For product managers and CX/UX designers, understanding their target group is essential in order to develop products and services that really resonate.
Personas, fictional characters that represent real users, are indispensable tools here. They make it possible to give the rather abstract target group a face, recognise their likes and dislikes and tailor everything from product development to marketing and communication to the specific needs of the users.
However, creating personas can be a challenge. It is often time-consuming and resource-intensive, as data must first be collected and clustered, target group-specific characteristics identified and then translated into vivid persona descriptions.
This is where LeanScope AI,, a new AI-supported tool, promises to change the game.
The development of personas, traditionally a process of several days or even weeks, depending on the depth of the necessary market and target group research, the number and complexity of the personas and the required creative and iterative customisation processes, can be done in a matter of seconds using LeanScope AI. LeanScope utilises the new possibilities of generative artificial intelligence to create credible personas quickly and efficiently, which can then be refined and adapted using user research findings.
In addition to the simple creation of convincing persona profiles, LeanScope AI now also offers the option of interacting with the personas created. This means that the personas can be addressed directly or conversations with and between them can be simulated. As a LeanScope user, this gives you further opportunities to gain insights into your target groups and obtain information about users and their needs.
I had the opportunity to talk to Thomas Immich, the founder of LeanScope AI. In it, Thomas introduces us to the tool in detail and we discuss with him the opportunities – but also the risks – that arise from the use of AI in the creation of personas and product development and marketing in general.
Dear Thomas. Thank you very much for taking the time to talk to us about LeanScope!
With pleasure!
Can you tell us what exactly LeanScope is and what inspired you to develop this tool?
LeanScope is a product discovery tool that POs or CX/UX professionals can use to find products that actually fulfil real user needs more quickly and accurately.
Overview of the possibilities and functional areas of LeanScope
Unfortunately, there are enough products out there that miss the actual core needs of the target group – whether digital or analogue. And it is quite cost-intensive to only realise such a misalignment after the launch and then have to correct it afterwards. On the one hand, you won’t reach certain users in the first place, and on the other, you may even lose those you had already reached.
What specific benefits does LeanScope offer product managers and CX/UX designers?
In my opinion, LeanScope solves a fundamental problem in the world of experience design: access to real users is too often blocked or simply too expensive in continuous use.
The centrepiece of LeanScope is therefore its sophisticated role and persona module. On the one hand, you can define the “jobs to be done” or tasks of your target group at role level. On the other hand, you can be very specific and define specific personality profiles and intrinsic motivation factors, as is usual with good personas.
Instead of these personas just “hanging around” on the whiteboard or in Confluence, LeanScope transforms them into so-called “persona agents” that you can interact with at any time. For example, you can brainstorm product ideas or prioritise user needs with them. You then see the world from the perspective of your users.
The threshold for getting input from the target group is therefore extremely low.
Can you give an example of how you use LeanScope in a real project for you or your clients?
Basically, we use LeanScope in every customer project of Centigrade from minute 0, because we believe that you shouldn’t discuss product ideas until you know who has which need, which user in which role and in which context of use.
But a really great example that I would like to mention here is the BIGEKO project, in which we are carrying out user research and usability testing with Centigrade, among other things. This research project deals with the question of how deaf people can be supported much better in communicating in emergency situations. Unfortunately, despite digital advances, there is still a gaping accessibility gap for those affected.
It already starts where a deaf person in an emergency situation is actually deprived of any possibility of making themselves understood in the individual emergency institutions in a short time. The emergency helpers or control centre dispatchers at the fire brigade, police or in hospitals are of course not trained in sign language and are therefore unable to receive the important and urgent information about the emergency promptly or accurately enough to initiate the correct action.
What it means to be deaf from birth is one of the many things that we as hearing people cannot even begin to imagine. We had to make this painful experience during user research, because of course we don’t get the insights that are important to us in the conventional way. Traditional user interviews, for example, are simply impossible without involving an interpreter before, during and after the interview. Naïve as we were, we thought that we could conduct the user research in a purely digital, written form, but a person born deaf can often only read or write to a limited extent or with great effort.
It also turned out to be extremely difficult to get an emergency worker in the control centre of a police or fire station to take part in a user interview, as their precious time is understandably very limited. In order to provide the team with sufficient user research information, we actually demand a disproportionate amount from the users in both cases.
This is where LeanScope comes into play: We first used LeanScope AI to generate proto-personas for deaf people and various emergency responders. We then did a lot of desk research and also uploaded relevant studies on this topic to LeanScope. This gave us a pretty good “grounding” and an orientation to prepare us for the upcoming “real” interviews.
For example, we were able to enter into a “conversation” with our proto-personas very early on in order to develop a feeling for their sensitivities and actual limitations. Of course, the AI-controlled deaf personas were able to communicate with us in writing without any restrictions, but at the same time they made us understand how difficult the actual emergency situations are for them and how poorly most naive solutions would work in practice.
We were actually able to build up empathy with our target group before we actually came into contact with them. It was a fascinating experience for me personally. Carla Biegert, a Master’s student, is currently writing her Master’s thesis on the topic of “Empathy enhancement through persona agents”. A few of her most important findings can already be found in a corresponding blog article.
But we weren’t quite finished with the project yet: after our “real” user research with actual people had taken place, we fed this information back into LeanScope in order to turn our proto-personas into validated personas and to mark our false assumptions accordingly.
The other project partners, especially those implementing the project, could then write to these validated persona agents at any time to check whether individual feature ideas meet the actual needs of those affected or to expose nonsensical ideas at an early stage.
To minimise the hurdle of slipping into the shoes of those affected, we have even written a Teams add-on that allows you to add the persona agents to your own Teams contact list and write to them directly.
How exactly does LeanScope work when creating personas? Can you take us through the process?
The great thing about LeanScope is that as a PO or UX professional, you don’t have to start with the famous white sheet of paper. The AI generates very plausible proto-personas based on simple keywords. Among other things, we have incorporated our Centigrade agency experience from many years in the UX project business.
Persona of a user researcher created in LeanScope AI
From then on, the process is actually similar to that described for the BIGEKO project: the proto-personas are enriched with desk research or user research findings and then incorporated into the brainstorming process in order to arrive at better product ideas or feature prioritisations.
Sometimes the process also starts directly on the basis of existing customer data. For example, we have customers who use their CRM data to segment a small number of perfectly “cut” personas from the thousands of (anonymised) customer data available there.
How flexible are the personas created by LeanScope using AI? To what extent can or should users customise them further?
The basic philosophy of LeanScope is that any information generated by the AI can be challenged, deleted or overwritten by a human at any time. Whether you want to change certain motivation or frustration factors or simply replace the automatically generated photo – anything is possible.
We are also currently working on ensuring that subsequent manual changes are also subjected to appropriate plausibility checks. For example, it would be inconclusive for the persona user if I manually uploaded a photo of a 50-year-old man to a 35-year-old female persona. In my opinion, human control, plausibility, coherence and trustworthiness are very important criteria for the basic acceptance of the entire approach. And, of course, citing sources when required, i.e. good explainability, is something I consider essential.
You can now also interact with the personas created via LeanScope. How exactly does this work and what are the benefits?
As already mentioned, you can chat with the personas in writing. An interesting twist here is that I can also chat with the persona in a specific role, because the questioner always influences the respondent’s reaction, regardless of the question itself.
For example, if I, as a shift supervisor, ask a machine operator at which points in the workflow he makes the most mistakes, the machine operator might say that there are no such points, so as not to get himself into trouble. It might be different if I asked this machine operator the same question in the role of a close friend. It would be different again if I asked the machine operator this question as himself, i.e. if I started talking to him. LeanScope also supports this and sometimes reveals quite surprising findings.
In my podcast UX Therapy AI, I show that much more is possible. On the one hand, you can get the persona agents talking to each other so that they can exchange information autonomously without the need for human intervention. It’s really crazy what dynamics this creates.
Secondly, LeanScope now has a voice output module so that you can simply listen to the persona agents without having to read. Product owners can have a conversation generated between two user groups, let’s say a tax consultant and a client, and then take this spoken conversation on a business trip as a podcast 🙂
We are also currently working on a second version of our Person-a-Mat, a kind of ticket machine for printing posters of AI-generated personas for on-site design thinking workshops. You can now even interact with the personas by voice or, to put it bluntly, “chat” with them naturally. We presented the first version of the machine for the first time at the Mensch & Computer Conference 2023 and were subsequently even invited by Bosch to moderate their Generative AI Design Workshop.
Can you tell us about a conversation with such a persona that particularly surprised, amused or frightened you?
Wow, yes! There were so many of these moments that I have to think about what to choose.
Maybe I’ll start with the surprise. One of the first personas we generated for testing purposes was the garden gnome Gregory Smith. When I asked him which app he would spend money on, he replied: “I’m a garden gnome, I don’t need an app!”. This may sound trivial, but it’s still surprising, because I at least was used to ChatGPT always giving me an answer that was reasonably satisfactory. In this case, however, Gregory remained fully in character and didn’t make a point of buttering up his counterpart by pretending to have a “digital affinity”.
The persona of the garden gnome Gregory Smith generated via LeanScope AI
I generally find personas that have certain character traits that are not necessarily politically correct funny. For example, we asked a cynical teacher persona why she didn’t want to abolish parents’ evenings if she was so frustrated by them. The – naturally cynical – answer to this was literally: “Oh, abolish parents’ evenings… A tempting idea. But then we would miss out on the joys of robbing parents of the illusion that their child is the undiscovered Einstein. Where’s the fun if we don’t occasionally inject a little reality into parental dreams?”
Well, and scared? During our last UX Therapy AI episode, I was actually shocked that it was virtually impossible for Henning and I to convince the two persona agents, Pascal the PO and Valentin the salesperson, that it would be a good idea to involve the UX researcher persona Eva Schneider to do a little user research before implementation… in other words, before the features requested by the customer are implemented in a hectic manner and then perhaps not needed at all.
The PO still said that he thought it was a good idea, but not in this project at this time. Ultimately, however, they both remained stubborn until we wrote the following into their company’s strategy: “No projects may take place without user research.” But that can’t be the case… I would have liked a little understanding.
Yes, that was indeed frightening… namely “frighteningly realistic”, because unfortunately I know it in part from my experience with our Centigrade customers.
What kind of data does LeanScope need to generate personas? How do you deal with the issues of data protection and data security?
We take these issues very seriously in our partnership with LeanScope, because Centigrade has grown up with traditional industrial companies from the German SME sector, which in turn attach great importance to the topic. The very first version of LeanScope was even completely offline-capable and did not even require an internet connection. Incidentally, the usage scenario for this was around 5 years ago: “UX researcher is on the train, has poor wifi and has to process his findings from the on-site user research in LeanScope.” Unfortunately, not much has changed so far ?
Offline capability is of course not really practicable across the board with an AI substructure, but this basic architecture of LeanScope actually enables us to potentially switch to a locally running open source LLM as soon as we want to. So away from OpenAI and towards OpenLLama, for example.
In terms of the “expediency” of our data processing, we start very small. As already mentioned, a role description is actually enough to create a good proto-persona. This is why proto-persona generation already works in the freely available version of LeanScope. If you want to use more data in order to generate further information from it, you have to activate it successively and individually so that the purpose is always in balance with data security.
As LeanScope never “talks” directly to OpenAI or any other provider, we can also remove or anonymise real names beforehand. For example, we can also transcribe meeting recordings locally on our own servers to remove any personal reference before the transcript flows to LeanScope.
Ultimately, for me, the persona approach itself is the answer to the rightly very restrictive GDPR: as we burn all bridges to the original people for the personas obtained from accumulated customer data, we can talk to and with the personas about anything without having to worry that we are denigrating a real person. I casually call this “persona-related” data – and it is completely uncritical.
What concerns might users have about the use of AI-generated personas and how do you address these concerns?
As always with such progressive topics, there are of course many concerns. Some of them are also justified and need to be addressed more or less urgently.
One concern is that personas could provide false information or that they appear credible even though no user research has taken place. We want to counter this in LeanScope by not using personas without user research backing to provide certain information in the first place because there is no basis for it. Proto-personas are then more “closed” in conversations than validated personas.
Assumptions can be marked as such in LeanScope and then verified through user research.
Another concern is that, as a designer, you might even become too involved in the user’s world and therefore find it harder to “let go”, e.g. when it comes to compromising on the solution in terms of the project budget. Carla also addressed this important “detaching” phase in her blog article.
How do you deal with the challenge that AI-generated personas may be too stereotypical or not diverse enough?
The short answer is: personas even have to be stereotypical in a limited way, because otherwise they would not represent a simplification of reality and would be unsuitable as models. The long answer is of course far more complicated.
On the one hand, we should create diversity by simply having a diverse set of personas instead of developing a product based on a mono-persona. However, it is also important that this well-intentioned diversity reflects the “piece of reality” that contributes sufficiently to product development.
In my opinion, there are no advantages to designing an app for very high-priced kitchen appliances, for example, and then working with overly young personas due to excessive diversity ambitions, which then make up too small a part of the actual target group due to the assumed income levels.
Conversely, it is extremely important to me that great products also see the light of day for niche target groups or minorities and are not just aimed at the mass market. In my opinion, personas can provide excellent support in immersing people in different realities beyond the mainstream.
I think that most designers who grew up in the city, for example, will not find it easy to put themselves in the shoes of a farmer until they use the new version of LeanScope to generate a farmer including a “Journey of Pain” ?
Or, as mentioned in BIGEKO’s example, a hearing designer will find it difficult to empathise with a deaf person. And in the same way, a healthy designer will find it difficult to adopt the perspective of a child with ICP.
But the fundamental question was whether the AI has enough information to generate this diversity. I am of the opinion that time is working in our favour. The models are becoming more and more diverse. This has been particularly noticeable since the latest update of DALL-E 3, as more diverse persona photos are being generated much more frequently than before, so I’m very confident here.
Ultimately, the same applies here: do proper user research and include this data in your personas! This is the best recipe for integrating the topic of diversity cleanly and correctly.
How do you see the future of AI in product development and UX design?
One of my more frequently commented statements on this is: “In the medium to long term, AI assistance will mean that the UX designers & engineers who use it will make those who don’t use it redundant.”
I also believe there will be a shift towards user research, because in my opinion this is the one discipline in product development that can help AI to become decent. UX design will also be around for quite a while, but the job description will (have to) change significantly towards conversational UI and prompt engineering and, above all, learning why LLMs spend what they spend.
However, UI design in the sense of visual icon & screen design and layouting has long been a thing of the past for me.
But it’s a complex question. For those who want to delve deeper into it, I have summarized my thoughts in my article UX meets AI.
What would you recommend to product managers and UX designers who want to work with LeanScope? Are there any best practices or special tips for using the tool effectively?
Be curious, try out lots of things, even generate completely exaggerated personas to illuminate the spectrum. Ask the personas unusual questions! Compare the differences in the conversations with or without user research information. And of course: follow my UX Therapy AI podcast, because I’m already exploring the limits of the tool there and you might be able to save yourself a few journeys of discovery.
On the other hand: isn’t the journey the destination?
Gibt es bereits Pläne für die Weiterentwicklungen von LeanScope? Was habt ihr noch damit vor?
Yes, “more than ever”. We want to take LeanScope to the next level and are currently talking to investors to accelerate development. The traction is there, the product-market fit is there, the time is right. Now it’s time to shift up a gear!
What exactly we have in mind is of course still partly confidential, but I can already say this much: it’s worth following LeanScope AI on LinkedIn, because that’s where the important information will appear as soon as possible.
Dear Thomas, thank you very much for the interesting interview and the many insights you were able to give us!
