An easy and low-risk way to implement an AI solution is using a copilot (or agent assist) tool.
Since these don’t interact directly with customers, they’re great at helping your team develop AI skills, like how to train an AI solution or how to use AI to work more efficiently.
Here’s an overview of Zendesk Agent Copilot and a few of its competitors.
What are copilot tools?
Copilot tools are AI-powered solutions designed to enhance the efficiency and effectiveness of customer service agents.
These act as "co-pilots" by providing real-time support, automating routine tasks, and delivering intelligent recommendations during customer interactions. They’re commonly referred to as “agent assist” software as well. These aren’t directly customer-facing, the audience is the agents themselves.
Their key features typically include:
Real-time assistance by suggesting responses that can be sent out to customers. These are often based on macros or historical data from your help desk.
Automation of repetitive tasks. They often include API-driven actions, like processing a refund or reducing the steps in updating a customer’s email address.
Rewriting or adapting responses. That can mean expanding a set of bullet points into an email response or adapting the tone of an agent’s draft to make it more friendly or professional.
Summarizing ticket content to help customer support agents quickly catch up on the interaction.
In essence, they should help your team work faster by making information more accessible or tweaking their responses. Sometimes they’re also built to smooth out some of the more cumbersome processes you agents have to do regularly.
How to measure the ROI of copilot tools
We’ve written extensively about how to measure and track the impact of software like this on your team. These are the most important KPIs to keep in mind:
Average Handling Time is the first place you’d ideally see an impact from an effective copilot tool.
Number of tickets resolved per agent should naturally increase, if AHT goes down.
You might see an increase in Customer Satisfaction, if the above changes lead to faster, more accurate responses, or a faster resolution time.
Faster time-to-competency is a common outcome. Many copilot tools are especially helpful during the onboarding process.
Typically, the biggest challenge with realizing any of these is that it requires high adoption across your team. That might be challenging, especially for the more tenured and experienced agents who already have very comfortable and ingrained processes.
In these cases, investing more in advocacy, getting buy-in across the team, or going for a staged implementation aimed at your newest agents first might be helpful.
What’s Zendesk Agent Copilot?
Zendesk Agent Copilot is the native agent assist feature that Zendesk is rolling out and testing.
Its capabilities are very closely aligned with the description above. The goal is to reduce the time agents spend on routine tasks, enabling them to focus on more complex customer interactions. The way it does this is very simple:
It proactively suggests replies and actions.
These replies have to be predefined and inputted by Zendesk admins.
Agents can choose to accept, edit, or send these directly to customers.
The current iteration of Zendesk Agent Copilot is quite limited. It’s available for customers with the Advanced AI add-on. That said, they’re focusing on a few very specific niches for now:
Customers who have enabled intent predictions. Note that the intent prediction has been trained on industry data and your business should fit in one of the specific industries Zendesk has listed for it.
Customers who operate in the e-commerce sector, use Shopify as their primary platform, and are open to using the Copilot for processing order cancellations and refunds.
While the Zendesk team is currently working on opening access to other industries and use-cases, this immediately makes Zendesk Agent Copilot not relevant for massive chunks of their customer base. While Shopify is one of the largest providers in the e-commerce space, it’s also the one integration that a huge number of AI providers are targeting, so there’s some intense competition here.
Its main advantage is that this level of focus should lead to a reliable and secure integration. Given that, it also has a few other disadvantages:
Zendesk Agent Copilot is trained separately from the existing content on your help desk. For example, if you already use Zendesk AI agents, you could save yourself a ton of training effort by integrating it with Agent Copilot.
Some of the training could also be based on historical data in your help desk or existing macros. Since so much of your customer interactions are already in Zendesk, it’s a shame that they focused exclusively on predefined replies by admins.
Many customer support teams have robust internal documentation. Connecting this to Zendesk Agent Copilot would also improve the quality of its suggestions.
Three other copilot tools on Zendesk
Agent assist or copilot tools are – unsurprisingly – booming.
Here’s a selection of alternative copilot tools available on Zendesk that are worth looking into.
Stylo Assist
(source)
Stylo Assist, like SwiftCX, is also powered by ChatGPT.
Its key features include:
Drafting replies. It uses the existing knowledge base, previously resolved tickets, and predefined macros to tailor replies.
Language translation. Messages can be translated to and from any language.
Standardized agent tone. It can help you maintain a consistent tone across all customer interactions. This feature can be helpful if hitting a consistent tone is a challenge in your team, but could also result in interactions needing a little more personalization.
Summarizing tickets. It summarizes the content of the ticket, which is especially helpful when they’re lengthy.
Researching answers. One of its best features is that it can pull together relevant resources and list them for you–so your agents can always refer to the original source and review it before using the answer.
Shopify integration. As mentioned earlier, a Shopify integration is one of the most common features AI solutions are offering now to automate tickets about order status.
Generate responses via help articles. This one’s a unique feature. If the AI tool doesn’t automatically identify the connection between the ticket and the article, agents can select the article and generate a reply based on it.
Advanced spam detection. Zendesk has its own spam filter but it does have room for improvement. Stylo may work a little better for that.
Stylo is significantly more powerful on the copilot side than Zendesk Agent Copilot is. They have a public demo so you can play around with it and see what it looks like.
Pricing starts at $15 per user per month. The Zendesk Advanced AI add-on costs an additional $50 per user per month (on top of the regular Zendesk plan), making Stylo a much more attractive alternative.
They’ve only just started working on a customer-facing AI solution, so if keeping your tools consolidated is a priority and you want to implement an AI agent as well, they might not be the best option.
SwiftCX AI Copilot
SwiftCX offers the trifecta of AI-solutions: a customer-facing agent as a chatbot solution, a copilot tool, and AI-driven insights that analyze customer data.
Its AI Copilot covers many of the features that the Zendesk Advanced AI add-on does:
Suggested replies are generated based on ticket history, macros, help center articles, and 3rd party sources like Notion or Google Docs, where many CS teams host their internal documentation.
It also analyzes customer sentiment and summarizes the ticket, so that information can be used to speed up the response or prioritize tickets based on sentiment.
The main disadvantage of the sources above is it’s easy for suggested replies to be out of date because they’re so heavily influenced by historical data. This type of setup should make it easier for your team to maintain documentation (e.g. add a new macro, change an internal help article) when they realize the AI copilot is pulling the wrong information.
That said, SwiftCX has a very sparse website with no pricing information and very limited screenshots to see what it looks like in action. That makes it hard to judge what the training process looks like.
It also looks like there’s no API integration capability whatsoever, which means there’s no way to include actions. That’s especially limiting if you are an e-commerce company, since those integrations are so common.
Easy Ticket AI
(source)
Easy Ticket AI is very similar to both tools listed above. It does many of the same things they can do:
Automatically generate responses for agents.
Recommend relevant articles and macros by learning from previous interactions.
Summarize the ticket and analyze customer sentiment.
Provide multilingual support by translating the ticket and its response.
Help agents maintain a consistent tone by rewriting their responses.
Its best feature is an AI assistant that your agents can interact with, rather than having to search internal documentation for the relevant answers. Because it’s already embedded in Zendesk, it reduces the amount of tab-switching and searching that your agents need to do.
It’s a very simple agent assist tool that doesn’t natively offer any additional integrations to other software. That said, they’re versatile enough that you could contact them for support with developing additional features that are relevant for you.
Pricing starts at $20 per month per agent, which is a reasonable starting price.
Efficiency gains through incremental improvements
Helping your CS team work more efficiently is rarely dependent on finding the perfect tool that will have a sweeping impact. It’s rather about making a series of incremental improvements over time that each have a small impact.
As a support team, developing your tech stack requires thoughtful adoption of tools that address specific pain points. Taking the time to experiment with each one is usually the best path forward.
At Swifteq, we’ve tried to embody this approach. We’ve developed a series of Zendesk apps that automate different tasks or assist your agents in different ways, via translation or improving the Zendesk search, to help you improve the pain points your team might be struggling with.
Written by Nouran Smogluk Nouran is a passionate people manager who believes that work should be a place where people grow, develop, and thrive. She writes for Supported Content and also blogs about a variety of topics, including remote work, leadership, and creating great customer experience. |