With VAL 2.0 launching, we want to share how we got here, including the parts that did not go to plan. VAL 2.0 did not come from one big idea. It came from years of getting some things right, getting some things wrong, and learning from both.
We focused on the fundamentals
From the start, we put most of our effort into the basics. Collecting data from many different sources, and onboarding new sources quickly. A workspace where people can see their data and work with it. Dashboards built into the platform. Workflows to schedule and control data transformation and system activity.
At the same time, we learned the industry by working closely with our customers, dealing with their issues day in and day out. Every messy file and every number that did not tally taught us something. That experience is the most valuable thing we have built, even though you cannot see it in a demo.
What we chose not to do
We were also clear about what we would not focus on. We did not focus on user experience. We did not build automated data checks; our team caught data issues manually, together with our customers. And we did not spend time on visual features and extras.
This was a deliberate choice. First, it was about priority: what is fundamental and what is not. Second, it was about maintenance. Even if we built these features, how would we maintain them if they became complex? A feature we cannot maintain only solves the problem for now, and creates a bigger one later.
What we learned the hard way
We first tried a self-service approach, where customers would set things up on their own. It did not work. Our customers are SMEs, and most of them are not equipped to run a data platform themselves. They should not have to be. Our focus has always been to make things work for our customers.
So we took on the work ourselves, and for a while that is what made VAL work.
But taking on the work also meant we became the last mile for our customers' data. We were responsible for the outcome, even when the data changed within the customer's organisation or in the systems upstream. That made communication extremely important, because the numbers are only as good as the data that comes in.
As our customer base grew, so did the load on our team. We saw how much our team was carrying, and we knew that building more solutions we could not maintain would only solve the problem in the short term, not the long term.
Looking for a better way
In 2024, we started thinking seriously about how AI should be part of VAL, and about a new way of working that would remove the need for complex maintenance. We started to imagine what working with VAL could look like if it was completely different. For example, chatting with your data instead of opening a dashboard.
Around the same time, we built a domain model that could be replicated and promoted across customers. This centralisation sped up delivery a lot. But it also meant we had to centralise every customer's customisation, and that became a maintenance problem of its own.
There was also one challenge we kept facing. Different users have different needs, and use different terms for the same thing. There is a limit to how many dashboards and reports we can build for each of them.
Instead of adding more complex screens and more dashboards, we went down the AI chat path. That brought a new problem: accuracy and consistency. We introduced skills, which guide the AI on how to answer. They helped, but they still could not match the reliability of a programmatic approach. Meanwhile, our data volume was growing very quickly.
So we went back to the drawing board.
Redesigning VAL
To make this work for SMEs, three things had to be true. The output must stay consistent and repeatable. Users need flexibility in how they interact with their data and in what they get out of it. And they must be able to create what they need without the cognitive and technical overhead.
That meant redesigning the platform so it is configured not by the end user, not by us, but by AI, while the end user stays in control and manages it. Because AI does the configuring, users no longer carry the cognitive and technical overhead. And because it no longer depends on our team, the maintenance and support problem is solved in a sustainable way.
It was not only the platform that had to change. Our internal processes had to change with it. Scaling the platform also means scaling our internal capacity with the same number of people, so AI now augments how we work internally, not just the system our customers use.
We started the redesign together with AI document scan, a feature our users had asked for a lot. It also covered the last source of data we were not collecting at that time. From there, we took a hybrid approach in everything we do: AI where flexibility matters and a programmatic approach is not available, and programmatic logic where the answer has to be exact. That is where we are today with VAL 2.0.
Why VAL 2.0 is significant
VAL 2.0 is the point where we can create different screens, experiences and outputs without depending on our people. The user experience and visuals we chose not to build before can now be created by AI, for each user.
It also solves the data checks we used to do manually. Monitoring thousands of outlets is not humanly possible, and that is where AI comes in. It keeps watch all the time and flags data changes early, so we can talk to our customers about data readiness before the numbers are affected.
Our years of experience are now captured in skills that the AI uses. With these, the AI helps users configure and create what they need, on top of fundamentals shaped by the hundreds of scenarios we have handled since we started.
And it is not only what users see on screen. The engine that runs the platform, meaning how data comes in, how it is organised and what runs in the background, can now be set up by users themselves.
For us, this is what matters most. The years we spent on the fundamentals, including the choices that were hard to explain at the time, are what make this possible. AI does not replace that work. It lets that work scale.
What comes next
Our role is changing. We will focus on building and maintaining the data blocks and logic for the industry. We will create solution templates that give our customers a faster start. And we will teach them how to work with AI, so they can create what they need from the data blocks and logic we maintain.
Through all of this, our core focus stays the same: making sure we deliver more than what our customers pay for. Our benchmark is simple. How many people would a customer need to hire to achieve the same result, how long would it take them to get to the same stage, and how does that compare to what they pay us? If hiring would cost them more and take them longer, we are doing our job.
How we are rolling it out
VAL 2.0 is open for preview today, with general availability in October 2026. We will be working with our customers to bring VAL 2.0 into how they work, step by step.
You can see VAL 2.0 in action on our launch page.
VAL is no longer the same platform. The fundamentals are still at its core, and now they are supercharged because VAL is AI-native.
To our customers who worked through the tough problems with us, and to our team who carried so much along the way: thank you. VAL 2.0 is built on what you taught us.