How to nail the mix of automation and personal help in live chat for support

It seems like every few years the frenzy over automation and AI for chatbots whips up again, and the arena of live chat for support is not an exception. In the public imagination, a super-intelligent computer that can do tasks, help people, and guide people has always been popular and tech that is “smart” has been on fire in recent years (think Siri, Alexa, Google Home).

This excitement over AI and bots isn’t just for consumers; it’s also for customer service. The hype had perhaps never been higher than in 2016 when Facebook announced that its Messenger platform was ready for bots. Suddenly, you could get anything from shopping help to daily news with the promise of smart assistants and smart customer service around the corner.

But despite the big promises of every wave, we are yet to see a great, fully automated live chat support. While chat tech and AI are getting better, they’re not perfect — and customers don’t want mediocre service.

So instead of eschewing customer service agents for an army of bots, companies have quietly adopted the automation that’s customer-ready while maintaining service agents who take over to help customers with a personal touch. Here’s how.


Automating live chat for support where it counts

Since automation and bots can’t do all of your support, you have to carefully choose what is most useful for you to automate. The places where automation works well are simple queries that otherwise would take up valuable time for your customer support agents whose energy is better served elsewhere.


Customer intake

One of the biggest drains of time for customer service agents is intake. Collecting basic information from every person opening up a chat is no small feat. With good live chat for support software, you should be able to automate qualifications from customers and work with your CRM to pull up or create customer files.

Take this example from Freshchat’s bot that shows how quick it is to set up an intake. You can easily customize your chat to get your agents the exact information they need to start a conversation off on the right foot.

chatbot workflow

When you automate this part of the chat, every customer gets a standard and efficient user experience, and agents can put their time towards solving complex problems. When they’re called into a chat, they are prepped and ready to help without having to deal with rote questions.



The bulk of questions a company receives are not unique or hard to solve — for an e-commerce company thinks, “what’s your return policy?” and for software think, “what are your enterprise features?” These aren’t questions that customer service agents need to field via live chat because your company has probably already covered them in an FAQ or help docs page.

When a customer uses live chat for a query that has a readily available answer in your documentation, you can automate them to FAQ features, or you can create automated responses that give them quick answers to their problems. That way, your service agents aren’t copying and pasting your return policy 30 times a day and can focus on leads and queries that need their personalized service.


Handoff to humans when questions get complex

While automation can certainly help get things started and can point customers in the right direction, you’re still going to need to figure out a handoff point between a bot and a human for those cases that are not appropriate for bots.

That handoff point is going to be different depending on why the chat needs to go from bot to human. Here are three of the most common cases:

  • If the question has no available automated response. Every now and again a customer will have a problem that hasn’t been encountered before or hasn’t cropped up with enough frequency for you to automate an answer or write a help doc on it. In these cases, the customer should be handed off to a human after their intake information has made it clear that their problem is not standard.

Take this example where a customer is proceeded to live chat when their query about sizing had no automated answer available. A customer service agent is able to ask what measurements will work instead of putting the customer on hold.

live chat human help

  • If the question has multiple parts. Live chat automation is great for a one-part answer where customers can quickly be directed to policies or docs that satisfy their queries. But when a question or a complaint has multiple parts, there’s a much smaller likelihood that a bot will be able to get a customer what they need quickly. One way to hand off a customer with a multi-part question is after they show in intake that they’re having problems with multiple systems (e.g., billing and account login). Another way is to provide an option for a customer to request more or different information after receiving an answer from a bot, at which point they are turned over to a human.
  • If a customer gets frustrated. Sometimes dealing with a bot can be frustrating, especially for older customers or customers who are having a hard time describing their problem. One way to help get customers out of an automated loop is to opt them in to live support after a certain number of negative responses (e.g., “Did this help solve your problem?” “No.”). This will catch the customer before they become so frustrated they have to leave the chat to find another avenue of support.

Exactly what points your customers need to be handed off may vary a little from company to company. It’s possible that a simple multi-part question for a financial app might be easy for a bot to handle with documentation, whereas a multi-part query for a retail order might spell trouble.


Personalization can’t be faked

While bots can be cheerful and call people by their names, they can’t handle personal service or high-touch conversations. This is a limit to automation, but it isn’t a failure in service. When you shift a customer over from a bot to a person, you’re ensuring that the customer feels well-served by your live chat for support.

The key to making the switch from bot to human is letting your customer service agents process all the information from the automated interaction before they start chatting with the customer. Repeating problems over and over, especially if they’ve not been solved by the first line of customer service, can frustrate customers.

This will also stop agents from giving duplicate advice. Pointing out the same tips that the bot gave will do your customers no good. It’s better to keep someone waiting for a minute while your agent reviews their information than to have them dive in and immediately hit your customer with the same questions they’ve just answered and given the same advice that didn’t work.

Keep in mind that when an agent takes over for a bot, they may need to white-glove the situation a little bit, especially if it is with a frustrated customer. That extra personalization and attention can help smooth over annoyances that a customer may have experienced with a bot if the bot was unable to help them.


Live chat for support automation is part of a dynamic solution

There may come a day when AI and bots can beautifully handle all of your customer’s questions and problems. Until that day comes, view lives chat for support automation as a part of a custom solution that your company builds. Good live chat software will help you put in automation and bot interactions that stop agents from having to do grunt work. At the same time, they can quickly solve easy customer problems. The rest of your support infrastructure can help your automation to make it a cog in your well-oiled customer support machine.

(Cover illustration and images by Karthikeyan Ganesh)

Leave a Reply

Your email address will not be published. Required fields are marked *