AI Transformation Can’t Happen Without Humans

By Zhecho Dobrev

AI Transformation Can’t Happen Without Humans

A customer called and requested to return a product. The customer explained to Sue, the agent she was speaking with, that she needed to return a product that her husband ordered. He died a couple of days before receiving the package…

As you can only imagine, the woman was struggling to speak and her emotions were at the surface. What do you think was the first thing the agent said? Offer their condolences? No! Sue immediately asked the customer, “…do you have your receipt?” and jumped right into processing the return. You might think “Is Sue just a cold person?” but that wasn’t the case. It’s just that Sue was just missing empathy training.

You would think this story is a bit extreme and a one-off, wouldn’t you? I thought so as well. That was until I walked into a couple of contact centers shortly after reading it. A bank that I had a long-lasting working relationship with was about to go through a merger with another bank and asked me to visit the two banks’ contact centers.

Amongst other things, one of my most striking findings was the lack of empathy by agents. Similalry to the story mentioned, an agent was calling an old lady who was interested in a personal loan. The agent started the loan application and one of the first questions was about marital status. “Are you single, married, divorced or widow?”, the agent asked. “Widow”, said the lady, to which the agent answered with “OK” and moved on to the next question.

Another memorable story from my call listening was when a business owner called the bank to complain about the messy cables the POS installation crew had left behind. The agent listened but said they couldn’t do anything as “these are the cables we have”.  The customer was taken aback and enquired again, “So, really, you won’t do anything? It’s because of small things like this that I’ll cancel all my accounts”. The call was not even logged as a complaint!

I used that call later during training sessions. I asked agents “What do you think the lady really wanted”. Only a handful were able to realize that what the lady really wanted was to be listened to and treated as a valuable customer (i.e. her complaint to be taken seriously). I asked the agents: “This lady is a store owner – how many stores does she own? Does she have credits, etc.? How much does it cost to send a crew back (it turned out they use vendors) versus losing the customer?”.

To put the above examples into perspective, empathy was rated as the no.1 most valuable contact center agent attribute by contact center managers[i]. Another report found that intriguingly, customers who did not have their query resolved but had an interaction which understood and acknowledged their emotions very well were much more likely to be satisfied (20%)  than those that had their query resolved but had an experience where they felt their emotions were not understood or acknowledged at all (8%)[ii].

So at this point, you might be thinking – OK, the agents need training, and when you train them they will show more empathy and customer satisfaction ratings will increase. Not so fast, I would say.

Changing a behavior isn’t that easy.

I did in fact train the bank’s agents on those agent behaviors proven by AI to drive positive customer sentiment scores at the end of the interaction but… the contact center management team, which had all changed since my first visit, was too busy to attend the training and there was no one to keep enforcing the new behaviors. On top of that, due to the higher volume, the bank hired hundreds of new agents, who were not trained on those proven behaviors.

How AI could help

Both banks had already invested and kept investing in the development of AI bots. Those can of course take away some of the traffic from the already tired and stressed agents. But, as we saw, that doesn’t solve the empathy problem. What could help is  AI models that can provide a sentiment score rating on every single interaction. Those are so advanced that they give more weight to phrases occurring toward the end of an interaction because research has shown that they have a stronger prediction of outcome than those spoken earlier in the call.

What’s more, agents could be automatically scored on the different behaviors proven to drive customer sentiment. One agent might be better at showing empathy, another at effective questioning and so on. This could help the quality assurance team, who instead of randomly picking and listening to calls can cherry-pick which calls to listen to and use in their coaching sessions with agents. This too can be automated to an extent where agents are automatically pointed to some coaching & feedback learning modules that are automatically scheduled in their diaries.

And then there is the data democratization aspect…

We know that the contact center could be a great source of insights, but we also know that executives hardly spend any time in it. And it’s hard to blame them, as listening to many random calls is like looking for a needle in a stack of hey. Here, the AI-generated call scripts and summary notes not only save agents time but also allow for easy keyword searching. This, coupled with the call sentiment analysis, allows executives from all functions to jump right into those parts of the call that are of most interest to them without leaving the comfort of the chair in their vacation home.

Finding the root causes…

One of the best uses of AI is in tracking the end-to-end customer journey behavior. Wouldn’t it be good to eliminate unwanted and unnecessary contacts? And what about those who just abandon the cart, the interaction, or leave the company and never even bother calling? Nowadays, however, there are AI-based platforms that allow organizations to track end-to-end customer experience, which will allow continuous improvement teams to find behavior bottlenecks and use behavior nudges to shape the desired customer behavior.

AI can help, but don’t put the cart before the horse!

Some years ago, we worked with a water utility company that was planning to invest further in social media channels for contact. We did research and found that those channels were not driving value for the company. No wonder as 94% of the contacts came from the phone and many were about blocked sewers. Now that’s something that you won’t brag about on social media, would you? And when things were complex (e.g. repeat issues) or urgent (e.g. flooding or lack of water) the phone was the channel of choice for customers. The point is this – just because everyone is investing in one thing, it doesn’t mean that that would be the best thing for you.

Another company that we worked with recently, a Canada-based energy company, had switched the way they received orders from their commercial customers requesting refills to their heating oil tanks. Customers had to call a 1-800 central number instead of calling or visiting their local office, as they used to do. If you think about it, with this decision, they had taken away the feeling of a relationship with the company. Instead of visiting the local office where they are known and respected and can chit-chat for a moment, they had to call a central number where they had to repeatedly provide instructions to the drivers to find them in the remote corners of Canada.

This led to hundreds of complaints and many customers walking out on the business. The company soon reverted to the old ways.

Now, was the company wrong to pursue some ways to drive more efficiency using technology? Not necessarily. However, they should have thought about how to use technology to preserve the feeling of a relationship (which my research has proved to be a major driver of value for organizations). Caller ID and 360° view of the customer would certainly have helped. They could have also used AI-based predictive behavior routing to connect with the agent that is best positioned to build rapport with the customer.

In conclusion, AI is more than hype and can be used to make meaningful improvements to the customer experience. However, organizations:

  • shouldn’t be solely focused on cost-cutting opportunities and look at where AI can help to improve the customer experience, reduce churn, and generate more revenue.
  • shouldn’t forget the human element. Not only are calls that reach the agents more complex and require more empathy, but most contacts don’t originate from the contact center. They are created by what the rest of the organization does e.g. supply chain, finance, marketing, product development, and digital teams. Therefore you need to engage and change the minds and micro-behaviors of the rest of the organization as AI transformation can’t happen without humans. That’s what we at Human2outcome specialize in.



[i] 51% of respondents in a survey of 189 contact center managers and directors rated empathy as  the no.1 most useful and valuable contact center agents’ attribute (US Contact Center Decision-Makers’ Guide 2024 – 16 th edition)

[ii]  Source: “EMPATHY IN CUSTOMER SERVICE”, & Genesys, 2020