26 May Artificial Intelligence as a Service
Now that we’re entering the customer-focused era, CEOs have an important choice to make. It’s not about whether to harness the power of machine learning to improve customer services, but how. Because the required skillset needed to develop machine learning workflows and apps is in high demand, it can be a complex exercise in seeking the right talent.
Instead, companies looking to explore cognitive aspects of their customer service to enhance operations and workflow support should consider outsourcing it. Many companies looking to harness the power of machine learning are seeking out companies that offer AIaaS (Artificial Intelligence as a Service) to improve customer services.
If you think that machine learning (a form of Artificial Intelligence) need not impact you or your business in the next five to 10 years, consider the following:
- Gartner predicts that by 2020, 20% of companies will dedicate workers to monitor and guide neural networks
- 55% of consumers’ customer service interactions now begin online rather than on the phone or in person. This percentage jumps to 65% for consumers ages 18 – 34
- Nvidia plans to train 100,000 developers on deep learning AI in 2017
- Global consultancy, PWC, named AI as a technology CEO’s should consider now due to its potential business impact and commercial viability
To demonstrate there is real value and business impact in utilizing machine learning to improve customer service, let’s review real-world applications already in place.
Text-based NLP
One of the most ubiquitous forms of machine learning found in customer service today is the chatbot. You’ll often see it in the form of a callout in the bottom left or right corner of website that offers visitors the chance to “ask a question” or “ask for help.” Chatbots are enabled through the application of NLP (Natural Language Processing), the same technology used to enable Siri, who aims to help you just as a human would . However, when machine learning is combined with NLP and large sets of data – think other text-based conversations in which the computer successfully answered a query – a computer is intelligently able to respond to website visitors with a response as a human would and improve as the data set grows.
Something to keep in mind is that since this is one of the most popular forms of AI found on many websites today, many consumers are aware these are not human powered. For companies using the AI-powered chatbot as a means to aid sales or customer service, it’s wise to be forthcoming when the “ask me” friend at the bottom of a website is powered by a bot or by a human.
And, because consumers today want to interact with companies the way they want (think email, text, in a chat window) on the device they want (mobile device, desktop or tablet) and in a humanlike, intermittent manner (think of a text message exchange that might go for days based on the availability of either party), it’s wise to offer customers choices in how they communicate with you, such as:
- Allowing the customer or prospect to ask their question not just during regular business hours but when the business is closed as well
- Asking the customer if they would like the response emailed or sent via SMS to them when answered
- Allowing the customer to move the conversation to their preferred method of communication, such as email, and keep it there until resolved
Voice-based NLP
In addition to being able to analyze text-based conversations, NLP is also already being applied to live or recorded customer service calls. In fact, this form of AI is already in effect in a rudimentary form with virtual receptionists or operators that allow callers to answer a question verbally instead of pressing one for customer service, for example.
However, by applying machine learning to NLP, a number of AI startups can help companies improve customer satisfaction by applying a sentiment analysis to a conversation to help predict whether a prospect will purchase in the future or customer will cancel by analyzing the tone of voice and verbiage used on a call.
Combine that data with what you have in your CRM on that customer and you could prevent that customer from leaving or send the conversation to their sales representative because AI helped to identify potential up-sell opportunities. Or, AI may be able to identify a solution that the customer service representative was unable to uncover. A workflow could be applied to alert a human to review the potential solution and if approved, an automated call or email could be sent to the customer to alert the customer the issue has been resolved.
Payments and Collections
As you might already be aware, machine learning is already being used to determine if a borrower is suitable for additional credit or a new loan by applying such factors as credit score, age, income and other factors without the help of human in online lending, for example.
Even better yet, companies can use machine learning to be proactive in collecting near term or past due payments from existing customers by “using web-sourced data to more accurately predict borrower delinquency,” according to the fintech experts at McKinsey & Company. By giving customers payment options and by communicating those in their preferred methods, such as email or SMS, this significantly reduces late payments and accounts being sent to collections. McKinsey & Company added that, “companies using machine learning have been able to reduce their bad debt provision by 35 to 40 percent.”
Finally, companies can easily fight fraud with machine learning that aggregates buyer payment data and turns that “into behavioral signals that are predictive of fraud, and block payments that have a high probability of being fraudulent,” according to a recent Forbes article.
While these are just a few of the examples of how machine learning can be applied to your business, the take away here is that while many of these applications improve customer service, they can also improve your bottom line by heading off problems before they happen.