The contact centre is changing. In the past, call centre agents had to process a large volume of standard calls, really quickly. But with the deployment of new technologies like artificial intelligence (AI) and robotic process automation (RPA) these agents no longer have to carry out incredibly repetitive tasks and can rather focus their attention on tackling more complex customer concerns.
For call agents, RPA makes it possible to complete simple tasks across back-end systems, which reduces the amount of time spent on admin, says Adriaan van Staden, senior sales manager at call centre tech vendor Genesys South Africa. RPA is in its broadest sense an application that is governed by business logic and structured inputs, aimed at automating business processes. Using RPA tools, a company can configure software, or a “robot,” to capture data for processing a transaction, triggering responses and communicating with other digital systems. RPA lets businesses automate day-to-day, basic rules-based business processes.
AI accelerates system efficiency
More advanced systems use artifical intelligence or machine learning, a subset of AI that enables systems to learn and predict outcomes without explicit programming, embracing methods and algorithms – such as neural networks — that let applications to better their performance as they ingest more data.
For customers, these RPA and AI-powered solutions can solve problems more efficiently and better serve their needs. Finally, for contact centre managers and CIOs, AI-based systems can track each agent’s competency. These insights can be used to ensure that your always provide the best service.
But true AI is still a new area of technology and the scope is not very well defined, explains Mark Walker, associate VP for sub-Saharan Africa at IDC. Not only in Africa, but also across the globe. Progress is being made, though. “Contact centres offer a great business case because these solutions can be used across the entire value chain,” Walker said.
With AI, it’s best to walk before you run
Some CIOs are taking small steps, Walker said. Typically, they start out with automating the front end using chatbots, natural language processing (NLP) or natural language understanding (NLU) to provide automated responses. The next step along the AI journey generally involves using machine learning to run something like predictive questioning. And then they’ll start using AI engines to run analytics that complements the role of human agents.
NLP, and it’s sub-structure NLU, make it possible to grasp what question a customer is asking, in addition to understanding the context of the conversation. In practice, NLP knows that Johannesburg is a noun, while NLU understands that Johannesburg can also be a City and a suburb, notes Brandon Meszaros, CEO of CGX, a customer experience unit of Johannesburg-based business, Digital Solutions Group (DSG).
When handling “low value” interactions, CIOs can use AI to reduce support workloads by letting the technology handle repetitive tasks, Meszaros adds. “It’s not about replacing people but rather about using technology to boost their overall efficiency.”
For example, call centre quality assurance (QA) usually happens at a very slow pace and it simply isn’t possible for humans to listen and review every call. But Pommie Lutchman, founder and CEO of Ocular Technologies, a customer engagement and digital experience solutions provider in Johannesburg, explains that an AI engine can listen in on every call and provide a QA report in as little as 30 minutes. And the ability to scale means there is incredible potential when using these systems.
Please hold for the next available chatbot
Levi, a pizza bot available on Facebook Messenger, can help you find your nearest Roman’s Pizza branch, tell you about the latest specials and, eventually, be able to capture and place your order. “Like most fast food brands, we get a lot of tier-one support queries,” outlines Matthew Jackson, head of digital at Roman’s Pizza, the large restaurant chain in South Africa. “We have a relatively small support team so about a year ago, we decided that we needed a solution that would automate these ‘simpler’ interactions and could escalate any complaints/issues to a human being when required.” Using cloud, machine learning, cognitive search and Maps APIs, Levi is equipped to deliver results that are relevent to the customer. Ths is a critical part of the equation because a bot’s ineffectiveness should not add to a customer’s frustration.
“One thing that we didn’t expect was just how much learning Levi had to do ‘in the wild’, versus in a development environment. Farily quickly, we learned that how you expect people to use the bot and how they actually use the bot are too very different things,” Jackson says. Customers have even sent nudes to Levi in the hope of securing a free pizza, he says.
Using AI for predictive analytics
Over at Momentum Metropolitan Holdings Limited (MMI), a South African-based financial services group, Carol Atkinson and her team are looking for ways to future-proof the business. Atkinson is managing partner of the brand’s disruptive innovation team, Exponential Ventures. They are excited about what AI can do but she cautions that the technology is so nascent that any particular applications they have are still in their infancy and fail to incorporate what she describe as “full blown AI”.
That being said, MMI have deployed machine learning in their contact centres as a predictive analytics and propensity-modelling tool. This allows the company to identify upsell and cross-sell opportunities, states Atkinson. Machine learning uses different unstructured methodologies and algorithms to look at things in a totally unique way, she continues. This means that your clean data doesn’t have to be clean and you can identify unstructured correlations that humans may have missed. “What we’ve learned is that working in this way is quite different. It is an iterative process, so you certainly aren’t going to build Rome in one day. But if you’re not dabbling now, you won’t know how to use these innovations in the future.”
Bad bots are worse than no bots
Bots may be the order of the day but many people are building bad bots, asserts Lutchmann. If you build bots that come across poorly – they aren’t natural language-enabled, don’t do conversational AI, aren’t able to run cognitive search or real-time tone and sentiment analysis and they fail to understand colloquialisms – your investment won’t be very beneficial to the brand, or the consumer.
Chat bots shouldn’t be deployed only to perform a subset of the functions that a discerning human being can perform, adds Wynand Smit, CEO of contact centre solutions and optimisation business, INOVO. CIOs need to take the time to train bots so that they have the same competencies and capabilities as a regular, human call centre employee. When AI bots are used effectively, a bot can provide a customer with the phone number he needs, and then send along the customer’s information to a call agent so that the agent has context to handle the customer’s query even before picking up the phone.
Using AI to automate poor processes, though, is not a smart approach, notes Smit. And at the outset of any AI deployment, CIOs must be sure to take a step back and evaluate if their processes need to be improved before they throw technology at any of the problems they may be trying to solve.
So, can AI revolutionise African call centres? Perhaps. But there’s a long way to go before these intelligent machines become a contact centre commonplace.