ChatGPT is the support automation wunderkind. Harnessing the power of large language models or LLMs, OpenAI’s gen AI chatbot changed hearts and minds (pretty much single handedly) about bot-to-human interactions overnight.
In the excitement around the rapid pace of change — and the rush to adopt this new technology — the value of conversation design has been somewhat sidelined within the support automation world. But when it comes to creating the best experiences for customers (and supporting agents in their roles), this all-or-nothing approach to generative AI sees everyone lose out.
Generative AI shouldn’t replace conversation design. Instead, we should be embedding gen AI into existing conversation design tools: bringing customers the best of both. We’ll take you through how generative and conversational AI work together. And our CEO, Reetu Kainulainen, will weigh in — sharing why he believes this approach is the future of support automation.
Why should you choose generative AI for customer support automation?
For anyone who has played around with ChatGPT (and let’s be real, who hasn’t?) the customer-facing benefits of gen AI are clear: natural-sounding conversations, instant answers, and 24/7 availability. Memories of clunky, frustrating, old-school chatbots have quickly faded — and customers now expect the fluid conversational experience that gen AI bots provide.
As well as elevating the experience for customers, gen AI has made it easier than ever for brands to get started with CS automation:
“Generative AI makes automation significantly more accessible, because large language models (LLMs) already have the context of our reality. This means they don’t need training, they just need guidance.”
- Reetu Kainulainen, CEO & Co-founder, Ultimate
This guidance might be a brand’s help center articles, their knowledge base, and documentation on internal processes. The challenge then for automation platforms (like Ultimate) is making sure the generative AI model only answers within those guidelines.
And then there are the productivity and efficiency benefits that support teams see (but more on that later). For now, here’s a quick run-down of the top generative AI use cases in customer support:
- Democratizing bot building — generative AI has made bot building accessible to the masses, no IT resources or tech expertise required
- Deflecting FAQs — with a gen AI bot connected to your help center, you can serve all of this information around common queries to your customers in a human-like way
- Analyzing and summarizing support questions — support teams can use generative AI to understand the content of long emails or chat conversations and create a concise summary to get agents solve cases more efficiently
- Generating replies — gen AI can provide agents with suggested replies to any incoming message, meaning speedy resolutions and happy customers
- Keeping conversations on-brand — gen AI bots like UltimateGPT can instantly adopt your brand persona and tone of voice, keeping experiences consistent across channels (check out the demo below to see this in action)
And this is just the beginning for generative AI in support automation. As this relatively new technology continues to mature, more and more gen AI use cases will appear.
Why should you choose conversation design for customer support automation?
Where generative AI is flashy and (can be) unpredictable, conversation design is sturdy and reliable. Conversation design-based bots are tried and tested — and have been helping global brands to level up their support for years.
You can build out a process and know that it will run smoothly time and time again. Leading automation platforms (cough, like Ultimate) will allow you to design conversation flows that include custom-built backend integrations with your other business systems — like order management software or a payment processor — to automate complex queries end to end and increase your automation rates.
Read up on conversation design best practices
This is how Reetu describes it: “Conversation design is your ability to take control of the conversation and the model, to run a more structured process. So if you need to qualify a customer based on certain criteria, or understand when a customer is coming from a certain region, or if someone is a VIP versus a free trial customer. Maybe for a VIP, you want to ask further questions and then route them to an agent — so now you’re in control of the process.”
Here are the key benefits of choosing conversation design bots to automate your support:
- Making use of historical data — if you have a large body of historical support data, conversational AI tools can analyze this to see which questions your customers ask most frequently (so you know exactly what to automate first)
- Providing full control over conversations — with conversation design, your support team can decide on the precise wording of a dialogue flow, at every step of the automated interaction
- Increasing automation rates — if you’ve already launched a gen AI-powered help center bot, you can then use conversation design tools to build out more advanced flows and increase your automation rate
- Extending self-service — by connecting to your back office systems you’ll give your bot access to the same information as your human agents, allowing customers to resolve more complex cases without any need for agent involvement
“With generative AI, it’s hyper conversational — but these models won’t consistently run a process for you.”
- Reetu Kainulainen
Some automation providers are offering entirely generative or entirely conversation design-based bots — but in an ideal world you shouldn't have to choose. And with Ultimate’s AI-powered automation platform, you don’t have to.
How is the hybrid approach changing the world of customer support?
One of the biggest changes Reetu sees is how generative AI allows support teams to achieve results with automation much faster. With a gen AI bot, you can simply connect to your existing help center. The bot will then analyze your support articles and start resolving FAQs in the most conversational, human-like way.
“It makes the experience more conversational and it increases automation rates because the generative bot can handle the long tail of customer queries. Whereas in the past, it took much more effort to get to a good automation rate and a good level of customer satisfaction.”
When automation rates rise, not only does this increase support efficiency but it takes more of the tedious, repetitive tasks away from agents. This means they have the time to focus on cases that require their problem solving skills, empathy, and creativity. “The agent-facing use cases are unlocked with this technology. Generative AI helps to augment agents, which means we start to see a world of stress-free support teams,” Reetu says.
Using generative AI to provide accurate summaries or keep conversations on-brand helps drive support efficiency (as well as increasing both agent and customer satisfaction). Add the customer-facing benefits to this, and you might be wondering why we’re not going all in on gen AI. Here, Reetu explains why:
“Purely relying on gen AI gives you ease and freedom and flexibility, whereas conversation design gives you control. So these are very complementary technologies. Generative AI is the final missing piece to make support automation very human-like, but without conversation design you lose control.”
Shaping the future of customer service
We’re going to hand it over to Reetu to give the final word on this topic. Here, he discusses how using generative AI and conversation design together is the future of customer service automation:
“I think it’s genuinely exciting how conversational gen AI is off the bat. So just pointing it at your knowledge base and giving it a tone of voice makes it already very conversational. But the thing with these knowledge base bots, is that they’re answering FAQs. They don’t capture all support queries — so you still have to connect them to your business systems.
I think what’s really exciting is where we can start using large language models within the conversations. So instead of outsourcing the whole conversation to the LLM and hoping for the best, you’re using generative AI in some parts of the conversation.
For example when a customer asks “where is my order?” you can instruct the LLM to collect certain pieces of information from the user, and then hand this back over to the dialogue flow. Then you can do an API call for instance, and based on the API call you might have four different outcomes. In three cases you might hand back over to the LLM to give the customer a status update in a nice tone of voice. But in one case — say for a delivery issue —you can actually take back control so you can create a priority ticket for this or escalate to an agent.
I think the future will be this blend of generative AI and conversation design — where it’s easy to build hyper-conversational use cases, without losing control. This is what we’re excited about, and what our customers are excited about.
I see a world where you will put the generative AI solution live, start seeing how it’s behaving, and then start taking control of the sections of the conversation that you want to. This means the focus moves away from having to worry about intents or AI model training and the plumbing of a solution — and allows companies to focus completely on improving the experience for end users. So I think we’re truly going through a renaissance of AI right now.”