Our approach
Technology that adapts to people’s lives
We believe in discreet, useful and respectful technology that supports autonomy and makes everyday life simpler for those who provide care.
Our guiding principle
Understand the context, intervene with restraint
Our solutions start from real life, respect each person’s rhythms and step in only when needed, with support that is simple, clear and non-intrusive.
-
Person-centredness
We design with and for people, starting from their real needs.
-
Privacy by design
Privacy is a technical and design requirement from the very first prototype.
-
Responsible innovation
An ethical, transparent approach, with clearly stated limits.
-
Practical usefulness
Solutions designed for everyday life and tested in real environments.
The research that guides us
Knowledge comes before technology
We develop our solutions in dialogue with international research on assistive technologies, on artificial intelligence applied to everyday life, on the autonomy of people with cognitive vulnerabilities and on smart home environments.
The body of work we have reviewed spans more than twenty years of publications, from the early 2000s to 2025.
- 2001–2008 Canada
Intelligent cognitive assistance systems
Research on systems that interpret daily activities and provide contextual prompts to people with cognitive difficulties.
Reported result
An experimental study published in 2008, involving six participants, reported for the subgroup with moderate dementia an average increase of around 11% in steps completed independently and an average reduction of around 60% in the interactions needed with the caregiver during the task studied, handwashing.
These results are limited to a small study and cannot automatically be generalised to other populations or technologies.
- 2014 Germany
Assistance with daily activities
Research on systems that assist people with cognitive disabilities in their daily activities, which explored:
- breaking activities down into understandable sequences;
- recognising the stages of a task;
- graduated prompts;
- support adapted to individual abilities;
- interventions during household and personal care activities.
- 2017–2018 United Kingdom
Personalised smart homes
Research on designing smart homes tailored to people with cognitive difficulties. Topics include:
- adapting technology to individual characteristics;
- involving caregivers in defining requirements;
- understanding household habits;
- person-centred design;
- integrating sensors, environment and care.
- 2025 United Kingdom
Home observation with sensors and machine learning
Research on observing behaviour at home through sensors, data analysis and machine learning, with longitudinal observation of up to two years in some studies. Areas of interest include:
- identifying routines;
- recognising changes in behaviour;
- analysing sequences of activities;
- interpreting data collected over time;
- assessing the limits and reliability of the observations.
- 2014–2021 Europe
Research programmes on assistive technologies
The European Active and Assisted Living programme involved a total of 21 countries over its cycle of activity, according to the final evaluation published by the European Commission in 2022. It is a useful reference for understanding how technologies for personal autonomy, home life and caregiver support have evolved.
These are scientific references that we have reviewed and that inform our design thinking. Pantoufle did not take part in these studies, and their results do not concern Vesper.
The research in numbers
Data to understand, not to promise
A few figures help to situate the research. They refer to the studies cited, not to results achieved by Pantoufle or Vesper.
- 2001–2025 Time span
of the scientific references reviewed, from the first cognitive assistance systems to recent research on home observation.
- 3 Countries
Canada, Germany and the United Kingdom: where the main research strands documented here come from.
- 21 Countries
took part in the European Active and Assisted Living programme, according to the evaluation published in 2022.
- 6 Participants
in the 2008 Canadian study on cognitive support during daily activities: a very small sample.
- 11% Independent steps
average increase reported for the subgroup with moderate dementia in the same study, during the experimental task.
- 60% Caregiver interactions
average reduction reported for the same subgroup and the same experimental task.
- Up to 2 years Observation
longitudinal observation period considered in some British studies on behaviour at home.
From listening to skills
Before we teach technology, we listen to people
Design starts from an analysis of real needs. Through interviews with families, caregivers and organisations, we gather information on daily activities, recurring difficulties, habits and the situations in which technological support could be useful.
The interviews are not only a way of collecting opinions about the product. Above all, they help us establish which skills artificial intelligence must have in order to provide appropriate assistance.
-
Listening
Needs, everyday difficulties, individual abilities and the forms of care already in place.
-
Analysis
The information gathered is translated into situations, activities, behaviours and support needs.
-
Defining skills
What the AI must be able to do: recognise a situation, understand an activity, notice an omission, make a suggestion, call a caregiver.
-
Design
The skills become functional requirements, behaviour models and possible automations.
Automation and contextual intelligence
Recognising an event is not enough
Its context must be understood
We design systems that combine sensors, artificial intelligence and automation to provide support in daily activities.
The approach goes beyond traditional home automation based on fixed rules: the aim is to interpret information from the environment, compare it with the person’s routines and assess whether it is appropriate to intervene.
-
Observation
Collecting only the information needed, through ambient sensors and integrated devices.
-
Interpretation
Reading the information in the light of habits, activities and behaviour sequences.
-
Assessment
Distinguishing situations in which help is useful from normal or irrelevant variations.
-
Intervention
Voice prompts, contextual guidance or notifications to the caregiver, as needed.
-
Follow-up
Observing the response and assessing whether further support is needed.
Deciding not to intervene is also an important design choice.
An assistive system should not multiply notifications, constantly interrupt activities or reduce the person’s autonomy. It must strike a balance between support, discretion and respect for individual abilities.
Technology architecture
Artificial intelligence, close to people
Physically, too
Vesper is based on a computing device installed in the home, which integrates ambient sensors and artificial intelligence models processed locally.
The system provides concise, filtered notifications for caregivers, instead of continuous video supervision.
Three aims behind this choice
-
Confidentiality
Avoiding the continuous transmission of audio and video streams to external servers.
-
Operational continuity
Allowing the main functions to run locally even with a limited connection.
-
Responsiveness
Reducing reliance on remote processing for assistive functions in the home.
Documented technology status
A working prototype is not a product validated for general use. Here is where we stand.
- Home proof of concept - Vesper
- Completed
- Local artificial intelligence processing
- Working
- Detection of deviations from routines
- Working in the prototype
- Contextual voice interventions
- Working in the prototype
- Caregiver app
- In development
- Pilot with external families
- To be launched as planned
- Independent validation
- Not yet completed
We do not promise absolute safety, infallible detection or the ability to prevent every incident: an assistive system supports people; it does not replace them.
Pantoufle’s method can be applied to a range of assistive needs. Vesper’s initial operational plan focuses on young people with cognitive disabilities who live with their families.
How we work
From research to everyday life
A five-stage process, repeated with every improvement: what we learn in the field feeds back into research and into the definition of skills.
-
Research and listening
Scientific studies, needs analysis and interviews.
-
Defining skills
Translating needs into capabilities required of artificial intelligence.
-
Technology design
Integrating hardware, software, sensors and automation.
-
Trials
Technical testing and progressive evaluation in real environments.
-
Continuous improvement
Analysing results, revising skills and updating solutions.
Trials and measurement
Usefulness is not declared
It is demonstrated
We follow a progressive path that distinguishes research, prototyping, trials and validation. Showing that a technology works is not enough: we need to understand whether it can actually be used, consistently, in everyday home life.
-
Concept
Defining needs, objectives and requirements.
-
Working prototype
Technical testing under controlled conditions.
-
Trials
Testing with people, families and caregivers in real settings, according to defined procedures.
-
External validation
Independent assessment with documented results.
Indicators set out in Vesper’s validation plan
- Monthly hours of support required for each active installation.
- Actual device set-up time and the scope for standardising the procedure.
- Actual monthly operating cost per installation.
The plan provides for at least five completed pilots as the first stage of operational testing.
These are planned objectives, not results already achieved.
Other dimensions of evaluation
These are methodological proposals still to be developed, not indicators that have already been measured.