Main illustration: Molly Egan
Automated customer service isn’t a new concept. We’ve all navigated our fair share of automated phone menus or interacted with support bots to get help.
But much has changed, both in usability and customer perception. Voice recognition technology has improved, AI solutions can interpret customer feedback, and chatbots have started to answer the questions they receive, not just pass them off to a human. With this increase in functionality comes a change in how customers actually view automation: less as a nuisance, and more as a bona fide, personalized option for getting help with their issues.
As the head of our Support Ops team, I work daily with my teammates to develop automated solutions that don’t frustrate customers, but empower them. Our goal is to personalize every experience through automation – to deliver exactly what each specific customer needs, when they need it, and without anyone yelling in frustration for a human.
Here’s an inside look at how we’ve implemented personalized, automated customer service at Intercom, and ideas about how to successfully implement automation as part of your own support offerings in the future.
What is automated customer service?
Automated customer service is a way of helping a current or potential customer without requiring human intervention. It works best when customers need answers to recurring straightforward questions, status updates, or help finding a specific resource.
One of the most popular automated support options here at Intercom is chatbots. Our bots use AI to provide customers with links to existing resources, like knowledge base articles and FAQs. They can also route customer conversations to the team best equipped to handle their questions and can even provide answers to simple customer questions like, “How can I add more users?”
Why automate your customer service?
Over the last decade, live chat has become the de facto standard for companies wanting to offer top-tier customer support. Chat is faster than email, more personal than knowledge bases, and way less frustrating than shouting into an automated phone system.
But I get it – live chat isn’t always the most convenient option for customers. Some customers might be multi-tasking, in the middle of their work day, or on the go. For people who don’t have time to have an in-depth conversation with a support rep, automated support options – like chatbots – become just as important as real-time support. Customers can ask your chatbot a question and read the answer between meetings, or get a link to a helpful article and read it during their commute. There’s no pressure to take action immediately. Plus, they can still follow up with a support rep if they need more help. With both live chat and automation in your arsenal, you can truly meet your customers anywhere and at any time they need you.
“For people who don’t have time to have an in-depth conversation with a support rep, automated support options – like chatbots – become just as important as real-time support”
Moreoever, automation helps both your customers and your company. It lets you employ a smaller support team, reduce your labor costs and save money long term. And automation rules allow customer questions that can’t be answered by a chatbot to instantly be routed to the right person and team, ensuring there’s no confusion over who handles what or which questions have already been answered.
When implemented well, automated customer service allows businesses to help more customers at scale without drastically growing headcount. The speed and savings can be game changers for your business… but only if you implement those solutions thoughtfully.
Redefining “personal support” in the age of automation
Think about how your support team communicates with customers: they address them by name, use the customer’s account information to tailor their advice, and take notes to help their fellow team members with follow-ups.
Automated support solutions should be held to those same high standards. Even if they can’t have a conversation about the weather, they can still provide advice that’s personalized to a customer’s specific question or situation.
It used to be that businesses had to sacrifice personalization if they wanted to drive more efficiency through automation. But with advances in automation technology, that trade-off is no longer necessary. The very definition of what it means to be personal is changing, as our CEO Eoghan McCabe explained in his keynote last year:
“I got a great question earlier this year: ‘If we’re trying to make internet more personal, what are we doing building a bunch of bots?’ These questions had us think a little more about the motivation behind being personal, and try to better define it.
“Did being personal mean people on the other end of the line? Or was it more about what only people could do thus far? We started to define personal recently as treating the customer as an individual, respecting their time and dignity and ultimately getting them to their ideal outcome.”
Here’s how we’ve implemented automation that respects our customers’ time and gets them to their ideal outcome at Intercom.
Implementing automated customer service that feels personal
Making automation feel personal takes time and thoughtfulness. You have to make sure to strike the right balance to avoid having your personalization come across as creepy. It’s great when websites suggest support articles before you reach out to support and chatbots offer resources based on the page you’re viewing. But a chatbot using data enrichment tools to address a customer by name is probably not a good idea if this is their first visit to your site.
At Intercom, we work hard to balance automation and personalization in two critical areas of customer support: issue resolution and customer routing.
Instant issue resolution
When customers write in, the most personal thing we can do is to respect their time and get them to an answer as quickly as possible. So where possible, our support chatbot – Answer Bot – will answer that question for them immediately.
Our chatbot can also look at keywords in the customer’s message and suggest matching articles – but only if it detects that the user hasn’t visited the Help Center in the previous 15 minutes. If they had, the articles likely didn’t address the person’s question. In that case, redirecting them back to a resource they’ve already seen is probably more frustrating than helpful.
“When customers write in, the most personal thing we can do is to respect their time and get them to an answer as quickly as possible”
Of course, there will be questions that Answer Bot isn’t set up to address and we direct those conversations to our support reps. We could have designed Answer Bot to direct users to the Help Center, a forum, or some other form of self-service. But while that would have been the most efficient workflow for our support team, it would have been a prime example of impersonal automation, one that doesn’t treat the user’s time with consideration.
Even when Answer Bot can answer a customer’s question, it always checks if the answer it provided gave the user what they need. If not, they can request to talk to a human at any time. Offering instant access to a human is how we stay personalized while avoiding the dreaded “I can’t answer that” chatbot loop, where a customer gets stuck asking questions to an unhelpful chatbot and can’t find an “escape hatch” with which to talk to a real person.
Contextual customer routing
The other area where we heavily apply automation is customer routing. For conversations not addressed by a bot, our assignment rules take care of routing 80% of conversations to the right place, with the rest routed to an escalation inbox to be examined by a human.
What this “right place” looks like varies, so our assignment rules consider the customer’s history with us (if there is one) and decide where they should go to get the help they need. This is how we keep routing personal: we take into account multiple factors – not just account spend – to determine where our customer should go. These factors include:
- Message keywords
- Products in use
- Company type
- Page URL
- Customer tenure
For instance, if someone starts a conversation from our Developer Hub or their message contains technical words like “python” that clearly indicate a technical topic, we know they will likely benefit most from talking to our Customer Support Engineers. Not only does this save our team time by automating the number of manual escalations up to technical CSEs, but it inherently respects the customers’ time, removing an unnecessary touch from a non-technical resource and getting the customer in front of the person that is best equipped to handle their question, shortening resolution time.
How automated customer service is paying off
Using instant resolution and contextual routing has led to some pretty big wins for our support team. Here are some specifics:
- Our rate of automated resolution (ROAR) is about 4%. This means 4% of all customer issues are resolved without a human being involved.
- Every 1% of ROAR roughly equates to a savings of $100,000 per year, so that 4% resolution rate translates to $400,000 saved annually.
- 75% of our automated resolutions come from Answer Bot, while the remaining 25% come from suggested articles.
- 60% of our routed conversations land in the right inbox on the first try.
While a 4% ROAR might sound low, it’s actually a pretty huge number for us, as you might guess from those annual cost savings. 4% is also on par with B2B companies like ours that tend to see more complex questions from customers. B2C companies can get their ROAR up to 10-20%, since many of their questions are far more transactional in nature and thus are more easily resolved by automation. We’ve seen customers for whom Answer Bot resolves 29% of the queries it gets involved in and improves customer response time by 44%.
Ultimately, we think this 4% ROAR and 60% routing success rate are just the start. We’re confident that we can continue to improve these KPIs and route even more customers to the correct place on the first try, all without ever involving a human.
When automated customer service isn’t the right solution
Ultimately, there are some situations where automation isn’t useful. This is usually when you’re in a situation where you can’t personalize the kind of support you’re offering. This might be because you don’t have the necessary context on your customer to treat them individually. Or it could be because your current chatbot can’t interpret that information to make the appropriate routing decisions, like sending customers knowledge base links or automatically routing them to a member of your support team. In these situations – when it’s not personalized – automation becomes a blocker instead of a valid support method.
Here’s what I mean by a blocker: think back to the hours-long phone calls with your cable provider or bank. Have you ever yelled “customer service” into a phone five times, trying to get to a human being? By the time the phone menu gives up on trying to resolve your issue, you were probably too frustrated to properly articulate what your problem was.
“When it’s not personalized – automation becomes a blocker instead of a valid support method. What you needed in that situation was an ‘escape hatch’”
What you needed in that situation was an “escape hatch.” Therein lies the danger of poorly implemented automation. If your customers get blocked by a chatbot or gets routed to the wrong team, they’ll be just as frustrated as they were when you yelled at that phone menu. But this time, the risk is even greater, since it’s so much easier to cancel, tell friends about your unhelpful support, or both.
Poor personalization can turn satisfied customers sour, to the point that they cancel their account and take their business to your competitors. Or they may instead flock to your real-time support options, increasing the need for staff and raising your associated costs. It’s a lose-lose situation.
Ultimately, some customers just shouldn’t be subjected to automation:
- Customers with sensitive issues: Requests for upgrades or cancellations should be answered by a human. When a customer is trying to give you money, you can’t allow a chatbot to jeopardize the relationship before it even begins. If they’re thinking about canceling, poor automation might make any negative feelings even worse, or ruin any chance at saving the relationship.
- High-touch customers: Customers with lots of questions, and those who need hand-holding through difficult processes or explanations, would benefit from working with a human. Most of the time, these folks are more than willing to wait for a person to talk to if they know they’ll get the help they need.
- New customers or trials: Your newest converts might end your relationship early if they find themselves continually blocked by chatbots, no matter how “helpful.”
To prevent issues with these three types of customers, my team maintains a list of questions that we don’t allow to be answered by automation. Customers who ask about pricing, who are identified as at-risk or “high-touch,” or our trial users, are automatically routed to a team member for assistance. Though AI is learning to handle complex problems, for the time being, these customers will get the best service possible if you send them to a human, not a bot.
Automated customer service is useful, but it can’t stand alone
As you work to reach a 2%, 5%, or even higher ROAR, put substantial effort into your entire support strategy. On its own, automation won’t solve the problems of your customers – it needs to be supported by a strong knowledge base and answers from your support team. Without those resources backing it up, your bots will do little more than annoy customers who are desperately trying to seek solutions to their problems…and we’ll be right back where we started with phone systems.