For many reasons, email is the best proving ground for AI and still the most popular digital customer service channel
As companies race to bring AI into their workforce to meet customer’s quick-rising expectations for effortless and convenient customer service, many struggle with figuring out the right channel to initially launch a support AI.
Customer service automation started to take off as Web chat and social platforms like Facebook Messenger and WhatsApp positioned themselves as ideal channels for 1:1 customer engagement. While these channels offer amazing opportunities for customer service, email has proven to be the ideal channel to launch AI for customer service. This is because email provides an ideal environment for an AI to learn, boost accuracy and confidence, while keeping customer satisfaction high throughout the entire training process.
In this article we’re going to explore the many reasons that customer service email is the best initial channel to launch AI for customer service. Skip to a specific section, or read our in-depth analysis of why email is supreme.
- The trouble with first-generation first-generation chatbots
- Omni-channel is important, but with AI let email take the lead
- The key is asynchronous conversation: Why email is the ideal channel strategy to roll out AI-powered customer service
- Email wins the immediacy race: Comparing customer expectations for immediacy across channels
- Beyond asynchronicity: Even more benefits of email
- The future is every channel, but email is your best first bet
In the past few years, we’ve seen many companies introduce basic chatbots. These first-generation bots that were dedicated to customer service, though, overwhelmingly failed.
Instead of delivering on the promise of convenience, these bots damaged the customer experience. This is because basic chatbots lack natural language understanding (NLU) and limited the user to clicks or keywords. Chatbots have very rigid decision trees, and it’s nearly impossible for users to redirect the bot in a new direction once it starts on a journey. Customers had little control over the conversation, ultimately leading to desperate attempts to be connected with a human on another channel (with customer satisfaction [CSAT] diminishing, and costs increasing, with each passing second.
The best chatbots are more often actually classified as Conversational AI, but the terminology between the two has become quite ambiguous over time. Rules-based Chatbots differ from modern Conversational AI platforms for many reasons, but primarily because AI has advanced NLU capabilities. People are free to explain their issue or ask a question in their own words, as if they were interacting with a human agent. Getting an AI Agent to perform at this level, though, requires specific training which is well suited for the email channel.
While offering omnichannel customer service is becoming increasingly important, launching an AI across multiple channels at once is not attainable for most companies, or frankly, recommended. This is because each channel has a different mode of communication. Think about it: how you text, send an email or talk on the phone are extremely unique.
Chat and social messaging are typically a stream of very short messages. You might send multiple in a row, perhaps using abbreviations, short-form text or digital slang like emojis. On the other hand, emails are longer, sometimes with multiple questions, or more in-depth information and attachments. This is because the number of back-and-forth interactions on email is generally less, but contain much more detail within each exchange.
When it comes to adopting customer support AI, each channel comes with varying levels of risk and requires nuanced training to provide a positive customer experience.
Discover the key questions to ask when scheduling a chatbot demo.
The key is asynchronous conversation: Why email is the ideal channel strategy to roll out AI-powered customer service
Like a human employee, AI learns on the job. While AI systems are initially trained using historical customer service data, a Support AI continuously improves based on real interactions, learning the various ways a person might ask a question. Over time, AI learns how sentiment and context impact what a customer needs and the best course of action to take, truly replicating a human agent.
On email, companies can enable AI to learn on the job without ever interacting with a real customer. When a new ticket comes in, AI can recommend responses to human agents, instead of automating a response directly back to the customer. AI will learn and re-tool based on how human agents respond: reinforcing a positive behavior or adjusting wrong behavior.
This is possible because on email, there is no expectation from a customer for an immediate response. The delay when an agent is accepting, rejecting or editing an AI-recommendation does not even register for the customer. Compare this to a chat conversation where people expect an immediate response and synchronous conversation. This “agent-review” time (and ensuing AI training) would not be possible due to the expectation of a response in a few short seconds.
Once AI has boosted its accuracy behind the scenes, it can start responding to emails directly, freeing up human agents to solely focus on more complex tickets, which is where true cost savings and operational efficiency will take hold.
Launching a support AI is a zero-risk way for companies to gain confidence and boost accuracy.
Email Support AI will always meet customer expectations: Comparing expectations for immediacy across channels
When thinking about what channels to extend customer service automation following email, a good rule of thumb is to consider how quickly people expect a response.
On social messaging platforms like Facebook Messenger or Twitter DM, faster responses are expected as compared to email, but expectations are still not for real-time, synchronous interaction. This does allow companies to have an agent-review period and behind-the-scenes training, although there is more pressure for a faster response time. This “incubation” period will likely be shorter as messages are typically easier for an AI to parse, as compared to email.
Now, let’s consider Web chat. Here, waiting even 30 seconds is not acceptable to many customers. This is because they can’t go about other business and walk away from the screen, coming back for a response when it’s convenient for them. A person is stuck on a Web page until their issue is resolved. Having an agent-review mode is not ideal here due to these immediacy requirements. If a company has gone through the process of agent supervision training on messaging and email, though, a much more accurate and confident AI agent can easily be launched on chat.
Finally, we get to emerging voice support channels like Google Assistant and Alexa. Synchronous, real-time interaction is the norm. Because these channels are more complex and expensive to introduce a support AI, we recommend exploring these channels only after more traditional digital channels have launched and are performing with high accuracy.
More than offering great real-world training benefits, email is actually how people want to get in touch with companies. In a recent survey, we found that 47% people prefer to contact a company for customer service over email than any other channel. Email beat phone (23%), Web chat (23%) and social messaging sites (2%).
Email is inherently more convenient than other channels as it is a part of every consumer’s daily habits – from Generation Z to Baby Boomers. Email is readily accessible on all of their devices, there’s always a record of the conversation and people don’t need to wait around to carry on a back-and-forth chat or phone call. The entire interaction is on a customer’s terms – exactly how it should be.
Email also provides great cost savings opportunities:
- Companies tend to have more historical email data as compared to other channels as it is the original digital customer service channel. This decreases initial training time and associated costs.
- Email continuously generates a high volume of new tickets which helps the AI to ramp up quickly.
- If a customer is not satisfied with an AI response on chat, voice or social, they will reach out on another channel, driving up the cost of resolving the ticket. When an email Support AI is trained under agent supervision, a human always has the final say in what is sent to a customer.
Customers want convenient, effortless resolutions. To meet these demands, AI adoption is ramping up. Use of AI by customer service teams is projected to increase by 143% over the next 18 months1.
While it might sound counterintuitive, companies should take AI one channel at a time as it will result in better customer experiences and most cost savings in a shorter amount of time. For many reasons, email needs to be the first channel companies tackle with AI, or they risk negatively impacting customer satisfaction while increasing costs.
Can we analyze your historical email data and identify the best issues you can automatically resolve with AI? Get in touch.
- Salesforce: https://www.salesforce.com/blog/2019/03/customer-service-trends.html – learn more about our Salesforce chatbot solutions by clicking the link.