AI is everywhere and is becoming more prevalent in our lives both at work and at home. According to a recent report from the Capgemini Research Institute, Emotional Intelligence – the essential skillset for the age of AI, EI will be a must-have skill in the future, with demand likely to rise six-fold within the next five years. The report offers a country-based analysis regarding executives’ beliefs on EI in the workplace with 74% of executives believing that EI will become a “must-have” skill. The report also focusses on the current statistics regarding EI in employee assessment process and the potential benefits for promoting EI skills during staff recruitment and training.
I personally think AI will change the job profile of companies, many of the traditional roles have already been automated and with more sophisticated AI, more roles will be done by machines that are supplementing human intelligence and helping humans to evolve their job skills. According to the Future of Jobs report by the World Economic Forum, by 2022, more than 50% employees, globally, will require significant re- and upskilling of six months or longer. The graph below shows the projections for 2022 from the World Economic Forum.
The type of employee is changing also. Recent studies by Instant Offices show 35% of the global workforce will be millennials by 2020 and given that 72% of millennials will leave their job within 5 years, are we meeting the needs of our workforce today? What are some of the challenges?
- Can we adequately and efficiently reskill our workforce?
- What are the new areas of opportunity for employment that do not currently exist?
- Can we cope with the change in attitudes in our workplaces with more creative and innovative staff?
AI can help with these challenges and be used to make our workforces more emotionally aware. Some of the potential applications would be, what is the best new role for an individual based on their experience? How can we ensure the most efficient and slick retraining programs? How can we make a more personalised experienced for our employees as well as our customers? My research corroborates the Capgemini research that EI will be the must have skill in the next 1-5 years.
The second consideration of EI in the context of AI, is the question of how do we make AI emotionally aware & should we?
Given the recent customer trend for more personalised experiences, we could easily stipulate this trend moving into our emotions, considering not only what customers want but also understanding how they feel in that moment and modify the customer journey based on their feelings. When we decide on a recommendation based on a customer query, we anticipate a set of feelings and thoughts which govern that behaviour and the actions we take. Behind these actions are thousands of emotionally aware judgements we make.
Currently there are two ways we can learn in AI, the first is using known outcomes to train a model that finds patterns and trends in data to give the best result given the data input (method 1). The second method involves observing our environment and making decisions accordingly. The outcomes of these decisions teach us how to make better decisions and so on (method 2). This is the way humans learn. This flexibility enables us to respond to new stimuli and make novel decisions.
If we use method 1 for training our AI, the AI is not emotionally aware as it is not understanding the emotion of the customer and acting accordingly but applying the emotional intelligence of previous human agents to a similar problem. Even if we train the AI to recognised emotions such as anger or happiness, we are labelling these emotions for the machine and it is learning our interpretations of these emotions which is not based on the data it is receiving. If we use method 2, until the agent has had enough experience to have learnt effectively, it would be like talking to a child and we do not employ untrained staff for this reason.
A better approach is to combine methods 1 and 2. Build an AI agent to use current outcomes and labelled examples and then as more data is collected, allow the AI agent to learn new patterns on its own. The AI doesn’t need to know which responses are from angry people, it will associate all the similar responses together and call it whatever it likes. The AI then offers bespoke solutions based on similar behaviour within this profile and then learns from the responses and records feedback, improving the outcomes each time.
Now that we have an emotionally aware AI engine, how is that better than what we have today?
There are the universal benefits AI brings such as consistency, repeatability and scale but also, we begin to understand empirically the role of emotions in customer interactions. We then can very quickly change the offer to the customer if they change their emotion mid correspondence and we can try completely novel solutions and observe the emotional response to this. We have built a flexible and stable agent that can handle complex customers and understand how they feel.
Going back to our first impact of AI on the workplace, what does our previous human agent do when replaced by a non-human agent. They can either retrain to manage the AI, spend more time innovating solutions for the business or be available if the customer wishes to speak to a human agent. Perhaps then the most important part of an AI agent like we discussed, would be the ability to know when to revert to human, based on how the correspondence is progressing. After all, if the context is highly emotional people prefer to talk to a human or are we simply in a transition period until we can no longer discriminate between AI and human?
Jonathan Kirk, Data Scientist, I&D Insight Generation
Jonathan is a Data Scientist in the Insights & Data practice in the UK. He studied Biology (MSci) at the University of Nottingham, specialising in Behavioural Ecology before taking up a career in data science. He has over four years’ experience, in both the public and private sector, delivering solutions to solve key business problems using both well-known and bespoke statistical and machine learning techniques.