Ryan Steinberg is the Associate Manager of Global Support Operations at Intercom. He joined the company four years ago as a Customer Service Representative, and after about a year moved on to leading the Support Operations team. The three people he manages are in charge of pretty much anything that involves the Customer Support team and numbers — capacity and headcount planning, volume projections, building individual and team KPIs, doing analysis for any ongoing projects. “All the good stuff,” as he calls it.

In his spare time, Ryan also writes for Intercom’s blog — Inside Intercom. He recently penned an article on automated customer service, which is what our conversation focussed on.

Automation in customer support is nothing new. Long before chatbots, if you called your internet provider and had to go through a phone tree until you finally got connected to someone — that was automated customer service. It didn’t, and still doesn’t, offer the best experience, but it was already there, hoping to deflect issues and get customers connected to the right person. Live chat and chatbots serve the same purpose.

What is still relatively new is using machine learning and natural language processing in chat, to run through conversations and automate the simple actions that a support team member would frequently do, such as searching the Help Center for articles pertaining to customers’ questions. This, in turn, helps support teams focus on conversations that need a human touch, instead of repetitive tasks or simple questions.

But despite the technology no longer being new, customers still feel a bit skeptical about interacting with a machine. Or rather, they’re frustrated by it. Not because they’re dealing with a bot, but because the bot might not be able to answer their questions or keeps suggesting the same irrelevant Help Center articles.

We don’t want to put people in the “bot jail,” where they’re basically having to talk to a robot with no escape hatch.

One of the guiding philosophies of Intercom’s Product Team is to think from the customers’ point of view. So when developing chat solutions, they aim to avoid putting people in a situation where they’re stuck talking to a machine, with no human-representative escape valve. It’s a matter of knowing what to automate or not, keeping in mind that making the customers’ experience better is the ultimate goal.

For Ryan, chatbots are not just about cutting costs; they’re about giving customers an answer to their question instantly as opposed to having to wait for a human to respond, which could take hours, or even days in worst case scenarios. Every time they’re able to give customers a quick, automated reply that solves their issues, it’s a win.

And it happens more often than one might think. Intercom currently has a 4% ROAR — rate of automated resolution — which is the percentage of all conversations that were resolved by a piece of automation technology. Article suggestions is what Intercom primarily works with in terms of automated replies. An algorithm runs through the text the customer types in, then looks through all Help Center article and suggests the ones that might be relevant, solving 1% of all conversations.

The remaining 3% come from Answer Bot — another automated solution that looks at similar questions and past conversations to come up with answers. The total rate might seem small — only 4% of all interactions with customers — but it translates into $400,000 saved every year.

It is an incredible number, not only because, you know, it’s $400,000, that’s nothing to scoff at. But this is $400,000 that you’re saving every single year, which is a pretty incredible thing.

The number is only expected to grow in the coming years. Intercom currently has B2C customers with a ROAR between 20 and 25%, while B2B customers experience lower ROAR since their users tend to have more complex questions that are harder to automate. That being said, Intercom is working on improving their help center, building more answers and tweaking the ones they already have, but also using new technologies like the Inbound Custom Bots that helps support teams triage issues more efficiently by giving customers options to click through — much like a phone tree, but in live chat. Not only that, but Custom Bots also allow for businesses to setup bots in multiple languages. With all this in place, they’re confident that the ROAR for B2B customers will rise to 6 or 7% by the end of 2020.

Despite live chat and automation now being an essential channel for customer support, there is still room for improvement. Ryan points the real-time nature of live chat as one of its biggest challenges, because it prevents the customer from being able to do other things while waiting in a queue. A synchronous experience makes it so that, if the customer isn’t there by the time an agent replies, their conversation will be closed and they’ll have to wait in line all over again.

This is why Intercom offers a business messenger that allows an asynchronous support experience. Customers can ask a question, leave, and then resume the conversation at a more convenient time.

We know that you’re not going to be honed in most of the time on just having the support conversation; you have meetings that you need to run to, you have to go pick up your kids, you’re cooking, you’re doing whatever you need to do in the rest of your life outside of this support interaction.

The fact of the matter is that not all customers expect real time support when resorting to a chat tool. Sometimes, setting the right expectations is much more important. Having a bot in place that lets customers know their place in a queue or informs them a support team member will reach out to them soon are a few ways to do that. If a customer writes in with a very simple question like a feature request, for example, they probably won’t mind waiting up to a business day for someone to get back to them, as it is not an urgent issue.

In any case, having live chat and automation in place that gathers information beforehand is what makes the difference, so that when customers actually do talk to a support team member, they already have the context they need to get a resolution to their issues a little bit faster.

Until recently, we might have said the chat revolution had fallen flat. Companies looked at chatbots as a cheap quick fix to solve customers’ issues, instead of thinking of how best to pair automation with live chat to deliver an improved customer experience. But thanks to the latest advances in chatbot technology, companies can now offer support automation that their customers will actually appreciate, and chat could easily become the sweet spot for customer interactions. And that’s why Ryan Steinberg believes it is the future of customer service.