You can teach bots to be compassionate with customers, Anthrolytics’ Peter Dorrington explains how

By Jim James, Founder EASTWEST PR and Host of The UnNoticed Podcast. 

Anthrolytics is a London-based company that uncovers the motivations behind human behaviour — why people do the things that they do and what they’re likely to do next. According to co-founder and Chief Strategy Officer Peter Dorrington, they do this by looking at emotional motivation and adding that to other kinds of motivators. Why is somebody buying a particular product? Apart from factors such as price, convenience, and purpose, they ask: How do the customers feel about the products, and how does that work over the lifespan of them being a customer?

At Anthrolytics, data science is combined with behavioural science to answer that question. However, they don’t only identify why people do what they do — they ultimately help businesses make better experience decisions, particularly when it comes to their digital channels where there isn’t a human at the one end of the conversation. In sum, they turn motivation into action.

Image from LinkedIn

How they Help Bridge the Digital Divide

With COVID, the past 14 or 15 months have seen more businesses moving into a digital operating model. And this means that they often don’t have a human talking to another human being. Many of the consumers have learned to self-serve and businesses have relied on bots. For most organisations, this is a very satisfactory, competent job: It does what they want them to do. However, there is this disintermediation — and it has really affected the experience economy. 

Customer experience, as a discipline, has been around for about 25 years. It’s not something new but what businesses have done are the easy aspects of it, e.g., listening to customer programs, doing customer journey mapping. However, what customers are saying is that they want empathy; they want compassion. They want to be more than a customer number. They want to feel that they’re in a relationship with you. Once they feel that, research shows that business metrics will go off the chart. Customer satisfaction will leap up, and so does loyalty and economic activity. Your customers will spend more money more frequently — and they will enjoy doing it. 

What Anthrolytics is focusing on is how to replicate or produce empathic experiences in digital channels. This one is possible and it can be likened to teaching a bot how humans feel. If someone says he’s hungry, then the rational need is to eat. But if he wants a juicy burger and fries, then it becomes more emotional. Doing so will make him feel happy. However, it’s not the only emotion that the customer wants to feel. He wants to feel less afraid and more certain during uncertain times. Therefore, at the heart of human-centered design (which is where a number of businesses are moving into), there’s a need for empathy. You need to blend compassion with competence and once you’re able to do that, business results will be outstanding.

Screengrab from Anthrolytics’ website

The Process of Teaching Bots

Bots run automated tasks over the internet. But how do you teach them to be compassionate? How do you help them when they couldn’t see the person on the other end?

According to Peter, it’s a two-step process. And the first step is to use natural language processing or understanding technology to analyse what customers say. There’s a specific kind of question that they’d typically ask, i.e., asking the customer to describe something, like their experience or a podcast episode, to a friend or a stranger. When customers do that, they tell you what’s on top of their minds about the particular experience. In the process, they’re leaking emotions. The first stage involves analysing those and figuring out which things do people remember and talk about and how they feel about them.

Now, when these are incorporated into a decision matrix (where you have all other factors like price or star ratings), you can then associate an event (e.g., listening to a podcast) with an outcome. Then, you can make decisions about what outcome do you want and what’s best for both parties. So the bot has a decision-making algorithm that doesn’t only tell about the logical side of things but also the emotional aspect — if a customer is likely to be feeling anxious or uncertain or unclear about what he does. Based on that, you can choose a different tone of voice or endorse a different product for that customer, something more suited to his wants and needs. These are things that can be taught to bots in any form of automation where there is a choice.

One reason why this can be complicated is that a lot of previous models over the last few years only worked when the customer had a completely free choice. Behavioural economics talk about free choices but very few of us have such. For instance, if you want to buy an Italian supercar and you can’t afford it, it’s not a free choice. The operational part of the decision-making is to use these insights to let the system make better decisions on behalf of your business and your customers. When it does, you get something that feels a lot more humane.

The real trick is being able to do this proactively. Before somebody starts to exhibit fear or anxiety, anticipate where your customers might be right now. With that, you could reach out and offer reassurance before they have to dial in asking what they need to do. This is an incredibly human feeling and experience that everybody enjoys.

Understanding a Customer’s History

If you try to model an individual person, you’d want to understand what his history did to put him to where he currently is. When you use traditional techniques, Peter says that it’s undoable because there’s not enough data and it’d take too much computing power. 

Image from Unsplash

A customer’s history informs one’s habits and opinions. It took Peter two years to figure out how to do that but what he found out was how you could extend that across an entire customer base of tens of millions of individuals — and update it every single day — based upon what has happened to a particular customer. Rather than trying to do regression (which is building a big complicated model), his company came up with a much more straightforward technique that’s efficient and requires relatively little computing power. However, he points out that it’s not going to be 100% accurate, as in any other behavioural model. 

To make it clearer, take this as an example: You’ve probably gotten offers from a marketing department where they’ve used a predictive model. They’ve put you into a segment, let’s say, Customer Segment A1, where there are a lot of other customers. Angry customers within that segment will respond differently from happy customers. If you treat them both the same, you’ll probably irritate both of them. But, if you could tell the difference, then you could treat those customers differently according to their needs. And that’s where the slight differentiation of the golden rule comes around: You’re more focused on treating customers the way that they want to be treated — not the way that you want to be treated. After all, the customer should be at the center of your thinking.

In this sense, you’re using legacy data from that customer within your organisation rather than what they’ve done before they got to your organisation.

Everyone talks about customer journeys in customer experience. They plot the journey and see that there’s a happy journey. Then there are derivations off of that. In reality, it’s about customer landscapes because people’s lives are messy. They are not sequential or linear — people have overlapping journeys and there are detours. When you take the approach that Peter and company have taken, your history will be a bit like dead-reckoning navigation. It’s about where you are right now, what direction you’re going in, and how fast you’re traveling. 

However, it doesn’t really matter how fast you respond if you’re taking the wrong treatment or direction. It’s not necessarily that you’re making wrong decisions — you’re not just making optimal decisions. And you can make a better one. From the human point of view, it’s not about reducing choices or saying that certain people only get certain offers. It’s about offering everybody the best possible choice that meets your customers’ and your business’ needs so that everybody walks away as happy as they could given the circumstances.


If you’ve got everyone’s data and you’ve got how they behave, you’ll be able to personalise both the offer and the way you deliver it. In the digital context, the most talked-about version of that is what’s called hyper-personalisation, which adds real-time context to personalisation. Apart from knowing what a customer wants, you delve into the emotional, empathic context of that: How is the customer feeling? Because this will influence his decisions. It’s estimated that 98 to 99% of decisions people make on any given day are dictated by the subconscious and are influenced by emotions and habits — they don’t follow logical rules.

Image from Unsplash

This is one of the things that many businesses get wrong. When they’re talking to their customers, they think that getting more facts helps. For some customers, this is true. But for others, the last thing that they need is more facts. What they want is help and guidance — something that is more narrative saying what is not right for them and offering a product or service that is. If an organisation could give a customer what he needs, he’s more likely to trust that organisation even if it’s too complex to understand intellectually. 

There are tools and techniques to understand these things. However, these shouldn’t be used to do dark psychology, which is to influence people to do things that are not in their best interest. These should be used to treat customers like human beings. And if you’d be able to address their needs — both emotional and rational — it will be a better experience for them. Subsequently, it will build a better business relationship. It would also be nearly impossible for your competitors to reverse-engineer because it’s based upon a relationship that’s unique to your business. Your relationship with your customer is not something that can be replicated.

Customer Relationship in the Digital Context

Establishing a personal relationship with a customer is more difficult to do in digital environments because there’s no human-to-human relationship. 

To be able to do that, Peter talks about two important aspects of empathy, the emotional side of things. With emotional intelligence training, you can understand what people are feeling and why they’re feeling it. It’s called cognitive empathy. The compassionate aspect comes when you take action as a result of recognising your customer’s emotions. If someone is angry, what are you going to do about it? When you link cognitive empathy with compassionate, action empathy, you can display it through a machine. You don’t need to be a human being to do that — it simply entails making the right decision that reflects what the customer values.

For big businesses, this one is easy because they’d only need to automate and operationalise for it to be done to millions of customers every day. For smaller businesses, on the other hand, there are a number of ways how this can be accomplished (e.g., customer feedback). However, the first thing you need to do here is to ask the right question — instead of asking how your customers feel, ask them to describe their experience. 

As stated earlier, having them describe their experience will help leak emotions. If you have no infrastructure, you have to sit down with these verbatims and identify emotional words. If you’ve got a bit of infrastructure, you can use AI tools for natural language processing and let these tools strip out those words for you. You’ve probably seen word clouds where the most frequently used word is the biggest. That word can be associated with emotions, like anger and fear. Armed with that, you as a business can make better decisions. Whether you’re a small business doing things manually or a big company automating the process, what’s important is how you use the data. You should feed that directly to your marketing or customer service systems. 

Image from Unsplash

Keep in mind how important these interconnected bits are. Because when something goes wrong, your customer won’t blame your delivery driver or your supplier — they are going to blame you. You are responsible for any failure no matter how unfair it might seem from your point of view. So you really have to understand what your customer genuinely cares about and why do they care about it, and design your experience around that.

This is also where having a decision tree of templates becomes essential. Imagine if you’re phoning with a complaint to a call center and you get routed to a happy-clappy individual. If you’re angry, a customer service representative that’s upbeat is the last thing you’d want to hear. What you’d want is someone who a bit more somber, someone who’s going to say that they’re taking your problem seriously. You need to have a script, but it should be slightly different. You have to use certain words and tones depending on the customer. And for Peter, it doesn’t have to cost your business more money. If anything, it can save you money if you do that in the right way. You’re avoiding having a failure of process and a failure of thought.

To learn more about digital empathy, visit, which offers different resources. Their company also provides a free discovery workshop for people who want to find out more about this matter.

This article is based on a transcript from my Podcast The UnNoticed, you can listen here.


Cover image by Austin Distel on Unsplash.