Digital companions – what’s the value and how do we build them?

Today’s digital assistants would struggle with these complex questions. They can deal with clear requests – ‘play Imagine by John Lennon’, ‘will it rain tonight’. But they cannot provide good advice tailored to your personal context.

At least, not yet.

Innovators are turning their attention to the next generation of digital assistants, what we might call Digital Companions or Digital Coaches. These will take nuanced requests, understand your personal context, map it to their training on complex scenarios, and give out good life advice.

They may help tailor your fitness regime, organise your work schedule, help you change negative thinking, or give you dating advice (might we one day see a digital Cyrano de Bergerac?). These may be apps on phones, or avatars in the metaverse.

These companions will help and support you. The goal is not – as some technologists predict – to automate away human thinking, but rather to provide useful advice that helps you make better, more informed decisions about your life.

This level of sophistication is becoming a very real possibility. Both tech and non-tech companies should start considering the business opportunities they present.

What’s the value of digital companions?

The first question to ask of any new technology opportunity is ‘why should we do this’?

People are desperate for good advice in everything from physical and mental health, to work and dating. Since there is no ‘average’ person, generalised online advice is always imperfect. Meanwhile, professionals (sports coaches, nutritionists, councillors, life coaches), can be pricey, limited in availability, and with no guarantees of quality.

Digital companions offer alternatives. They are not a direct replacement for all human interaction, but they have benefits; they offer advice on demand, without judgement, and – if you share the right data – may know more about you that a human expert.

A fitness coach could create a tailored training plan, push you to train at the right time, tell you to speed up and slow down as your vital signs change, and advise you when and what to eat to convert fat to muscle based on your metabolism. While some of these fitness services exist, they do so in silos.

Digital psychologists may be a way off, but simulated sessions with a digital counsellor, based on structured programmes such as cognitive behavioural therapy or addiction treatments, are a possibility. Apps which nudge behaviour using such technique are already on the market but could become far more contextually aware, personalised, and interactive.

This is a need people will pay for, and therefore a big business opportunity. Millions of consumers happily part with $30+ per month for health and fitness services. Mental health programmes represent a large cost for health services and insurers rely on pool of professionals that falls short of demand. The UK’s National Health Service, one of the world’s biggest buyers of HealthTech, now has more than a million people waiting for such support. Scalable, cost-effective alternatives could be attractive.

What about the business model? Some will be subscription services. Others may be given away in return for data that allows you – or others – to build valuable products, such as personalised therapies. They may be developed as standalone products by tech companies, or complementary to existing products – sportswear, health foods, dating apps – creating both a new revenue stream and a cross-selling platform.

How do we build digital assistants?

Given the complexities, how can we start to create these personalised companions?

Although we may one day see companions that do everything, in the short term the key will be to constrain them by providing a clear remit on what goals they can advise on: Fitness, diet, lifestyle, etc.

This allows you to clearly define the product: what data it needs, what technologies and software, what training and expertise. Such companions are not a single technology, but a carefully designed and integrated stack of interconnected hardware and software.

As you start on this journey, here are some key things to consider.

Identifying the data you need. What do you need to know about the user to provide meaningful advice? A fitness coach may need to know heartrate, nutrition, sleep, daily schedule. A life coach may need mood, routines, friendships.

Once parameters are identified, you need to find ways to obtain or derive that data, which are acceptable to the user. Smart watch sensors can already tell us heart rates, and could soon do blood pressure too. New connected devices can add context – Lumen collects breath samples to calculate metabolism, for example. Facial recognition and computer vision can monitor how users respond to an avatar’s advice. Heart rate, affective analysis, and skin temperature can track mood. Social media profiles can provide signals about behaviour, interests and relationships.

Gather deeper insights from data. We will need more than direct measurements. With the right algorithms, combinations of signals can create a richer picture. Patterns of movement or heartrate can tell us whether the user is running, swimming of cycling; or if they are going through a tough time.

Such algorithms are hard to get right – not only do they need to discern complex (good) signals from noise (unwanted signals), but in order to work for all people, they need to be able to do that regardless of whether the user is sweaty, tired, ill, and work for all body types, skin tones, and cultures. Developing these algorithms needs not just technical expertise, but subject matter experts to ensure human factors is considered in design.

Use expert data to train the companion. The companion then needs to match the user’s data profile to the optimal advice. Creating that will require AI algorithms that have been trained on the breadth of human knowledge and real-world data in the field in question (nutrition, CBT etc), and match the most relevant data to the user’s personal goals and context. This of course is no easy feat.

Again, this needs both AI experts and subject matter experts – in health, fitness, psychology, and so on – to work together to identify the right training materials, and then build and train a functioning companion.

Validate them. These devices will also need rigorous testing before launch, to ensure they perform in a wide range of environments, especially if they operate in regulated areas like mental health where stringent clinical studies must be performed before efficacy and safety are confirmed.

Plan for privacy and data security. The data on which a companion’s recommendation may be hacked or reverse engineered to discover the underlying personal information, so it’s important to consider how access is protected. While consumers are demanding better integration, a seamless experience, and better intelligence and insights, these benefits are only achieved with additional amounts of personal data, thus increasing the risk of exposure or data manipulation. Unwanted data exposure could impact behavior of the digital coach (giving bad information), or give leverage to unscrupulous hackers.

These are just two examples of the potential legal and reputational risks if the sensitive data they use is lost, stolen, misused, or revealed. In creating these companions, the full range of data security considerations must be considered and built in from the start.

Build the infrastructure around it. Finally we need the infrastructure to support these companions in the real world: cloud hosting, analytics services, data management, data security, customer support. These companions will process a lot of data, much of it highly sensitive, so getting the backend right is just as important as the product itself.

The path to digital companions

The first generation of digital companions will go beyond the current ‘assistants’, and provide tailored guidance across different aspects of our lives; organising our schedules, keeping us physically and mentally healthy, finding activities we like, and giving us daily advice to live life to the best. The journey will be complex, bringing together wide-ranging technology and expertise into one place. There may be big rewards for those that get it right.

---

How we can help

Capgemini can support companies to plan and develop bespoke digital companions, through access to expertise in software, data, algorithms, AI, security, sensor selection, integration, curating subject matter expertise, scaling, and lifecycle management.