AI for the automotive industry

In mid 2025 we were approached by Vladislav Smolyanoy (now CEO of The Vienna AI Company) who back then had already demoed an early version of his vision of LISA, a voice agent for car dealerships, to stakeholders in the field. He introduced us to the idea of a comprehensive suite of AI tools for car dealerships and told us about who's already on board and where this project could potentially be going. Having seen our work on 0cred, Vlad wanted to work with us to turn his vision into a reality. We happily accepted the offer and got to work.

Today, The Vienna AI Company is the fastest-growing startup in the AI automotive industry in Europe, and we helped the team raise two rounds along that journey.

LISA was never a one-feature project. From day one it was a full suite of AI tools for car dealerships: chat, transcription, phone agents, workflow automation, integrations and more. The scope moved fast, so we had to break everything down into small, shippable blocks.

We worked very closely in a small core team, shipped weekly, tested weekly, and kept adjusting in real time. We also took this on at a relatively low rate and paired it with an equity deal because we believed in the upside and wanted to build for the long term.

We went through many rounds of design, messaging, dashboards, but also core product functionality. Some versions were too flashy, some too technical, and some just not right for dealership owners, so we kept refining until the story, interface UX and AI functionality finally clicked together.

The client really wanted an Apple Intelligence-style direction and came to us with a pre-designed logo and some other branding directions. We were admittedly not the biggest fan of those from the start (we didn't think it would work with the mostly male car dealership audience and thought it looked pretty tacky) but as requested we gave it a shot. So we explored colorful pinkish gradients with glassy layers.

After we assessed the client's needs and preferences, we started with super rough wireframes and implemented the first glassy version of both the dashboard and the landing page to test that direction in a real product context.

In the first few weeks of a project we like to work closely with our clients and be in constant communication either in Figma async or on a call, to get a feel for the product and the direction they want to go in. This can result in lots of comments (as visible in the image on top).

In parallel, we also had to design the parent-company brand for The Vienna AI Company. For that logo, we explored many different Vienna landmarks (e.g. Stephansdom, Hofburg, ...) but eventually ended up picking the Vienna State Opera as the core reference and translated it into a mark that fits the rest of the ecosystem.

After testing the first direction and talking to customers, we shifted to a style that felt more familiar in the Austrian car dealership space and easier to trust for a traditional audience.

At this point, our Figma file had grown into a behemoth of a file with lots of different screens and components. (It still looks like that btw, it REALLY needs a cleanup but we sadly haven't gotten to it, too many things going on that are all more important than a figma cleanup).

The new direction looked tame and boring, but this was on purpose. Our clients wanted something that looked and felt familiar. It has a clear CTA, carousels, and simple to understand illustrations with almost no visual overload.

For most of the project, the core team was just Philipp (our founder), Semere (one of our devs), and Vladislav (founder of The Vienna AI Company). Philipp and Semere were both coding and coordinating execution, while Vladislav was focused on sales and feeding his own, the investors' and our first customers' feature requests to us. We mapped features, jumped contexts constantly, pushed back on scope creep and kept shipping under pressure. We also did this at lower rates with a vested equity deal.

If that kind of setup is interesting to you, feel free to reach out to us, we occassionally take on projects like this, if we see the vision.

From the get-go we already had real customers, so we built with constant customer feedback directly in mind and on the ticket list. That made this project harder than the usual startup scenario of building for a future customer base, because alongside shipping new features we also had to handle bug reports, fixes, and rapid change requests in real time.

The third and final direction was blue again but cleaner. Vladislav (our client) came up with the idea to integrate LISA as an avatar into the product because the target audience clearly wanted to "see AI". We combined that with the blue wheel logo, introduced orbs, and removed most of the glass to land on a cleaner dashboard direction for LISA.

This redesign was meant to be more attractive to investors and younger people, having learnt that dealerships usually convert through other channels anyway. It also felt more modern and clean, which is what the target audience wanted.

The avatar became a central branding anchor for us. It gave the product a recognizable face, made the AI angle immediately obvious, and helped us keep the visuals modern without making the interface harder to use for the core dealership audience.

Our new hero animation showcases some core tasks that LISA can automate. This is great to showcase most of LISA's core features at a glance. Building these types of animated heroes that tell a story is an adler.studio staple, we've done it for many other projects and could also do it for you.

Up to this point, most visible updates were on the landing page. The core product, including the AI web dashboard, call system, and other core flows, was still running on the older pink glass UI. That gap had to be closed, so the next phase was bringing the actual product experience in line with the new brand direction.

Branding was only one part of the job. The product side was heavy too: many connected modules, lots of edge cases, and different user journeys that all had to work together. And the main problem: the product to this day still isn't set in stone, it's constantly evolving and built iteratively. Our team helped with planning and scoping out features as well as actually building them.

A good example of this complexity is the automation builder. At a glance it looks similar to n8n-style workflows, but it is fully proprietary and much more on-rails for dealership-specific operations. No n8n was used.

Disclaimer: for this project we can only share a limited part of the stack publicly. We can not go into much detail about the AI tech involved.

The initial phase of LISA had a super hard deadline: the product had to be live-demoed in November in front of important industry stakeholders, roughly five months after project kickoff. For AI products that is always a little scary, because live demos can break in unexpected ways, due to the undeterministic nature of AI. Pre demo day we tightened UX, cleaned up visual hierarchy, prepared the core flows, and nailed the demo. Since then, it has only been going up.

After Demo Day and into the new year, we stayed close and kept shipping with the team. This became an ongoing product partnership, with Philipp Parzer moving away from the development role and more into supporting as fractional CTO / Head of Product across roadmap, delivery, and putting the product decisions coming from C-Suite, investors and customers on a timeline.

Since then, another round has been raised off the back of strong execution from both our development team and the founding team.