How are AI-powered personas changing the way we develop products and truly understand target groups? This is exactly what our managing director Thomas Immich discusses in an interview with Annett Bergk. The detailed conversation has been published in the book Praxisleitfaden für Künstliche Intelligenz in der Unternehmenskommunikation (Practical Guide to Artificial Intelligence in Corporate Communications) – and provides exciting insights for anyone who wants to use AI strategically and responsibly.
Classic personas are often well-designed profiles – but they are rarely used actively in everyday project work. For Thomas Immich, personas need to do more today: they should help make better feature decisions and reflect real insights from user research.
With generative AI, personas become capable of dialogue and learning. Teams can talk to them, discuss ideas, and question assumptions. New research findings are continuously incorporated—the persona evolves along with them.
Bringing research to life
With the LeanScopeAI, Centigrade often starts with an AI-generated proto-persona. The dialogue results in interview guidelines for real users. Their feedback is integrated iteratively, turning the persona into an interactive sparring partner – with a “memory” and an understanding of context. The key added value:
- Translation of data into tangible perspectives
- Greater empathy through dialogue-based interaction
- Better decisions through continuous learning
- Early detection of misconceptions
One thing remains clear: AI provides support—the responsibility for a product’s goals and values still lies with humans.
AI persona meets Wahl-O-Mat
Thomas took the topic of AI personas a step further in a public experiment.
For the Rhineland-Palatinate election, he used LeanScopeAI to generate a typical SPD voter persona. He used an agent to automatically run this persona through the Wahl-O-Mat.
The surprising result:
The persona did not end up with the SPD, but with a high degree of agreement with the DIE LINKE party – the SPD only came in eighth place.
The key insight
Such results depend heavily on the modeling of the persona, the data basis, and the structure of the Wahl-O-Mat theses. In addition, Wahl-O-Mat only measures content matches—not actual voting decisions.
Things got exciting when the discussion continued:
- How can qualitative and quantitative data be used to develop more realistic AI personas?
- Could programs be structured in a more “agent-friendly” way in the future?
- And how can AI help make democratic processes easier to understand?
The experiment was less a political statement than a methodological impulse: AI personas are powerful thinking tools – but their significance stands and falls with the quality of the underlying data. Used correctly, they reveal perspectives, sharpen discussions, and help to think through complex interrelationships interactively – in product development as well as in a social context.
