There’s a cashier at my neighbourhood grocery store who makes the same mistake every time I get to the front of her line.
She keeps trying to sign me up for a promotional credit card. It’s not a big deal—but I’ve politely declined three times now. Every time I’m in her line, she forgets I’ve already said no and offers it to me, with a one-minute sales pitch … again. It’s almost enough to make me shop at another grocery store.
It’s exactly the kind of irritation companies are trying to remove from customer experience (CX). In the 21st century, when contact centres function as the navigational cockpit for customer relations, many brands are turning to AI and natural language processing (NLP) to ensure CX is as free of turbulence as possible.
The changing contact centre
When was the last time you actually called a company on the phone? If you’re like most consumers today, you expect to be able to reach out to any company in any channel you prefer, including email, text, social media and web chat—and yes, the phone.
This is why the broader term ‘contact centre’ is now favoured over the narrower term ‘call centre.’
Besides being omnichannel, today’s contact centres are data-driven. Businesses have a wealth of digital data about their customers, and with AI, they can use that data to respond—as accurately and relevantly as possible—to inbound queries and complaints made by customers to their contact centres.
AI in the contact centre
If you’ve traded texts with a customer service bot or had a web chat with one, you’ve experienced AI in action. Gartner expects that by 2021, 15 per cent of all customer service interactions globally will be handled completely by AI.
In an article for Towards Data Science, Alex Fly, co-founder of AI startup Quickpath, lists some things AI can do for the contact centre:
- Smart call routing: AI can route each call to the human call agent best suited to handle it, based on data about the customer’s query, each agent’s experience, the complexity of the call and so on.
- Contextualize and personalize: AI can analyze or detect the context of the contact centre interaction using information such as historical data about the customer involved (such as when, why and how they last interacted with the company). “Context is crucial in a call center because it enables agents to provide a personalized conversation,” Fly told Towards Data Science.
How NLP ‘speaks’ to customers
As outlined in an analyst note by Raghay Bharadwaj of Emerj AI Research, NLP is moving beyond chatbots. It’s really any technology related to human language, whether it’s in spoken voice or written text.
In examples cited by Bharadwaj, NLP can discern email from spam, categorize documents based on their text content, summarize relevant highlights from longer documents, automatically transcribe recorded voice conversations, turn written text into speech and extract relevant clauses from a legal contract.
As such, NLP is tremendously useful in a contact centre for deciphering the meaning of what a customer says or types, so the best response can be provided as quickly as possible.
Bringing emotion into it
The newer frontier of AI in the contact centre, however, is even more exciting: it focuses on sentiment as well as semantics.
“What we are talking about is a way for the AI technology to analyze the tone of voice or even the cadence of the language, to detect what the caller’s mood is,” Dr. Skyler Place, VP of behavioral science at startup Cogito, told Tech Republic.
The goal is to understand the customer’s emotions as well as the meaning of their words. The cadence of a caller’s speech—whether they pause, raise their voice, sound excited, bored or frustrated—can all help the contact centre deal with that person more effectively, whether it’s up-selling them or defusing their anger.
There’s even data regarding swearing and customer sentiment. Research suggests calls featuring customer profanity take an average of eight minutes longer for a human agent to resolve than calls without customer swearing. If time is money, then swearing is pretty bleeping costly for contact centres.
Emotional (job) rescue?
Although AI is now more prevalent in the contact centre, it doesn’t guarantee better results. The CFI Group’s Contact Center Satisfaction Index slipped from 71 to 68 out of 100 in 2019. The survey also found that customers who reached a human agent directly were 27 per cent more satisfied than those whose first contact was an interactive voice response (IVR) system.
What’s going on here? Well, AI doesn’t always get it right. When that happens, customers get frustrated, which exposes a company to risk of customer churn. That’s why analyst Zeus Kerravala argues in CIO that currently, “AI is not ready to be customer facing.”
Instead, he believes AI is best deployed as a way to enhance the performance of human contact centre agents rather than replace them entirely. “A practical approach is to use AI to make human agents smarter and more efficient,” he wrote in the magazine.
That’s where emotion analysis enters the CX realm.
Observe.ai is using NLP to detect a caller’s emotions during contact centre conversations. The company’s software automatically transcribes calls between agents and customers. The AI then analyzes each caller’s feelings and how they change based on actions taken by the agent. This data is used to train new agents and help experienced agents improve their performance.
Place, whose aforementioned startup Cogito competes in the same space, told Tech Republic this type of software could even send real-time alerts or tips to an agent’s screen during a call, such as “you’re talking too much.”
Remember those gloomy predictions that AI will quickly eliminate almost all human jobs in the contact centre? The type of sentiment-sensing AI we’re talking about here could prove those forecasts wrong by using AI to complement the skills of us mere mortals, perhaps saving our jobs in the process.
As for my grocery store dilemma, perhaps I could just remind the cashier that I’ve already said no to her credit card offer three times. It would be very human and very awkward but probably solve my CX pain point immediately.