Good customer service is easier said than done. With the promise of AI still being translated into reality, we’ve been wondering: what does the technology have to offer customer contact? And is it actually a technology problem in 2024? We discuss it with experts from Salesforce, ServiceNow, Freshworks and Evolve IP.
According to Jan Verbrugghe, Salesforce’s Senior Director of Solution Engineering, there is a need to elevate the service worker job. “You need to give young staff different tasks than just resetting passwords and arranging refunds,” he said. Instead, they need to do more technical and diagnostic stuff, he proposes. In addition, the work itself has become more difficult, Verbrugghe says. He says employees have to do more and more customization on a per-customer basis. Also, he says, sometimes proprietary offerings are so complex that the customer agent can no longer keep up.
Joran Klarenbeek, Business Development Manager at Evolve IP, acknowledges this. However, he observes that it is hard to hire qualified staff. On top of that, “It’s also not the kind of work that most people want to do for 40 hours a week.” At a time when cost savings reign supreme, investing in customer service is not the obvious go-to option. Still, it’s necessary to do so, and organizations have caught on as well, Klarenbeek argues. “They know they have to do something with AI.”
The non-human side of customer contact also needs improvement. Lubert Meuter, Senior Lead Solution Engineer at Freshworks, analyzes where things often go wrong with the well-known kind of chatbots without GenAI. “Nowadays, with a chatbot, you can get stuck in the programmer’s train of thought.”
The AI promise, concretely
But what does AI ideally have to offer? Surely the appeal in 2024 is primarily the use of AI to cut costs. Jeff de Graef, Customer Service Workflow Leader at ServiceNow, talks about the “reducing of the pain” that needs to happen first. Staff are under a high workload, so the quick win is to use AI to partially absorb and support this. De Graef has found that conversations about this go a lot smoother than about integrating AI within larger processes. A workshop on this can have big results within 2 hours, he says. “In addition, you do open up a conversation within the organization about the future deployment of AI,” De Graef says. Later, other parts of the customer journey may be addressed inside such an organization.
Meuter draws the benefit of AI more broadly, albeit with a critical note. “AI means you get more time for other things. But what do you use it for? Simple use cases yield a higher ROI for customers.” Automating simple work processes is the first step. But what next?
Meuter comes up with the following example: “If a customer call is successfully handled, AI can repeat the steps the employee took. You can then use that again in other conversations.”
AI should not become an IT party
AI innovations are flying around our ears. They are happening faster than organizations are capable of. Klarenbeek argues that organizations often don’t know where to start in the first place. “Many companies don’t even know what the five most common customer service questions are.” On top of that, basic mistakes are often made, Klarenbeek says. “At government organizations, we regularly ask, ‘did you communicate clearly on the website how to make an appointment?'”
The key is to take steps now and increase your AI efforts, Verbrugghe says. “Boards are already getting disillusioned, but they absolutely shouldn’t. AI has actually been a reality check for the maturity of organizations.” They are starting from misconceptions, he says. “You should not adopt AI and try to push a workflow into it. It should be the other way around.” The worst thing you can do is try to build AI yourself, Verbrugghe said. He’s seen plenty of AI nightmares, such as “home-built AI” in which employees can simply request the CEO’s salary and are let loose on all company data.
Verbrugghe: “Customer service has always been the ugly duckling in many business processes. This is due to the fact that the data supporting it is not structured. Most customer service employees are a jack-of-all-trades who know how to solve problems better than anyone else. But that’s often still relying on manual work. The really useful data is hidden in an ERP platform or knowledge base.” And this is while those who have to build the AI systems are not familiar with the exact processes that AI should be able to take over, he says. Meuter adds that employees’ years of experience are also simply not captured. It can be done, as indicated earlier.
AI adoption within organization is usually a little less dramatic, but regularly still problematic, De Graef says. “It is enormously difficult to provide an end-to-end AI infill in which all your work processes are connected. AI is not a silver bullet.” ServiceNow works with the customer to look at use cases where AI effectively makes a difference and where it doesn’t. De Graef says his company is always looking for ways to provide value for the customer by looking for bottlenecks and working from hype to concrete steps and clear expectations. This is not limited to just the front office, mid office or back office, but requires an integrated end-to-end approach he says.
Therefore, AI should not become an “IT fest,” Klarenbeek adds. It must involve all parts of the organizations. “You also have to make sure that all other software packages integrate into your AI solution,” he says.
This is difficult in practical terms because “enterprise customers also have different technology stacks internally,” says De Graef. Whereas for one company it is relatively feasible to integrate AI quickly, to others it is a significant task. Especially if the aftermath of all kinds of acquisitions or an accumulation of products over the years is still visible, Verbrugghe notes. “That’s all brownfield, nobody is starting fresh.”
Not just technical
Preparing for AI is not just a technical problem. “All IT systems are intertwined these days. You have to make sure those integrate with each other,” Klarenbeek says. You also need to know the business processes.
That knowledge is hard to come by at the C-level, an example from Verbrugghe shows. “At a press organization, I spoke with someone who had earned his seat at the C-suite by pointing out how bad the sign-up process for the digital newspaper was. The CIO knew nothing about it being a problem.” And in answer to the question as to why this wasn’t known before, “No one asked the department that even though they are the canary in the coal mine. Organizations are sitting on top of a lot of processes that they don’t know about.”
“Organizations need to ask themselves: what profiles do we already have on board?” said Verbrugghe. That can be an advantage for clients and at the expense of external parties: “The market that BPOs typically move into are low-cost, low-margin repetitive tasks. They’re suddenly getting competition.” These organizations, he says, need to re-energize. At the same time, these organizations themselves are increasingly at the helm, especially the business side, Verbrugghe says. “The IT threshold for building AI applications is dropping. More and more business instructions are going straight toward software development.”
De Graef acknowledges that organizations themselves need to nurture AI adoption beyond the technical issues. He sees an opportunity with younger generations, who are more accustomed to using tools like AI to achieve a goal and complete tasks. “Outsourcing to a bot” is a fairly natural way of working for them, and they can play a significant role in the implementation and adoption of AI throughout the organization, De Graef argues.
Is your organization ready?
Ultimately, IT workers do bear the responsibility for implementing AI. They’re the ones who will run into significant problems first. For example, De Graef points out that organizations simply do not have their business ready yet. Data governance and data quality are regularly substandard. As a result, AI deployments can’t happen overnight, even if the technology is ready. The push for a true enterprise architecture is still some ways off, with AI being a “wake-up call” for them, according to De Graef. “They have invested years in AI technology and AI expertise, but there still isn’t the desired data quality. The technical complexity has not decreased.” The key, he says, is for customers to work with their vendors to look at the bottlenecks to implementing AI.
Meuter mentions that sometimes customers want AI on the front end, focused on the customer side, or primarily internally. In the latter case, the IT department needs to be prepared to take on a new role. “IT is increasingly becoming an internal service provider for AI applications.”
Aren’t some organizations ready for it now? Klarna, for example, moved away from Workday and Salesforce to replace these solutions with self-developed AI. Verbrugghe thinks very few organizations are the next Klarna, even if they think they are. “Klarna is a greenfield with one business process: to handle payments after the fact,” he said. The vast majority of organizations are brownfields and thus should instead rely on their vendors, as Verbrugghe notes. Their existing IT infrastructure must remain, according to Verbrugghe, leaving a company like Salesforce to realize AI on a scale that normal organizations are not capable of.
Meuter does indicate that organizations can look at their departments separately to arrive at the simple use cases for AI. For example, there may be several Klarna’s within larger companies, so to speak.
Conclusion: driving change
2023 was definitely the year when generative AI became “hot.” But real fulfillment was yet to come. 2024, according to these experts, is the year of practical implementations. Nevertheless, ambitions have been curbed by the focus on cost savings.
Organizations have yet to traverse that roadmap, argues Jeff de Graef of ServiceNow. Salesforce’s Jan Verbrugghe also speaks of a “leap forward” that won’t happen until next year. “The time will be right for that, the technology is already there,” he says. To that end, getting to grips with the use case is key, believes Lubert Meuter of Freshworks. He says that some platforms already allow AI adoption right away, but other applications still require preliminary work to clean up data. “A single customer record is still a technical challenge,” he says. Joran Klarenbeek of Evolve IP expects the future of AI adoption to be bright in this. “People who are skeptical are going to see the benefits.”