Balancing AI Automation and Human Touch in Customer Experience

Zendesk customers reveal their tips for automating CX with a human touch

Increasing automation of customer service creates a quandary. How do you improve speed and efficiency without dehumanizing the experience for the people involved? As businesses adopt the latest AI enhancements, they are able to resolve a growing volume of customer issues entirely automatically -- but while removing the human agent may be faster and cheaper, it can result in an impersonal experience that is out of step with what customers expect from a business. At Zendesk Showcase in London last month, the CX vendor's customers spoke about how they keep their AI agents on brand, and I caught up with the company's CTO, Adrian McDermott, who is keeping a close eye on the growing role of AI in the customer experience. He tells me: This becomes more important as we go beyond automation as a 'party trick' to automation as a platform. So many of our conversations don't actually involve humans, with our customers. We need to make sure that the customer is actually having a good experience, it's low-friction, and it's enjoyable. Speed and convenience is a big part of that good experience, which is where automation excels. At Zing, an international payments venture launched in January by HSBC to take on the likes of Revolut and Wise, the decision was taken to offer customer support entirely through digital channels, with no inbound voice option. This meant ensuring that the messaging service was never more than two clicks away throughout the mobile app, with 86% of customers choosing this route over the alternatives of email and webform. Customers can request an outbound call, which happens in 3% of cases. Speed of response proved crucial, as Sara Peacock, Head of Customer Service and Operations at Zing, explains: Having no voice can be seen as making it difficult for customers to contact us. So we wanted to counter that. We don't want to make it difficult. We want our customers to know that if they need us, we're there for them to answer them... We set ourselves an extremely ambitious target of giving a first response to our customers in less than two minutes. With an 86% satisfaction rating on all inbound support tickets, the approach seems to be working, supported by a large library of carefully designed help articles for those who want to self-serve. Zing has also taken care to ensure their agents can express their personality in the chat, to add a human touch to the interactions. She goes on: We've spent an incredible amount of time and effort developing a tone-of-voice training for our agents, and the ability for them to be able to also add their personality to the chats they are having with our customers. If you ever happen to chat with one of our agents, you'll notice that they've created a digital emoji which they feel represents themselves. The customer can see that when they chat with us in our window. We also give them quite a lot of creative licence, which they love. They have the ability use emojis in the messaging window to really connect with our customers. And actually, what we tend to see is that customers start to mirror that kind of behavior. At beauty brand Charlotte Tilbury, the response time via chat is now within 30 seconds, and all tickets are answered within two hours. AI will help improve that further, as Harminder Matharu, its General Manager of Digital, explains: We don't see AI as a replacement for our agents, we've never seen it like that. We don't even see AI as a tool. We see AI as an actor within our flow. We see it weaving into the processes that we already have. That's really important. So with our agents, how can we make it more efficient for them to be able to respond to tickets? Can we take the incoming content that's coming through our 'Contact us' form, pre-populate a response, that an agent then simply needs to vet and hit send? When I was talking about a two-hour handling time before, imagine we can do this now in minutes, with a touch from an agent. Zero-touch response handling is also on the agenda, but it's important that these processes still stay true to the brand's values. He goes on: The kinds of queries that we get on a daily basis, we want to be able to handle that with the right empathy. And we want to be able to handle that with the right emotion. But if we can avoid our care agents having to handle something that can be pretty much self-served, that's where we're headed. Storio Group operates eight different photo processing and products brands across 14 European countries. Since its agents only operate during business hours, it was really useful to be able to add a chatbot that could be on hand 24/7 to provide automated support and help. Often, its customers are sorting through their photo collections over the weekend or in the evening, precisely when its human agents aren't available. When introducing Phoebe, its chatbot, Storio decided to give it exactly the same training and objectives as any human agent on the team. Wayne Grimshaw, Director of Customer Service & Shipping, explains: We built a whole operating model, and a whole competency, around who was going to look after Phoebe? The bot needs exactly the same support that an agent needs. What's the bot's objectives? Does the bot have a performance review? QA? All of those things, we said it's absolutely the fundamental needs of this newest recruit. And it really was quite fun, and it worked. The agents treated her as the newest recruit and the newest member of the team. Even in team meetings, it's like, 'Who's going to tell Phoebe that information?' when they came up with a new piece. Analyzing the bot's interactions has provided new data to help refine the voice of the agent and further improve performance, as well as driving better customer journeys. Several acquisitions by Zendesk over the past year have been designed to help customers fine-tune their use of AI and automation in CX operations. Zendesk QA is able to monitor every interaction and ensure that both human and automated agents are setting the right tone in interactions and responding appropriately to the issues customers raise. The acquisition of Ultimate in March expanded the range of customer support automation, while investments continue in the development of CX-specific LLMs. Last month, the company also announced the launch of Zendesk Ventures to invest in AI start-ups in the CX space, such as PolyAI, which provides a more conversational, natural language experience when handling inbound voice calls on topics such as order tracking and delivery updates. It's evident from hearing the customer stories at last month's event that most organizations are still in the early phases of harnessing AI for customer service and support. While it remains in this early phase of learning what works best, careful monitoring is essential to ensure that end customers don't lose confidence because of poor experiences. Zendesk CTO McDermott tells me: Customer service is all about trust. What we're doing is helping -- through post-facto audits and in-process changes and driving change through human-in-the-middle -- that's getting people comfortable with the level of automation that works for them and their users. Customer support and service is one of the earliest adopters of AI in the enterprise and it's clear that Zendesk's customers are eager to take advantage of what it can offer. But it's important not to get too enamored by the technology and remember that previous attempts at automating the customer experience have often fallen short of expectations. AI may yet prove its worth, but it's early days and enterprises must proceed with caution. QA testing and detailed analytics will be essential to ensure that the customer experience really is getting better.

How your business can best exploit AI: Tell your board these 4 things

How can organizations gain a competitive advantage from emerging technologies like GenAI and machine learning? Four business leaders share what they've learned. Emerging technologies promise big benefits on paper that can be tough to realize in practice. Research from the Capgemini Research Institute suggests the adoption of generative artificial intelligence (GenAI) is still at an early stage, with nine in 10 organizations yet to scale these nascent projects. However, boards exert increasing pressure on CIOs and their teams to gain a competitive advantage from innovation. So, what is the key thing business leaders have learned about AI so far? Four business leaders share their tips. Also: Time for businesses to move past generative AI hype and find real value Miguel Morgado, senior product owner for the Performance Hub at Eutelsat Group, said his firm's use of AI and machine learning is related to outage predictions and root-cause analysis, such as the effect of weather on a satellite dish. These explorations into emerging technology show the importance of high-quality information. "We do lots of tests with real data," he said. "And validating the models is very important. Because if you don't have an accurate model and then use it, it will be a case of 'garbage in, garbage out'. It's important to have a good data set." Morgado said his business is lucky -- the satellite company collects billions of rows of data daily for various use cases. However, the firm ensures this information is applied safely and effectively. "We still need to test models and the results over and over again until we validate the approach," he said. "The results won't be perfect -- there will always be a degree of imperfection. But it is an indication." Also: When's the right time to invest in AI? 4 ways to help you decide Morgado told ZDNET that other companies should ensure a skilled expert stays in the loop and communicates the significance of outputs to business colleagues. "Then you can get that person to say these results indicate a particular value or could potentially be taken as guidance," he said. "So, it's always, in the end, the user who decides if they trust the AI or not. My advice to other people is to ensure there is always a human element to your AI results." Ulf Holmström, lead data scientist at Scania Group, said his company is exploring how it might use AI for internal support processes. The company is investigating how to make the most of Amazon Bedrock and is keen to explore how it can use some of Snowflake's tools, including Cortex AI. Like other business leaders, Holmström pointed to the importance of underlying data and technology concerns. "Call it whatever you want to call it, but you need to have trust in data and infrastructure and governance, otherwise you can never scale, and you'll only do proof of concepts. And like other organizations, we need to put stuff into production." Also: Agile development can unlock the power of generative AI - here's how Holmström told ZDNET the good news is the implementation of technology seems to get easier all the time. Access to technical knowledge has been democratized through technologies like the cloud and generative AI. However, Holmström said new processes must be introduced for users to make the most of emerging technologies. "If we're going to implement AI in production, that comes with implications -- and one of the big implications is that we need to change our way of working. It means new business processes and a new type of organization," he said. "We need to have new skills and we need to change our way of working. That shift is difficult in all organizations, especially in legacy enterprises. But without that transformation, AI will never fly." Holmström said senior executives must drive the move to this new way of working. "Top management commitment is crucial," he said. "You can never do an AI transformation bottom-up. It must come top-down." Anastasiia Stefanska, data analyst for analytics and AI at holiday firm TUI, said it's important to think about how we can turn the huge quantities of data we collect into high-quality information -- and that task requires a recognition of human biases. Like other professionals, Stefanska recognized that ensuring your organization has high-quality data is a prerequisite to any successful AI project. However, it's not the only key issue -- smart professionals will ally a focus on data-quality concerns with a consideration of real-world biases. "AI is a simplified reflection of the real-world reality in which we live," she said. "On one hand, driving AI adoption with data quality in mind is paramount. However, a critical eye on the status quo of the real world can allow us to go beyond data quality. We can think in the direction of having an opportunity to solve the biases that are deeply embedded in the real world." Also: 4 ways to help your organization overcome AI inertia Stefanska told ZDNET how TUI uses the Snowflake platform to consolidate enterprise information and create a digital platform for data-led change. As part of this work, Stefanska and her colleagues watch how data is used and exploited. "That's why we say at TUI that the human eye is important. We acknowledge that the biases are there in the real world around us," she said. "My main message would be, 'Yes, data quality is important, but have a holistic view on whether you have a chance to convert a quantity of the existing data into something with a new quality.'" Richard Wazacz, CEO of foreign exchange specialist Travelex, advised other professionals not to walk before they can run. He recognized there's huge hype about AI. However, the fear of being left behind must not color professionals' judgments. "At the moment, we're not going to be early adopters of AI," he said. "But we will use AI when the case study for how it's helped others has been proven." Wazacz told ZDNET that his extensive business experiences, including as director at Octopus Energy, helped him develop a strong awareness of the times when emerging technology can play a key role. "I worked at Octopus and they've successfully used AI to help improve customer services," he said. "A lot of customer questions are answered through AI. Do I think there's an option for us to do that now? Yeah, and because they've proven it can be done, it's less of a risk." Also: Do AI tools make it easier to start a new business? 5 factors to consider Wazacz said his approach at Travelex is to draw on consultancy expertise from Mesh-AIto ensure digital investment is directed to the right places. Mesh-AI has helped Travelex establish a cloud-based data platform, with an initial focus on real-time reporting. The company will move into other emerging areas when the time is right. "I'm being very narrow in the scope," said Wazacz. "So, at the moment, that's what Mesh-AI is working on. They're excited about what they're doing. They're taking the approach of, 'We'll prove that we can make our customers self-sufficient because we'll win more business.' And that's what I feel like they're doing."