Interview with Wolter Smit of TOPdesk
Wolter Smit, CEO at TOPdesk, discusses why successful AI adoption depends not only on better technology, but on capturing organisational knowledge, choosing the right use cases and knowing when human judgement still matters.
AI adoption is moving from experimentation to a strategic priority, but technology alone will not determine whether organisations capture its value. The knowledge that sits with employees, the processes built around it and the judgement required in uncertain situations all shape where AI can genuinely make a difference. Wolter Smit, CEO of TOPdesk, explains why organisations need to capture and structure their internal knowledge, identify use cases where AI can deliver meaningful value and gradually increase automation while keeping human judgement in the loop where context, empathy and risk demand it.
Looking back from starting with a help desk tool in the 1990s to building an international business, what turning points most shaped your approach to leadership?
I started out in the age of DOS and, along with my colleagues, wanted to find a pragmatic solution to what appeared to be a simple problem facing customers – In other words, keeping track of what you’re doing and ‘taming the chaos’.
One of the biggest turning points for me was realising that, although a lot of what we do is technical because it’s IT, the way it is delivered is very human. Quite early on, I saw that people work better when they are engaged.
Although a lot of what we do is technical because it’s IT, the way it is delivered is very human.
So, we built a company with high levels of human engagement. If your own people are engaged, the customers will be as well. I see that reflected in the positive things customers say about our people at events.
If you have people who enjoy what they are doing, that leads to satisfied customers, and the rest will follow. You will grow and be profitable, partly because customers will talk to one another. That is as true in the age of AI as it has ever been.
What lessons would you share with leaders trying to stay relevant through the current period of major change driven by AI?
I’ve seen huge excitement around technologies such as 3D TV, the metaverse, blockchain and edge computing, only for the initial hype to subside.
But AI is different. I like to program in my spare time, and I see the power of AI in software development. Speaking to others across the industry, there’s a real sense that this is one of the most exciting periods of technological change we’ve experienced in a long time.
Coding models are trained on masses of data, but this approach does not work so well in the service desk environment. For AI to be effective, you need to capture the knowledge inside your people’s heads and make it available across the organisation. You need to turn that information into something AI can work with that is specific to the needs of your business. That knowledge can become a real differentiator.
Before introducing AI, what service management fundamentals should organisations have in place?
Organisations need to collect the accumulated human knowledge in the organisation and institutionalise it. It could be as simple as creating a knowledge base. That sounds easy, but it is relying on human nature – and nobody really likes populating a knowledge base. We can see from our own work that few organisations have extensive, well-maintained knowledge bases.
The biggest challenge is changing people’s habits so that when they answer a question, they capture that knowledge and it can be reused next time. There also needs to be a clear reason for doing it. Ultimately, it can remove the routine, boring tasks that cause irritation and apathy. AI can then free people to focus their human capabilities where they add most value.
What should leaders assess across their people, processes and systems before deciding what to automate?
Making AI scalable across a business is the hard part, and it will be a learning exercise. Many employees already use AI in their daily work, for example to help draft emails or messages, but that does not necessarily translate into something that can be scaled across an organisation.
The priority is to identify use cases that genuinely merit the technology and can tolerate non-deterministic outputs. AI projects can struggle when a process requires completely predictable behaviour, or when the quality of the output is simply not good enough.
AI projects can struggle when a process requires completely predictable behaviour, or when the quality of the output is simply not good enough.
That means assessing where AI can genuinely add value, where human input still matters and how much variation or uncertainty a particular process can tolerate. In some cases, AI may be most useful for summarising information or helping shape thinking, with people adding the context, judgement and creativity.
Organisations will ultimately need to take a trial-and-error approach to identify the areas where AI automation delivers real value.
What approach are organisations most likely to adopt when they want to scale AI?
Some business leaders think AI is a magic wand, which overstates its current capabilities. But agentic AI is real and we’ll see much more of it as it becomes more familiar and trusted.
The approach will be cautious at first, like driving a new car with adaptive cruise control. Eventually you trust the system because it is better at spotting a traffic jam ahead and adapting than you are. Agentic AI implementations will be step-by-step in a similar way.
In service desk management, for example, we can use AI to fill in tickets. It’s a small step, but one that can be very helpful when it works properly. The output and underlying data still need to be checked for errors or hallucinations. From there, organisations can move on to anomaly detection and identifying cases that require human intervention.
How should organisations decide which tasks AI can handle and where human judgement still needs to remain?
Using AI on routine tasks such as taking meeting minutes and providing summaries has obvious value.
But moving beyond that, you need to test each use case and increase the level of autonomy as your confidence in the system grows. You can then ask AI to flag the anomalies or situations it does not have sufficient confidence to handle. That is where human judgement comes in.
There are many reasons human intervention might be necessary, whether because of priorities, empathy or risk, or because a situation involves friction or emotion. These are all valid reasons for AI to request human help and oversight. Even if these situations represent only a small proportion of cases, they may carry a disproportionately high level of risk.
Looking ahead five years, how do you see AI reshaping service management, and what should leaders be preparing for now?
Five years is a long time in the current environment, so any prediction needs to be treated with some caution. As an experienced coder, I think there will be no more hand-coding, apart from in some very specific areas. I see that as a positive development. AI can increasingly take care of repetitive technical work, allowing people to focus on the areas where their judgement and expertise add more value.
I think what increasingly differentiates businesses will be how effectively they integrate their “non-solidified knowledge”, even if it isn’t entirely clear yet how we capture and structure that knowledge in a way software engineers and AI can use. What is certain is that the combination of AI and human capabilities makes this an extraordinarily exciting time.
