Skip to content

Share

  • Add this article to your LinkedIn page
  • Add this article to your Twitter feed
  • Add this article to your Facebook page
  • Email this article
  • View or print a PDF of this page
  • Share further
  • Add this article to your Pinterest board
  • Add this article to your Google page
  • Share this article on Reddit
  • Share this article on StumbleUpon
  • Bookmark this page
PA OPINION

EmotiBots – is AI ready to take on our emotions after COVID-19?

COVID-19 has created a physical health crisis. But it’s also creating a mental health one, with the World Health Organization (WHO) predicting an upsurge in the number and severity of mental illnesses around the world. That will have a major impact on organisations, especially when you consider the economic cost of stress, depression and anxiety (£30 billion in the UK in 2019, according to health insurance firm Vitality).

On the other side of the coin, when organisations get mental health support right, productivity rises. Studies by The University of Oxford and The University of Warwick revealed that happy employees are 13 per cent and 12 per cent more productive, respectively. As a result, organisations are increasingly looking to improve their people’s wellbeing. What was once a nice-to-have initiative has become a strategic and moral imperative.

But getting the right support to people at the right time, for an economically viable investment, is tough. This is where artificial intelligence (AI) can help. To leverage AI effectively, you need to understand how you want to support your people in the future, what AI can do today and what your business case is for wellbeing technology.

What future do you want?

As humans, we struggle to predict the future beyond immediate cause and effect – who would have imagined in November 2019 that a world-changing pandemic would emerge in just a few weeks? Yet we can anticipate possible futures (such as the widespread adoption of remote working) and work towards our desired scenarios if we take a methodical approach.

The first step in understanding your wellbeing AI journey is to define the future you want. We use our FutureWorlds™ methodology with global organisations to identify uncertainties and map the various ways they could shape the future so leaders can act today to navigate towards their preferred future.

1.   Identify uncertainties

While we’re not good at predicting the future, we are good at recognising uncertainty. For FutureWorlds™, you need to identify two key uncertainties – if there are more you want to explore, it’s important to run separate exercises with them in pairs. For example, when looking at the possibilities of wellbeing AI, we could look at the uncertainties around the capabilities of AI and the level of organisational support people will want.

2.   Map the possible futures

With two uncertainties in mind, you then map their extremes onto an X and Y axis to create a matrix of four quadrants, as in the image below. Each quadrant represents an extreme possible future, so write a few words about what the two axes mean for each future. For example, in a world with advanced AI and where we rely on self-service wellbeing support (the ‘buddy world’ below), people would have an ‘AI friend’ that helps them overcome challenges and avoid future crises.

3.   Identify your preferred future

With clear possibilities mapped based on current uncertainties, you’ll be able to build a picture of what would work for your organisation and build a desired future state. You can choose elements of more than one world and, of course, there are many other factors to consider. But when you’ve defined where you want to be, you’ll be able to build a strategy for getting there.

What can AI really do?

With an agreed future for wellbeing support in place, you need to know what technology is out there today and what it can do. This will help you decide what you can do in the short-term and what will take longer to achieve.

Chatbots

Today, there are numerous AI-powered chatbots that either link to resources or offer active emotional support. For example, we’ve developed a wellbeing chatbot that links to a learning platform of resources. That means people can ask what they can do about a stressful situation and the bot links them to the right content or person.

Other chatbots focus on direct emotional support. For example, Woebot and Replika provide coaching, counselling and/or friendship, recognising the intrinsic link between the human need to connect and mental wellbeing.

Intelligent wellbeing analytics

AI already has significant analytical powers, uncovering insights in vast data pools across industries. When it comes to wellbeing, our chatbot analyses what resources people are looking for the most, highlighting patterns of emotional needs so leaders can develop more targeted responses.

Soon, AI analytics could also be a valuable diagnostic tool. The Alan Turing Institute is researching how AI tools could personalise mental health profiles and advance the precision of early diagnosis.

Emotional augmentation

At present, AI doesn’t have the socioemotional capabilities to replace people in anything that involves complex emotion. However, ‘whisper agents’ are helping people better connect to others. For example, Cogito acts as a co-bot (collaborative robot) in contact centres, spotting signals from callers and providing guidance to customer service agents in real time. It also tracks employee stress levels and can suggest when people should be taking a break.

Calculate cost-benefit for scenarios

For any desired future, it’s important to explore the cost-to-benefit ratios of investment. As an example, let’s look at investing towards ‘Independence World’, where a chatbot can link people to relevant resources.

Imagine the leaders of an organisation of 5,000 people want to invest in a simple chatbot. According to the UK’s Office for National Statistics, the average number of sick days an employee takes per year is 4.4 and the average daily wage is £117. This means the organisation will lose an average of 22,000 days to sickness per year, at a cost of £2,574,000.

Now imagine they could design and launch a wellbeing chatbot for £100,000 and predict it would cut their average number of sick days to 3.4 per person, per year. If this was successful, the organisation would reduce their sick days to 17,000 and save £585,000 per year.

While there are many factors to consider, like the improvements in productivity, resilience, creativity, team working and physical health that good mental health support brings, this simple example highlights the impact investing in wellbeing AI could have.

Proceed with caution and hope

AI is an incredible tool with immediate potential, although it isn’t going to replace trained wellbeing practitioners anytime soon. The accelerating advancement of wellbeing AI technology and its increasing use show it can add real value to a workplace today, potentially mitigating the high financial and emotional costs of mental health challenges. And while it’s easy to make a good business case for investing in targeted wellbeing AI, prioritising people’s mental health is also the right thing to do.

Contact the author

  • Heledd Straker

    Heledd Straker

    PA future of work expert

    Heledd is a workforce futures expert and design thinker who helps clients think completely differently about their people

    Insights by Heledd Straker

Contact the people and talent team

×

By using this website, you accept the use of cookies. For more information on how to manage cookies, please read our privacy policy.