> ## Content Index
> Fetch the complete content index at: https://www.frontierprofile.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# Chris van den Belt, Head of Product Management, Ultimo
- URL: https://www.frontierprofile.com/chris-van-den-belt-head-of-product-management-ultimo/
- Published: 2026-11-12T16:37:00.000Z
- Updated: 2026-11-12T16:37:37.000Z
- Description: Chris van den Belt on how Ultimo's embedded AI and digital workers are cutting repair times for maintenance teams, and why the technology is the easy part of enterprise AI.
- Author: Ewan MacLeod
- Tags: Asset Management, AI Agents, Industrial AI, Digital Workers

[Chris van den Belt](https://www.linkedin.com/in/chris-van-den-belt/?ref=frontierprofile.com) is Head of Product Management at [Ultimo](https://www.ultimo.com/?ref=frontierprofile.com), a cloud-native enterprise asset management (EAM) platform used by maintenance teams in manufacturing, logistics and utilities, and part of IFS since 2022\. He joined Ultimo in 2009 as a Product Owner and has led product management since 2020, with a focus on manufacturing and logistics. His team has built AI into the platform in two forms: embedded AI inside the screens technicians already use, and "digital workers" that act on their own across maintenance operations.

Chris is a returning guest: we first profiled him on Conversational AI News in [June 2025](https://www.conversationalainews.com/chris-van-den-belt-head-of-product-management-ultimo/?ref=frontierprofile.com).

My questions are in bold - over to you Chris:

---

**Who are you and what's your background?**

I'm Chris van den Belt, Head of Product Management at Ultimo, where we build software that helps organisations manage their maintenance and physical assets, from factory production lines to logistics hubs.

I studied Bachelor of Communication at HAN University of Applied Sciences, and then started as an Application Designer at Getronics PinkRoccade, where we built SAP-based software for local governments. That's where I learned translating customer needs into something technical teams can build.

That's essentially what brought me to Ultimo in 2009 as a Product Owner. I spent a lot of time on the shop floor with customers, then back with our developers, trying to close the gap between the two. Some of my favourite work from those years was building new modules from scratch around problems customers were genuinely struggling with, like our EHS suite. Safety processes, including Seveso-related ones, are normally rigid and run in disconnected systems, so we built them into the maintenance workflow, where the work happens.

In 2020 I was given the opportunity to lead product management, so these days I'm responsible for the overall roadmap, with a particular focus on manufacturing and logistics. Those are industries where an unplanned breakdown can halt an entire operation, so reliability really matters.

AI came into my work when generative AI arrived and opened up possibilities where we struggled before. We experimented with several ideas. One was turning the unstructured failure histories of our customers into structured data. That quickly exposed a bigger gap: much of the input data was missing, so we began using AI to improve its quality and completeness. Today, an AI-first approach shapes both our product and the way we build and maintain it.

What still motivates me is seeing the effect on the people who use our software. When a maintenance team moves from constantly firefighting to planning, and their workplace becomes safer, that's the part of the job I enjoy most.

I live in the Netherlands with my wife and two kids. Outside work, I like to step away from the laptop and get tools in my hands instead, tackling DIY projects in and around the house.

**What is your job title and what are your general responsibilities?**

In short, I'm responsible for deciding what we build and why, and making sure it solves the problems our customers face.

In practice, that splits into a few areas. A big part is owning the product roadmap: weighing customer requests, market developments and technical possibilities, and turning them into priorities the development teams can work from. That means a lot of conversations and quite a few difficult choices, because there's always more, we'd like to build than we have capacity for.

I also lead the product management team, 13 product managers and product owners who each look after different parts of the market and platform. I spend a good amount of time with them talking through ideas, challenging assumptions and helping them make decisions, rather than making every call myself.

Then there's the customer side, which I still try to keep close to. I focus especially on manufacturing and logistics, so I regularly talk to and visit customers in those sectors and talk with partners and our own sales and consultancy colleagues about what they're hearing in the field. Those conversations are often where the best product ideas come from.

A typical week might include a roadmap review with development leads, a customer visit at a production site, a session with the team on an upcoming AI feature, and time with sales and marketing preparing a product launch. No two weeks are quite the same, which is one of the things I like about the role.

**Can you give us an overview of your business?**

Ultimo makes Enterprise Asset Management (EAM) software - the system maintenance and asset teams use to keep track of their equipment, plan and carry out work, manage spare parts, and keep people safe while they do it. Our customers are mostly organisations in sectors like manufacturing, logistics and utilities, the kind of companies where a machine breakdown or a safety incident has immediate consequences.

What sets us apart is how quickly customers can get up and running, and how easily the software adapts to the way they work. Many EAM systems are built for very large enterprises and take a long time to implement. Ultimo is cloud-native, highly configurable, and designed to be usable by the technician on the floor, not just the planner in the office.

Our business model is software-as-a-service: customers pay a subscription for the platform, and we work alongside a network of implementation and consultancy partners.

Our AI takes two forms. Embedded AI is built into the existing Ultimo platform, suggesting, drafting and completing tasks in the screens our users already work in, so people get more done with less friction. Digital workers are built outside that platform as an AI-native approach: they take initiative rather than waiting to be asked, work across systems and coordinate tasks on a customer's behalf. Ultimo has been around for more than 35 years, and both forms are already in use today.

As for scale and backing, Ultimo was founded in the Netherlands in 1988, and since 2022 we've been part of IFS, the global enterprise software company. Within IFS, Ultimo continues as its own product, aimed at organisations that want a fast-to-deploy, standalone EAM solution, while benefiting from IFS's international reach and investment. Today we support over 150,000 technicians managing more than 22 million assets for 2,500+ customers worldwide.

**What problem do you solve with AI?**

Our customers are maintenance teams in factories, warehouses and distribution centres, and their core problem is not a lack of knowledge. The knowledge exists, but it sits in the system where nobody can reach it now it matters.

Picture a technician facing a broken machine at two in the morning. Years of work orders describe similar failures, their causes and their fixes, yet that history is spread across free-text notes written by different people in different styles. The technician falls back on memory, calls a colleague or starts from zero. Meanwhile, experienced people are retiring faster than new ones can be trained, and much of their expertise leaves with them.

Repair time is mostly lost before any repair begins. Industry estimates put diagnosis at around 80 percent of total repair time, and poor failure reports are a major reason. An operator notices that a machine sounds or smells different, but by the time the report reaches a technician it reads "machine 4 not working." Skilled technicians are increasingly hard to find, and downtime in manufacturing can cost tens of thousands per hour or more, so that lost time adds up quickly.

We began in 2024 with embedded AI, built directly into the Ultimo screens our users already work in. We started with failure reporting: when someone logs a problem, Ultimo recognises the asset and suggests likely symptoms, so the reporter can capture what they saw, heard or smelled with a few taps instead of typing. The technician receives a far better picture from the start. From there we extended the capability along the repair process. Photos are described automatically so they become searchable, assisted troubleshooting finds similar past failures and groups them by root cause, and planners can turn a few short notes into step-by-step work instructions. In January 2025 we made embedded AI available to all new customers.

The results have exceeded our expectations. Across our cloud customers, median repair times have dropped 42 percent from their 2022 peak, measured across nearly seven million tracked failures, and the decline has quickened since we rolled out AI. Three in four customers who were with us through 2024 and 2025 saw their repair times fall. At Berkvens Doorsystems, team leads save 30 to 60 minutes every day because information is summarised automatically, and they find the AI's insights match their own analysis more than 95 percent of the time.

Embedded AI helps people work better inside the platform. Our next step was to bring AI into the team as a colleague that acts on its own. In 2025 we introduced our first digital worker, an EHS specialist that tackles underreporting. Instead of waiting for someone to file a report, it continuously scans work requests and shift logs and reports incidents, near misses and unsafe situations autonomously.

The reason for digital workers is the workforce gap facing our market. An estimated 40 percent of the industrial maintenance workforce will retire by 2030, and the supply of talent cannot keep up. Digital workers do not replace experienced people; they spread the expertise of those experts across the whole team. They come in two forms: conversational workers that users can talk to around the clock, with full knowledge of the asset, its maintenance history and its operational context, and autonomous workers that run 24/7 and act on triggers rather than manual input. We modelled them on existing roles in a maintenance organisation rather than on feature names.

In 2026 we launched two more digital workers, bringing the total to three. The maintenance technician guides onsite troubleshooting and records the work performed. The maintenance planner prepares work orders before the planner reaches the desk and briefs the team on priorities and emerging issues, drawing on maintenance data and production shift logs.

Together, embedded AI and digital workers mean that a less experienced technician can draw on years of collective knowledge the moment they open a work order. The cleaner data this produces also lays the foundation for the predictive work that comes next.

**What does the AI actually do, and how is it built?**

Currently, we don't build or train our own models. The AI in Ultimo runs on a public large language model from OpenAI, combined with vector search built on embeddings. For assisted troubleshooting, we turn each customer's failure history into vectors stored in that customer's own Ultimo database. When a technician or planner starts diagnosing a problem, the system retrieves similar past failures by meaning rather than keywords, and the LLM condenses the closest matches into a concise summary grouped by root cause. We are extending the same approach to equipment and OEM manuals, which we vectorize so users can find the relevant repair or diagnostic passage quickly. The prompts carry operational context, such as the asset and the failure report, and we keep personal data out of them. Prompt data is configured so OpenAI cannot use it to train its models, and no customer's data informs another customer's results.

**Who are your target customers, and what's your revenue model?**

Our customers are mostly mid-sized and larger organisations that depend on physical assets - manufacturers, logistics companies, ports and terminals, utilities and healthcare providers. More than 2,200 organisations use Ultimo, including Castrol, ConMet, Stolthaven, Bristol Port Company, Van Moer Logistics and Berkvens Doorsystems.

The customers who get the most out of Ultimo are those ready to move from firefighting to structured, planned maintenance. Castrol's plant in Ghent is a good example: they moved from spreadsheets to Ultimo, halved their equipment failures and now save €360,000 a year.

Our revenue model is software-as-a-service. Customers pay a subscription and we work with our own services team and implementation partners to help them get up and running. Our embedded AI features are included in the subscription, and the digital workers are charged as an add-on.

**What's the evidence it works, and what have you learned along the way?**

The clearest evidence is in repair times. We track mean time to repair across our cloud customers, and between 2020 and 2026, across nearly seven million recorded failures, the median dropped by 42 percent. Three in four customers who were with us through 2024 and 2025 saw their repair times go down.

I want to be honest about that number. Part of it is recovery from the COVID years, when workforce shortages and supply chain delays pushed repair times up across the industry, and part of it comes from product improvements that have nothing to do with AI. What convinces me AI is making a real difference is that the improvement sped up after we made embedded AI available to all new customers in January 2025, and it has kept going since. In 2026, customers using AI repair failures about 25 percent faster than customers without AI.

At the individual customer level, Berkvens Doorsystems has been one of the early users of our digital workers. Their team leads used to spend the start of each day pulling together information from different places. Now that's summarised automatically, saving each of them 30 to 60 minutes a day, and they find the AI's conclusions match their own analysis more than 95 percent of the time.

As for lessons, our biggest shift was moving from applying AI to our customers' data towards using AI to capture better data. For most of our customers, the biggest hurdle is getting the knowledge of operators and technicians out of their heads and into the system, so we are making that easier. Once that is in place, many other AI possibilities open up, because AI is only as good as the data it can work with. That taught us that better data capture must come first in any AI project.

The biggest lesson, though, came from using AI ourselves. We believed we couldn't credibly sell AI to customers without proving it inside our own company first, so over five months we deployed around 20 digital workers across departments like HR, finance and sales. What surprised us was where the effort went. Teaching people to use a chatbot takes about an hour. Getting an organisation to rethink how work gets done takes months of slow, unglamorous process work: we spent four months mapping workflows with department heads and found about 70 places where AI could help before we built anything. That changed how we talk to customers. Now, when they ask where to start with AI, the first conversation is about their processes and their data, not the technology.

**How do you handle the data question?**

It's the right first question, and we try to answer it plainly.

First, what goes where. Our AI features work with operational data: information about the asset and the failure report, or for troubleshooting, the customer's own history of past failures. We deliberately keep personal data out of what we send to the AI model, and where it could slip in through free-text fields, we can mask it before processing. The language model we currently use is from OpenAI and it's configured so that nothing we send can be used to train their models.

Second, what we train on. The short answer is that we don't train models on customer data. The AI uses each customer's own data, in the moment, to give that customer a better suggestion. For troubleshooting, we turn a customer's failure history into a searchable index, but that index lives in the customer's own Ultimo database. One customer's data never influences what another customer sees, and customers always keep ownership of their data.

Third, control. Every AI feature can be switched off by the customer's own application manager, and Ultimo works fully without them. That matters for organisations that want to introduce AI gradually, or not at all in certain areas. We also log when an AI feature has been used, so there's a trail if anyone needs to check how a decision was made. And we're clear with customers that the AI makes suggestions; a person stays responsible for the decision.

For regulated environments, the AI runs on the same secured platform as the rest of Ultimo: data is encrypted in transit and at rest, access is tightly controlled, and we run separate cloud instances per region so customer data stays where it's supposed to be. An internal AI Ethics Board reviews any change to how AI uses data and keeps track of regulations like GDPR and the EU AI Act.

**What's new at the company, and what's coming next?**

Looking ahead, we are focusing on two big areas. The first is expanding our digital workers. We will add depth to the three roles we have already launched, and over time more roles will join the team, such as a reliability engineer or a warehouse supervisor. That way, digital workers cover more of the maintenance operation instead of just three slices of it. We are building a team of digital workers, and the first three are now in the field.

The second is enabling customers to use Ultimo as a headless application. Some users will always prefer our user interface, but others want to reach asset management data and functionality through other channels. Rather than asking everyone to log in and move through screens, reports and dashboards, we bring data and functionality to people, in the flow of their work. Desk-based colleagues, such as asset managers, maintenance managers and reliability engineers, can run advanced analysis on stock levels or asset health across multiple sites, either on our Intelligence platform or by connecting Claude or Copilot to Ultimo. A technician in the field will use a native AI app with a fluid interface that combines natural language and voice, with the on-screen elements adapting to the task and the situation.

**If you had a magic wand, what one thing would you change about how AI gets bought and deployed?**

AI is often treated as a feature or a technology purchase, but the real challenge lies in the processes around it. Buying the tool is the easy part. The harder work is redesigning how teams plan, maintain and make decisions, and then managing the change that comes with it.

If I could use that magic wand twice, I would also cut the total footprint of AI. That means starting with low-value uses that consume a lot of energy for little benefit and making each remaining task far less energy-hungry. I hope innovation delivers that, because efficiency gains alone will not offset the growth in demand.

**What's your message for the large enterprises evaluating AI right now?**

My first message is that the technology is the easy part. Most large organisations can buy and switch on AI tools quickly. What takes time is changing how work gets done. Organisations must absorb not only new technology but a new way of thinking and working. We learned that ourselves when we rolled out AI across our own company, the real effort went into mapping processes and deciding where AI should help, long before we built anything.

Second, start with a real, everyday problem rather than the most impressive one. We could have led with predictive maintenance, because it makes a great demo. Instead, we started with something unglamorous, the quality of failure reports, because that's where our customers were losing time today. Small, practical wins build trust, and they also produce the clean data you'll need for the more ambitious projects later.

Third, don't overlook the physical assets that make production possible. If you run manufacturing sites, don't focus AI only on production processes. The machines, utilities and facilities that enable production also need to be managed and maintained, and AI can help there. When an asset fails, it costs money and can affect safety. AI is already delivering measurable results in asset management, and the risk stays relatively contained because a person remains in the loop.

Finally, treat this as a permanent change in how you run the business, not a project with an end date. Your employees and customers will judge your systems against the best technology they use anywhere, not just against your direct competitors. The organisations that do well will be the ones that keep people in control, give them new roles alongside AI rather than simply cutting headcount, and keep learning as the technology moves.

**What AI tools do you personally use day to day?**

I’m a big fan of Claude, especially combined with my ‘second brain’ in Obsidian. I organize the vault with the ACE framework (Atlas for what I know, Calendar for what happened, Efforts for what I’m driving to an outcome, from [https://www.linkingyourthinking.com](https://www.linkingyourthinking.com/?ref=frontierprofile.com)). Because it’s all plain markdown with a clear structure, Claude knows where everything belongs and how it connects. It turns raw meeting notes into filed, linked notes, briefs me before meetings on what was agreed and what is still open, and picks up each piece of work where I left off. The value is that my knowledge compounds with every conversation instead of disappearing into chat history, and I spend my time on decisions, not admin. Obsidian ([https://obsidian.md](https://obsidian.md/?ref=frontierprofile.com)) is where it all lives.

**What's the best new AI product or service you've seen recently?**

I was blown away by Dessa ([https://block.xyz/dessa](https://block.xyz/dessa?ref=frontierprofile.com)), a new voice model trained on the sound of human speech, instead of text. It is both impressive and scary, because you really can’t tell it’s AI. I can see how this is the next step to make human-agent collaboration more natural.

**Finally, let's talk predictions. What trends do you think are going to define AI over the next year or so?**

First, the interface changes. Today most software shows you its data model: screens, tabs, fields. That will fade. People will ask for what they need in their own words, through whatever tool they already use, and the software works in the background. A technician says what he found, and the work order, parts and history are handled for him.

Second, more work goes to digital workers. They plan, check, chase and report, next to the humans.

Third, robotic workers join the team. AI has sped up how fast robots learn, and I expect that pace to continue. Complex repairs will take a while. But routine work is close: transport, lifting, inspections. The sweet spots are dangerous places, repetitive tasks, and monitoring where a robot beats a human, like thermal, vibration, acoustic and visual checks.

Technology will not be the limit. Digital maturity will. Many maintenance organizations still run on paper work orders, while the front runners are already building on AI. I expect that gap to grow.

But the first step is small. Use AI to capture what is now only in the heads of your people: why a machine acts up, what the sound means before it fails, what the manual never says. Richer, more complete data is what digital and robotic workers need to do anything useful. It unlocks everything else.

---

Many thanks to [Chris](https://www.linkedin.com/in/chris-van-den-belt/?ref=frontierprofile.com) for taking the time to share his insights with Frontier Profile. You can learn more about [Ultimo](https://www.ultimo.com/?ref=frontierprofile.com) on their website.