Taking a fresh perspective on AI in contact centres

We have all been there. Punching numbers on our phones  – bank account details say – then listening to a set menu that tells us to ‘Press 9 to speak to an agent’ and frantically hitting that number nine button.

More recently we have been talking to our phone in an attempt to reach our bank – we’ve stated our account number and sort code, and loudly shouted our first dog’s name to get through security.

So we’re used to automation, even if we don’t love it. Speech recognition goes back decades – the first workable systems arrived in the 1960s  – but Artificial Intelligence (AI) has significantly upped the ante, offering a more conversational experience and personalised choices: ‘Your credit card balance is due – have you called to make a payment?’.

AI can deliver better experiences for customers. And that can translate into a more interesting and fulfilling role for contact centre agents, too. In a profession where levels of expertise are often too low, and job churn too high, that’s important.

A working example

Here’s an example from real life. The fully automated transaction we’re about to describe really does work with this level of skill. On second read, you’ll spot all the places where it could go wrong. We’ll talk a bit more about that later.

Picture the scene. You’ve missed your connecting flight. It’s late in the evening. You’re tired, stressed and stuck at the airport, not knowing what to do. You call your airline’s customer service, and the call is answered straight away. It’s a warm, friendly voice, with not a hint of stress or tiredness.

Not only is it picked up straight away, but you’re greeted with your own name. And the ‘agent’ is pretty sure it knows why you’re calling. So you’ve missed your flight? The next flight is at 9.30 tomorrow morning, would you like to rebook? Not convenient? How about midday? Okay, I’ve booked that for you.

The AI has recognised you from your voice ‘fingerprint’, and confirmed it from context and meta data such as your phone’s location and calling line identification (CLI). It knows you’ve missed the flight because you weren’t on the connecting one. It has pieced this together from booking records, flight manifests, and because you’re calling for help when you ought to be on an aircraft. That’s the only reason for this particular call, at this particular time. And the best solution is to get you booked onto the next flight, so you can get to your destination as quickly as possible. When the context of the call is as specific as this, the AI doesn’t need a person to figure out the best course of action.

Right place is definitely served by AI voice ID, because authenticating who you are is often challenging, especially when you’re stressed. Replacing the usual triage questions with a skilled prediction – have you missed your flight – falls into the same category. Since creating enriched experiences for your customers and your agents is your goal, then right time is well served too. We’ll get to the right AI piece in a bit.

Finely balanced

Within a workflow, and within individual transactions, there has to be a capacity to shift the equilibrium between AI automation, and human interaction.  A type of AI that makes its difference here is real-time speech analytics (RTSA). It monitors stress levels, speech clarity and workflow adherence, all while the call is in progress. So if the passenger in the story above was getting impatient with the AI or her voice was emitting stress levels above the threshold for AI-handled response, the AI would ask the caller whether they would like to be transferred to a human agent, who can provide some empathy, ensuring a fruitful human-to-human interaction.

At that point, having already captured a lot of information, the AI can deliver it to the human agent ahead of transferring the call so the customer does not have to answer the same questions all over again. It can then use the context it’s gathered to pull up guides and information that would help the agent to solve the customer’s problem. Imagine a concise, directional, and informative study guide placed at the agent’s fingertips. The interaction between the customer and agent is made more expert, more fluid, and allows more room for empathy. There is a much greater chance of a resolution without further escalation. Win-win.

Right AI

Too long a topic to explore here, but a couple of points worth highlighting from the long list of things we could touch on. The first is method. It is nearly impossible to build the perfect AI-driven contact centre from a blueprint. The method of choice is fast prototyping and ongoing iteration, workflow by workflow. That needs a toolbox of AI building blocks, shaped for purpose. We’ll be writing more on this.

Second thing, it’s helpful to disclose whose AI toolkit delivered that working example above:  Amazon Connect including Voice ID for voice recognition and authentication, Contact Flow Engine where the customer experience is both defined; and refined; and Amazon Wisdom harnessing RTSA to deliver agent guidance and deep knowledge.

In a nutshell

Contact centres are the hub of your customer experience and your ongoing challenge is to get them right. When they misfire, you’ll not only frustrate your customers but you won’t hang on to your staff either. AI enablement can greatly increase satisfaction within and with your organisation. The road to successful AI integration is about good process, fast prototyping and constant iteration.

Understand that AI will not fix all of your customer service issues – so figure out your right place, right time, and right AI . Then it will make a critical difference both to customer outcomes and staff retention.