
The table below compares the two sets of personas side-by-side:
The AI identified the same core segments I had, but the differences made me re-examine my own thinking. There was no ‘Foodie’ in the AI set – on reflection, this was a small number in my sample that had made an impression on me personally. The same applied to the ‘Veggie’ category. Incorporating this under Ethical Pioneers made sense. Similarly, Foodie could be seen as a subset of Traditional Sceptics.
What the AI added was rigour: it gave me the percentage of the sample for each persona and could list every respondent under each. It also identified crossovers and showed which respondents fit into more than one category.
The AI’s approach was multi-dimensional, cross-referencing all the factors. This ensured each respondent was classified based on behavioural patterns, not single statements, creating robust, actionable personas.
Case Study 2: Running Shoes – Where AI Changes Everything
At just thirty half-hour interviews, the plant-based meat study was manageable. With careful investigation and imaginative reasoning, I could create personas by hand.
But my running shoe brands project operates at an entirely different scale: 7,432 social media conversations from Twitter/X (27.3%), Reddit (68.6%), and compiled references (4.1%).
The AI sorted through this data to create six male runner personas:
‘The Value Hunter”
‘The Max Cushion Comfort Seeker’
‘The Performance Optimiser’
‘The Brand Loyalist’
‘The Pragmatic Beginner’
‘The Durability Demander’
For each persona, it provided detailed information on demographics, running profile, brand preferences, attitudes and motivations, purchase triggers, and pain points.
To differentiate between personas, the AI looked for patterns across a extensive range of factors: purchase behaviour, running experience indicators, pain and injury language, product knowledge indicators, decision-making language, age and life stage clues, lifestyle and context signals, shoe rotation and collection behaviour, brand interaction patterns, emotional tone and language, durability expectations and complaints, and purchase channel preferences.
It’s not just that AI manages huge volumes of data. It acts as an inference engine, drawing together comments, preferences and reported behaviour and making deductions in a Sherlock Holmes fashion. At this scale and complexity, achieving this manually isn’t just difficult – it’s inconceivable.
Why Personas Could Become More Central
This evolution in capability changes what’s possible. Here’s why I believe personas are about to become more central to research and strategy:
1. Scale Meets Depth. We can now create robust personas from datasets that were previously impossible to analyse comprehensively. Very large datasets from multiple groups and interviews, social listening data (as in this case), customer reviews, customer support conversations – all become sources for persona development.
2. Living, Evolving Personas. Traditionally, personas were created once and slowly became outdated. With AI-assisted analysis, personas could be continuously refined as new data is available, keeping them relevant and actionable.
3. Enhanced Targeting and Testing. AI-generated personas aren’t just analytical outputs – they become active tools for testing concepts, messaging, and advertising before launch, as demonstrated in the LinkedIn articles below.
Quantitative Analysis – Adding Even More Power
Here is an example based on a huge dataset of 2,700 respondents with detailed verbatim responses from two open-ended questions. AI is invaluable in creating a coding frame for each question. But we can take it much further and get even more value from asking Claude to create a set of personas. It will analyse the verbatim comments from both open-ended questions and also include the data on age bands, gender, social grade, work status, household income, marital status, presence of children, tenancy type.
Here is an example of one of the 12 personas
Persona 4: Affluent Eco-Adopters
Demographics
• Age: 35-64 (primarily 35-54)
• Size: 456 respondents (8.5% of sample)
• Gender: Evenly split
• Work Status: 73% full-time employed
• Income: £75,000+ (£100,000+ for many)
• Social Grade: Primarily A/B
• Children: Mixed
Defining Characteristics
Affluent Eco-Adopters are actively investing in sustainable technology and products. They’ve purchased electric
vehicles, installed solar panels, choose organic and sustainable brands, and view environmental responsibility as both
personal obligation and increasingly good investment.
This group largely continues their lifestyle with minimal disruption. Whilst noticing price rises, these haven’t forced
significant behavioural changes.
Verbatim Examples
• “I’ve opted for an EV and try to be better on energy management around the home”
• “I buy organic where possible, avoid fast fashion, and support local businesses. Money is not the barrier to
sustainable living for me”
• “I’ve not seen a real change other than prices rises. We’re fortunate to be in a position where we don’t have to
make difficult choices”
• “I am a lot more worried about high interest rates. I am saving a lot more and am very conscious of my spending”
Key Insights
• Financial capacity to align spending with environmental values
• Early adopters of sustainable technology (EVs, solar, heat pumps)
• View sustainability as investment rather than sacrifice
• Some guilt or awareness of privilege regarding financial security
The resulting 12 personas, could better be described as market segments since they are based on quantitative data. Being able to conduct further exploratory data analysis based on these segments is a major opportunity.
Where This Takes Us
Persona creation represents a perfect merger of human intuition and judgement, combined with AI rigour.
Additionally, with the help of Claude we also have the scope to test communications against each segment, creating genuinely targeted advertising and marketing.
The question is no longer whether AI can help create better personas, but how quickly researchers will embrace this capability to deliver insights that are both deeper and more immediately actionable.