Designing Tessa, a social robot for people with dementia

01Supporting people with dementia through design

In 2015, I co-founded Tinybots Tinybots changed their name to Tessa Care in 2026. Source Tinybots changed their name to Tessa Care in 2026. Source to build Tessa: a small, talking robot that supports people with dementia and other cognitive impairments. Family members and healthcare professionals write and schedule friendly spoken reminders through an app, and Tessa speaks them at the right moment, connected to the internet at the person’s own home.

We founded Tinybots on a design philosophy built around three deeper human needs: autonomy, connectedness, and competence. Rather than replacing human care, the goal was to increase independence for people living with dementia, autism, and other cognitive impairments.

02The path to founding a company

Before founding the company, we were studying human-robot interaction. Polygon was one of my research projects at VU University Amsterdam, testing whether a social robot needs a realistic, human-like form to be trusted at all.

There, I found that an abstract, distinctly non-human robot could still be accepted by people with acquired brain injury — abstraction seemed to help acceptance, if anything, rather than hurt it. It wasn’t about how real Polygon looked, but what people believed it could do for them.

That became the specific design philosophy I carried into Tessa: an intentionally simple, non-human form, left a little to the imagination rather than spelling out exactly what it could and couldn’t do, and simple enough that nobody expected more from it than it could deliver.

Polygon’s early wooden prototypes read as warmer and more approachable than a sterile, plastic robot would have, the same instinct that later shaped Tessa’s own wood and felt, materials with a long history of signaling warmth and care rather than clinical function. Joseph Beuys, the German contemporary artist, was one of my main inspirations for Tessa’s look and feel. He used felt throughout his work, from wearable pieces to entire rooms, as a material he associated with warmth, protection, and healing. Source Joseph Beuys, the German contemporary artist, was one of my main inspirations for Tessa’s look and feel. He used felt throughout his work, from wearable pieces to entire rooms, as a material he associated with warmth, protection, and healing. Source

Polygon’s focus groups brought two insights. The first was independence: people said they’d rather ask a machine for help than keep asking the people around them. The second was loneliness, a desire to talk to someone who wasn’t already family or care staff. Extending that same question, and those same needs, to people with dementia specifically, in a home rather than a lab, was enough for my co-founder and me to sense a real opportunity.

Back in 2014, before Tinybots existed as a company, I made a first napkin sketch to map out the idea. It wasn’t particularly beautiful, but the core concept has stayed unchanged to this day: a physical, creature-like device that gives verbal reminders, scheduled through a calendar-like app.

In October 2015, we founded Tinybots to find out whether we could build something that would meaningfully improve the lives of people with dementia.

03Iterating on Tessa’s embodiment

What followed wasn’t one prototype evolving step by step into the next, but several embodiment directions explored in parallel, each brought along and tested directly with residents and staff at retirement homes and healthcare facilities, then revised, then brought back again. Each round said less about whether Tessa worked at all and more about where a given direction still felt wrong, a shape too clinical, a material too cold, a gesture too robotic.

Based on that feedback, the embodiment was altered toward something brighter and more approachable. Wood and felt, mass-produced but still quirky enough to read as a character rather than a device, let people with dementia recognize Tessa as an individual without triggering the preconceptions a more clinical or humanoid form would have carried.

That work settled into a first version worth testing in the real world, but getting there meant solving a different kind of problem: turning a one-off, hand-built prototype into something that could be produced reliably, not once, but 120 times over, for a closed beta with four healthcare partners.

04Scaling hardware in the real world is difficult

While the hardware scaled from a single prototype to 120 units, Scaling hardware isn’t only about building more of it. A one-off prototype can be hand-tuned and supported individually; at 120 units, installation and support had to work identically across every single one, since that also determines how the service around it can scale. Scaling hardware isn’t only about building more of it. A one-off prototype can be hand-tuned and supported individually; at 120 units, installation and support had to work identically across every single one, since that also determines how the service around it can scale. the app running on it still had to be designed and built in parallel.

The first version was simple: text-to-speech reminders, scheduled remotely. That initially looked like a simple scope, but that includes an account registration process, a hardware pairing service, and connecting to WiFi. Those one-way reminders eventually grew into full branching conversations, Tessa listening and responding, not just speaking on a schedule.

The app itself started from a small set of basic flows: setup and registration to get a new Tessa connected to an account, and a simple way to plan and schedule reminders once she was already up and running. On paper, neither flow looked especially complicated.

The initial closed beta ran for twelve months, which revealed a major bottleneck in scaling. Our initial installation flow was overly complicated and getting the system set up and running in the first place was hurting adoption. This led to dozens of visits, Tessas not being installed, and on-site support during the closed beta.

The whole flow was mapped and reworked, and what had been a nine-step setup process came down to four. Everything had to work together to make the experience work: frontend, backend, the physical hardware, and the local WiFi network, all of it just to get Tessa online, registered, and securely accessible from a single account. Only then could our intended end-users start using Tessa.

As a final step, we changed Tessa’s eyes to reflect the step of the installation process she was in, a detail also reflected in the installation manual. Tessa starts as a WiFi access point, then takes your WiFi details once your phone connects, downloads her updates, and finally registers to your account.

05Creating a visual language to match Tessa’s embodiment

I created the original logomark and logotype back in 2015, built on a variant of a typeface I’d already designed called Polygon. That typeface was originally shaped to match Polygon’s own embodiment design, and I later adapted it again to match Tessa’s.

To keep the frontend consistent as the team grew, that same visual language became the Tinybots Design System: a Sketch component library standardizing color, typography, and roughly 80 components across every interface Tessa touched, the app, the administrative dashboard we developed later, and Tessa’s own WiFi setup screens.

The same principles carried onto physical materials too: the packaging, the quickstart guide, and the printed manual that walked someone through installation and their first scheduled reminder.

A few principles sat underneath all of it, the same whether an interaction happened on screen, in Tessa’s voice, or on a printed page.

06Moving from a closed beta to producing Tessa at scale

At beta scale, every Tessa was still finished largely by hand, refined unit by unit to fit its internal components properly. Getting from there to a fleet in the hundreds meant standardizing the assembly itself, not just the product it produced, robots leaving the line identical and ready for any customer, rather than built one at a time to order.

Reaching a first market-ready version took roughly two and a half years from founding the company: sourcing suppliers so parts didn’t all have to be finished by hand, and moving key components to injection molding to make assembly repeatable. But by 2019 it had shipped over 500 market versions, a number that kept growing after.

Scaling production also meant raising money along the way, several investment rounds I can’t say much more about here, but which made this stretch possible at all.

That shift, generic inventory instead of made-to-order, is what the administrative dashboard was built to track once production reached that scale.

07A short history of Tessa and the things we shipped

Over the years, we’ve improved Tessa a lot since founding Tinybots, from the embodiment and hardware itself to our own speech recognition.

08Looking back, the hardest part wasn’t the robot itself

What’s hard is doing any of it reliably at scale: it’s easy to make one of something, it’s much harder to make a dozen of the same thing, and it’s nearly impossible to go into the hundreds, especially once you combine a physical product, an app and platform, with a subscription model, hardware registration, contract management, and a support system that has to hold up in someone’s actual home.

By the time I left in 2019, more than 500 Tessas were already in homes and care sites across the Netherlands. Tinybots, since renamed Tessa Care, now also offers a newer product called Tessa Screen alongside the original Tessa Voice — developments after my time there. Source Tinybots, since renamed Tessa Care, now also offers a newer product called Tessa Screen alongside the original Tessa Voice — developments after my time there. Source That number kept growing after I left, but the core of what made it work was already in place.

A few years later, independent, peer-reviewed research backed up that original premise: a 2022 study found Tessa helped people with executive dysfunction in disability care complete daily tasks and feel more independent, close to the exact outcome we’d designed for. Van Dam et al., “Experiences of Persons With Executive Dysfunction in Disability Care Using a Social Robot to Execute Daily Tasks and Increase the Feeling of Independence,” JMIR Rehabilitation and Assistive Technology, 2022. An independent study on the efficacy of Tessa. Source Van Dam et al., “Experiences of Persons With Executive Dysfunction in Disability Care Using a Social Robot to Execute Daily Tasks and Increase the Feeling of Independence,” JMIR Rehabilitation and Assistive Technology, 2022. An independent study on the efficacy of Tessa. Source