The right combination of integrations can power your Automated Customer Experience efforts, helping you scale thoughtful, proactive, personalized brand interactions across the entire customer journey.
Best practices for deploying generative AI in customer service
Businesses are constantly seeking ways to improve the customer experience, and all eyes are on generative AI; 72% of respondents in a recent Adobe survey that polled both consumers and CX professionals believe generative AI will improve their experiences.
Companies with their finger on the pulse and their eye on the future are looking for guidance on effectively implementing AI and automation. This is where we come in.
As Customer Success Managers at Ada, it’s our job to work alongside customer service teams to ensure that they are maximizing the value of their automated customer service strategies with AI.
In this post, we’re revealing the best practices we use for deploying generative AI in customer service, following three critical (and chronological) steps. You’ll learn:
- How to automate your most common FAQs using a generative approach
- How to integrate across systems to power personalization and more complex use cases
- How to elevate generative AI and continuously improve automation using analytics and insights
Step 1: Automate your most common FAQs with a generative approach
There’s an art to how you build, deploy, and grow automation over time. This is most often referred to as your automation strategy, but you could also consider this automation maturity.
Most often, customers come to us with the urgent need to handle their most common inquiries. We call these questions the “low hanging fruit,” they’re better handled by automation than agents and are easy automation wins. Some examples of this would be specific information that can be found on your website, how to troubleshoot and reset a password, or the status of an order.
If you can relate, chances are you’re pretty eager to get this off the ground. With generative AI, you don’t have to waste time manually building out answers to FAQs — the AI can crawl your support documentation and generate these answers for you, accessing and leveraging the investments you’ve already made.
But first, you need to mitigate the risk that your AI will surface irrelevant, inaccurate, or harmful responses. After all, the key to achieving the most Automated Resolutions — along with resolving the customer’s problem without the involvement of a human agent — is ensuring the interaction is relevant, safe, and accurate and can be achieved in a matter of minutes with the right AI-powered automation platform.
It all starts with your Knowledge Base (KB). For many this is Zendesk or Salesforce, but the automation company you need can integrate with any KB — even a custom one.
Start by ensuring all the information in your KB is accurate and up-to-date. Then, you’re going to train the AI to use it effectively. If you’re new to knowledge management, there are best practices to follow to make the most out of generative AI for customer service.
Here are the cliff notes and some additional ways you can get automation off the ground quickly with AI.
Connect to existing knowledge sources
Before you connect AI to your KB, ensure your support documentation is centralized and works as the single source of truth for your agents and your customers. Generative AI works as an extension of your documentation; if your documentation still talks about last year’s product, so will your bot. While this may seem time consuming, thoughtful planning and preparation of your KB architecture upfront will save you a lot of time in the long run.
Here are a couple tips we share with customers as they get started:
- Ensure the categories in your KB architecture are mutually exclusive. This means that no two categories contain knowledge that overlaps, ensuring that you have a single source of truth for that information.
- Ensure the categories are collectively exhaustive. All together, the categories should cover every piece of information that your customers need to know or may ask about.
While you can still have specific troubleshooting documentation for your agents to use, most of your investment should be in your KB — this is how you build trust with your customers that you can actually solve their problems.
Create and maintain a brand persona
When using generative AI to create content, it's crucial to establish a consistent brand persona that aligns with your company's values and messaging. For example, if your FAQ and agents use emojis and exclamation points to convey excitement and build connection with your customers, your bot should too.
Some questions to consider when building out your bot persona:
- How do they speak? Is their tone friendly, plainspoken, playful, or sophisticated? Do they use emojis?
- What’s their name and visual representation? What types of visuals do they use?
- What answers do they have for your customers?
Once you have this foundation, you need to apply it — everywhere. That means infusing the bot persona into all your content and making customizations to your bot’s appearance. We’ve found that two or more bot customizations can boost engagement by 12-15%. It’s a no-brainer.
Speed up content creation
One of the most impactful ways you can use AI to get off the ground quickly is to use it as a writing assistant. And we frame this as an assistant for a reason.
While AI is an excellent tool to speed up the process, you’ll always want a human-in-the-loop. A bot builder or manager is integral to the process — they’ll review what’s been generated by the AI to ensure it’s safe, truthful, and helpful. That feedback will actually further improve the AI model.
Now, you may be thinking, “I don’t have someone on staff with AI or conversational design skills and experience.” Here’s the other major perk about using generative AI in customer service: advances in AI are making LLMs much more accessible to people who may not have a background in automation.
LLMs build conversational design best practices into their responses and the content it’s generating, making it a teaching and easy detection tool for automation building. What exactly does this look like? A bot manager could simply type in a few bullet points and the AI will use the built in conversational design best practices to reformat them. And it can do this according to the specific channel they’re speaking on.
With AI developing the first drafts and sourcing information from content you already have, you can accelerate the pace of building automated flows and getting them in front of customers to start making an impact.
Step 2: Integrate across systems to power personalization and more complex use cases
In order for AI to automatically resolve customer problems, it needs to be connected with other systems from across your organization. Bots need to do more than just provide information — they need to take action on a customer’s behalf.
This requires a tool that integrates with third-party systems, can access and process the information the bot gathers — like customer or order data — and trigger actions like preparing returns or resetting passwords.
Figuring out which third-party systems to connect to, and to what end, is what we call developing an API strategy. Here are some tips to get your API strategy locked in.
Ensure your CX team has access to a technical resource
One of the most common blockers we see when a customer introduces automation in their customer service is a lack of technical resources to help integrate the automation platform with other systems.
Have a technical resource available to help pull information like account balance, order status, account type, and so on. This allows the AI to take more action, like processing account upgrades, changing an address for an order delivery, and reviewing the account status.
Ensure your systems can give AI the information it needs
Your customer service automation should push and pull information across your technology stack to ensure customer data flows seamlessly through your platforms. This ensures customers can take action through your bot. Here are some steps to follow:
- Map out which actions you want your customers to be able to make in your bot. These decisions should be based on data — this will help you reach the highest number of Automated Resolutions as quickly as possible.
- Dig into where this data lives today (which vendors hold this information?) and audit their APIs with the help of a technical resource. Evaluate what options are available to pass information from one system to another.
- Once you have a clear understanding of the actions you want to enable in your bot and how to achieve them, prioritize deployment of each API integration. Again, this should be informed by data and potential ROI.
Step 3: Elevate generative AI using analytics and insights to continually optimize automation
To maximize the impact of your AI and increase Automated Resolution, it's crucial to create a center of excellence that focuses on continually improving your AI strategy. Over time, you’ll want to progress from simply measuring resolution to improving it.
Generative AI gives you faster, and more accurate, insight into how you can do this — pulling in information from multiple data sources to give you an overview of your customer data without duplicating effort or maintenance.
It can also create new answers to unforeseen questions dynamically — with no manual training. It gives you a smarter, more humanistic, and more resourceful chatbot with less effort to build and maintain. But there are still some best practices to take generative AI to this level.
Remove knowledge silos
Humans are a critical part of your customer service automation strategy. Key sources of expertise should be involved as you look for areas to elevate your AI. Great examples of this are star agents who can be promoted to help derive insights and manage the AI to contribute to its success. These resources should be accessible to the bot manager to ensure their impact is amplified.
To translate institutional knowledge from human expert to AI:
- Identify the experts
- Set up routing technology to connect them when needed
- Set up the tooling that allows them to offer their expertise to train your AI
Analyze and optimize
Automated conversations are a goldmine of information that your support team can use to improve documentation, and AI can quickly understand heterogeneous data compiled from different sources and distill it into useful insights. With instant reports on customer insights — from every conversation, complaint, or suggestion a customer makes — feedback will finally stop getting lost in translation. And it can increase CSAT and NPS scores.
With generative AI helping teams build, launch, and maintain automation faster and with less effort, customer service teams can spend more time on this — analyzing and optimizing automation. It’s even giving employees the opportunity to evolve their careers in customer service.
Your bot manager should be involved, if not spearhead, content innovation and audit initiatives. But their scope of ownership will evolve as generative AI is adopted across your organization. To ensure their focus is aligned with your automation strategy, we recommend following these principles:
- Transition from policy and content review to analysis and insights. As your CX strategy matures in its adoption of AI capabilities, ensure your bot manager transitions their time spent in the bot towards data analysis, review, and insights. Budget a total of 8-10 hours a week — five hours for data analysis and two to three hours to derive insights.
- Make your bot manager the bridge between cross-functional teams. They should work with content, product, analytics, revenue, and any other relevant teams to stay aligned. Cross-functional exposure is key to continue building a strong business case that supports your CX strategy across your business.
- Position your bot manager as a trailblazer. This is an excellent opportunity to shine a light on your in-house customer service expert. Your bot manager will have a unique and powerful understanding of the performance of the business through the customer’s lens. Use this to your advantage — AI can generate reports that give your team and the C-suite clear visibility into what’s working and what’s not. As the bot manager establishes themselves as the subject matter expert on AI — bringing back learnings, operational enhancements, and strategic guidance — they’ll get closer to the C-suite, with their insights reflected in executive team reviews and board materials.
While deploying generative AI for customer service can greatly enhance the customer experience, while improving efficiency for the business at unparalleled speed, it’s not a one-and-done effort. Encourage your team to stay up-to-date on the latest AI advancements and best practices and incorporate these insights into your ongoing strategy.
Rafael Silva is a Senior Customer Success Manager who loves learning how to use new technology to help solve the hard problems. Prior to working at Ada, Raf implemented payroll software and was a karate coach. He represented Canada as a national team athlete for six years. Andrea Freixa is a results-driven professional with 10+ years of experience in SaaS organizations and a proven track record of delivering business outcomes for customers. A skilled leader and strategist, she excels in building and leading CX teams in high-growth and startup environments.