Customer Support for Live Chat

Live chat service around the clock isn’t always necessary. You could start with one shift and expand from there.

The required headcount will depend on the chat volume, resolution times, coverage requirements, and factors specific to your organization. For example, let’s say 12 to 1 is the busiest hour of the day for your support team, and you want to respond in less than a minute. You’ll need to staff enough agents to meet that volume, even if that means you’re overstaffed other hours of the day.

If you don’t have your own workforce management team or software, you could explore tools such as the Erlang calculator, which allows you to calculate the optimal staffing based on your volume, handle time, and desired service level. Our article on how to build a call center staffing model will show you how to build a model that can also account for volume patterns.

Smaller companies may want to start by turning on chat during peak times only. That allows you to test your staffing and systems and ensure everything’s running
smoothly before you expand coverage.

Live chat demands experienced and adaptable support professionals.

Since it’s such an immediacy-driven environment, any support agent who spearheads a live chat service will need to be able to transition from one task to another seamlessly.

For example, at Peak Support, we typically have our agents master email service before moving them to live chat. While an email support agent can ask for help if they’re unsure of an answer, the immediacy of live chat makes that impossible. This is why live chat agents must undergo rigorous training and be able to navigate the faster-paced nature of real-time conversations effectively.

Equipping your agents with quick macros for greetings, handoffs, escalations and other common live chat scenarios can go a long way in boosting their response times and saving mental energy by erasing the need for typing the same responses again and again.

Gartner predicts that by 2027, chatbots will become the primary customer service channel for a quarter of all businesses.

Chatbots aren’t all powered by AI. You can also build a rules-based chatbot that responds based on the instructions you provide. Here’s an example of a rules-based chatbot flow:

An example of Intercom’s Chatbot Builder decision tree

If most of your routine queries can be solved with a concise, one-paragraph article, rule-based automation can be great for guiding customers down a decision tree and prompting them to select the most relevant option at each step. Too many options can overwhelm users, while too few may lack the necessary granularity. Typically, presenting 3-6 options at each step strikes the right balance

Another important item for a customer care chat experience is offering an easy opt-out. There’s nothing quite as annoying as having to rephrase the same question three or four times before finally getting connected with a real person. That’s why it’s essential to offer customers a straightforward way to transition to human assistance if needed. That doesn’t mean a way to transfer to human support needs to be the first thing you offer your customers — but don’t make them hunt to find it, either.

Smart automations and chatbot flows can have a massive impact on your servicelive chat success. When we optimized the website live chat experience for a client in the pet industry — including revising their chatbot menu, making the “chat with a human” option appear later, and optimizing chatbot flows — they saw a 75% reduction in human chat conversations, while maintaining a 97% CSAT.

AI assistants are basically supercharged chatbots.

Customers ask questions, and the bot can track down the information During non-business hours, directing website visitors to your self-service knowledge base empowers them to find solutions independently. For a better experience, set up a conversational AI bot to alleviate the burden on customers – they can simply ask questions and have the bot track down the information they require.

The AI model setup may not be as straightforward as some providers claim, so let’s review the key steps to ensure that you achieve the best outcomes with your AI tool.

For the AI answers to align with your brand’s tone of voice and offer a consistent customer experience, assign your AI model with attributes, such as “professional”, “friendly” or “casual” and experiment with the responses until you land on the desired tone. The persona should work for different types of queries from feature requests to messages about product outages. While creating an AI model with a dry British sense of humor may seem funny, it can backfire if it insults your customers.

Use historical chat conversations and solved support tickets for the AI to learn from real-world interactions and comprehend user intents better.

Put your AI model through a stress test with a large body of potential questions, including some that have nothing to do with your product and others that are very difficult, and attempt to overwhelm or break the AI to identify weak spots in its behavior. The purpose of the test is to ensure that your AI model can navigate any situation appropriately.

Running AI-handled conversations through the same quality assurance scorecard as your human agents is also a great way to ensure consistent performance and identify opportunities for improvement. In addition, you’ll want to measure other KPIs to ensure your bot is performing (see below).

Training conversational AI can be a full-time job for the first few weeks or months. But even once it’s good enough to launch, you’ll still need to maintain and train your bot in an ongoing manner – forever. That might feel like bad news, but really, all this means is that you need to think of your AI assistant the same way you’d think of a human agent: always growing, always improving, and never finished.