João Graça is the Co-Founder & CTO of Unbabel.
getty
When it comes to customer service, there's always a high-wire balancing act between cost and quality. Many teams today recognize that technology is one of the best paths to finding balance without losing your footing with customers.
In particular, artificial intelligence can dramatically lower customer service costs while improving outcomes like customer satisfaction (CSAT) and first contact resolution (FCR). However, that's only if you do it right.
Here are a few examples illustrating how AI done right can lower customer service overhead without negatively impacting KPIs.
Feed The Bots
Artificial intelligence is only as good as the data we empower it with. This is particularly true with chatbots, which are computer programs designed to mimic human conversation. If you've ever interacted with a bad chatbot online, you know how unpleasant and frustrating (if occasionally funny) it can be. In fact, the internet is littered with examples of chatbot fails. However, customer service teams understand how much money can be saved when chatbots are used wisely, so we need to get it right.
The key to success is to provide them with as much correct and appropriate information as possible. A robust set of online FAQs can serve an important dual purpose for customer service. First of all, they allow customers to search and find answers to their own questions. Secondly, the questions and answers (and variations of each) can be fed into AI-powered chatbots to create templatized answers and simplify outputs to reduce the risk of unintended gaffes or unhelpful responses.
Tools like topic detection and sentiment analysis — specific AI-powered technologies — can also be used to hone chatbots and make them more useful and accurate.
Additionally, make sure chatbots are being used in the ideal scenarios. Simple queries are the best use case for a chatbot, while humans should generally handle more complicated or emotional customer inquiries. A strong AI algorithm can tell the difference and route them accordingly, saving the company money on customer service without diluting results.
Speaking of results, when it comes to chatbots, measurement matters. Some of the metrics to pay attention to when testing and refining your chatbot strategy include:
• Number of abandoned chats.
• Number of chats that end without the customer getting an appropriate response.
• Time to resolution.
• Net promoter score for human agents vs. chatbots.
• Number of times chatbot conversations are rerouted to humans.
• Number of interactions before resolution.
You can use these metrics to improve quality over time without throwing the cost-effective chatbot strategy out altogether.
Keep Humans In The Loop
One of the biggest misunderstandings about AI (and technology in general, to be frank) is that it will steal jobs. It's true that many low-skill roles will eventually be consolidated into tasks that artificial intelligence and other computer programs can handle. However, this doesn't necessarily equal a net decrease in jobs.
In fact, much of the work that customer service agents do today is rote, boring and frustrating for them. Smart organizations realize that such work is likely to lead to high turnover rates. Instead of tasking humans with answering an FAQ manually for the eighth time in one hour, refocus workers on solving customers' pain points and resolving complex cases.
When customer service agents become problem-solvers and creative thinkers, their insights can be used to employ AI more strategically while also making their jobs more enjoyable and rewarding. Moreover, AI only works well with humans in the loop to some extent. Humans must build the algorithms, train them, refine them and continually update them to ensure that they operate as efficiently and effectively as possible.
Buy Off The Shelf
Not so long ago, any company that wanted to use artificial intelligence or machine learning needed to build its own algorithms from scratch. However, a lot has changed in the last five years. Many solutions for customer service (and beyond) are available off the shelf. In most cases, there's no need to reinvent the wheel and develop AI in-house, and it's often much cheaper to buy.
Even if you need to customize algorithms to your specific needs, toolkits like TensorFlow and PyTorch come with robust tutorials and communities. People with no machine learning background can learn and apply useful concepts without extensive training.
It's a brave new world out there, and the best thing to do is jump in. AI has gone mainstream, and the democratization of this technology has all kinds of potential positive impacts for customer service and beyond.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
