With AI everywhere these days, the expectations for customer service have never been as high as they are now. Customers expect nearly immediate answers, highly knowledgeable support agents, and ample help documentation.
But the reality is that most companies fall way short of this level of service because they're using outdated and slow technologies rooted in traditional CS methodologies.
But AI is changing all of that, particularly through generative AI fueled tools like chatbots.
Improving customer success with chatbots
There are two main ways to leverage a chatbot for improved CS.
One is to provide highly timely answers through a user-facing AI agent.
The other is to empower your support agents with better information and more context, which allows them to help users more quickly, through agent assist chatbots.
We'll come back to number one later, but let's first focus on number two.
Agent assist chatbots
With the advent of LLMs, a whole new suite of AI-powered tools has emerged to help support agents thrive.
These AI agent assistance tools come in many ways, shapes, and forms, but all of them are aimed at improving the way that support agents operate. You can think of them as an ongoing dynamic resource for agents to solve all kinds of problems.
We wrote about broad real-time agent assistance earlier in May 2024, but chatbots are a particular area of interest lately.
While many of us think of customer-facing chatbots like ChatGPT when we think of an “AI chatbot,” another way that these can help your customer support team is through an internal chatbot that can transform the user's query into meaningful resources for the agent to respond with: correct information, knowledge base articles, and other materials to share and send to the user asking the question. It supercharges their speed and knowledge with access to more primary, secondary, and tertiary information, and allows them to provide the best possible experience.
Why do agent assist chatbots help?
How do these chatbots actually help, you wonder?
Real-time suggestions
One of the primary ways that agent-assist chatbots can help agents is by providing real-time suggestions to customer queries.
We're not quite at the stage yet where AI agents, particularly for voice support, have been replaced by artificial intelligence itself, but there are still ways that the human can get better info, faster, and thus provide quicker and more accurate answers.
For example, there's technology now that can follow the interaction, identify what's being asked, and recommend a course of action, which the agent on the phone can then deploy (if they so choose.) That's really powerful!
Or, the agent can see the user’s complaints, and then simultaneously provide script modifications, courses of action, or links to other resources which the agent can then use their human knowledge to pick and choose from.
Integration with other technologies
Imagine that you are running a support center where agents need to access various systems quickly. By integrating the chatbot into this tech stack that drives their daily workflow, like a CRM and ticketing system, the chatbot can provide higher context.
Instead of giving a generic response, it might recognize the specific iteration or part of the product the user is dealing with and suggest the most relevant information.
This enhances the chatbot's ability to provide accurate and contextually relevant support.
For example, think of a traditional customer support system, which is more like a basic CRM, which lack nuance. It might have some data point and some notes from past encounters with the customer, but these are still static.
Sure, those basic notes can be helpful, but they don't provide full context and emotion and trends. That’s where an AI chatbot can help an agent get caught up to speed.
For example, it might note that a user has had three calls in the past week. They've been negative, and they haven't had their issue resolved. And it might suggest a specific course of action, or an escalation to a manager.
You can see how that's more helpful than a user calling into a support system and the agent having to read all of that themselves while also trying to start the conversation. It might inform the agent (Hey, we tried X before — try Y instead to solve this user’s problem.)
Knowledge base access
Let's face it, even the best support agent is no match for an AI bot's ability to access a knowledge base and search for the right article(s). One of the best ways these agent assistance technologies can help support agents is by connecting them to the right information at the right time, faster. It's a simple matter of speed.
Benefits of agent assist chatbots
What’s the impact of these chatbots?
Improved customer satisfaction
Agent assistance bots help agents do their job better. It's no surprise, then, that when agents perform better, users are more satisfied. Not only are they getting faster time to resolution, but they also have a better rapport with your agent because they feel that they're knowledgeable and solved their problem.
This not only improves direct metrics, like I mentioned, but also their general satisfaction with your entire product and their view of your customer experience as a whole. It will leave a good taste in their mouth. This can lead to higher CSAT and NPS scores, and less stressed out agents!
Increased agent productivity
Luckily, this isn't just a win for the customer and the user; it's also helpful for agents. Because they're getting support to aid in their responses, they are getting faster answers out to their customers, which means that they have time to focus on more complicated cases that require more involvement. It's as if they've been freed up to put energy into things that are truly rewarding and don't have to deal with all the repetitive BS!
How to implement an agent assist chatbot the right way
Before you implement an agent assist chatbot, you need to really understand whether it's needed in your product and how you're going to deploy it. If you have just gotten caught up in the big AI trend of the day, you may add a tool that's actually clunky and not useful in your team’s specific workflow.
What should you look out for?
First, you really want to make sure that it's necessary and that you have a demonstrated use case. Ask yourself:
- What problems do my support agents face?
- Can these be solved by a chatbot that can provide more timely and contextual information?
- If they can, what kind of information or data do you want the chatbot to connect to?
- What will it say and do for the agents? What will it not do?
- What kind of results mean success? What metrics will you use to measure that success?
If you feel confident that you need an agent assist chatbot to assist your users after going through these questions, you can begin to evaluate your options.
Depending on your tech stack and the level of complexity in your business, you might want to look out for what level of sophistication the LLM behind the chatbot has, and whether it integrates with your key support software.
From there, it's a matter of introducing it to your staff, training it, and then deploying it live.
I would recommend you take some time to discuss the involvement of this AI tool with your team. There's a lot of anxiety in the customer support world about the growing role that AI has, and this puts folks jobs at risk or at least makes them feel like they're at risk. You'll want to emphasize that this is a tool to augment and empower them to spend their time on more complex cases, not to replace them – (if that is your plan.)
Are agent assist chatbots the best AI-powered tool to improve your CS?
If you got plenty of funding and plenty of time, then you might be able to add a lot of different AI tools into your workflows. But the reality for many startups is that time and money are short, and that wasted energy on software that drive only incremental improvements and outcomes can be devastating.
I think there's a lane for agent assist chatbots – they definitely can provide value and empower your agents. This can be particularly true if you believe in maintaining your support staff workforce and believe that the product requires a human touch.
That said, one thing we've learned through the development of our Copilot and interactions with our customers is that sometimes, the best customer support strategy is actually to provide answers immediately, in context, at the right time, without a support agent ever getting involved.
Now this isn't a novel concept, certainly.
But it's been recently rejuvenated and reinvigorated by the ability of user facing AI agents to address the 80% of queries that revolve around 20% of topics (yes, CS is Pareto-based.)
This is not only a better user experience, with faster time to resolution from smart and dynamic AI responses, but it also deflects tickets and reduces strain on your support staff in the first place.
A final word on AI agents in customer service
I don't mean to sound too black and white, yes or no.
It's not necessarily an either or world. That'll depend on your org's resources, time, and needs.
You could have a user-facing Copilot as well as a back-end agent assist chatbot that helps your support agents with queries that did not get answered through Copilot and needed a human touch.
However, our vision is to create a Copilot that is so natural and so accurate that users rarely need to go beyond it for support. That's why we've evolved beyond static text answers into an ability to launch in product experiences like product tours and nudges and checklists, all of which offer your team multi-layered opportunities to respond to customer needs.
While the world of customer support may be changing fast, as products emerge and disappear rapidly, one thing is abundantly clear: AI is here to stay, and will play a very important role both on the back end and front end, whether through agent assist chatbots or customer facing AI Copilots, all of which help drive faster service, improved satisfaction, and reduced costs across the board.
