Is SaaS Evolving? We Ask WATS CEO Tom Andres Lomsdalen
AI in enterprise software has become one of the most hyped and least clearly explained developments in recent memory. More and more predictions are surfacing every day about its trajectory, often from opposing points of view, and it leaves manufacturers in the middle trying to figure out what any of it actually means for their operations. We sat down with Tom Andres Lomsdalen, CEO of WATS, to hear what he thinks about it.
There’s a lot of talk right now about SaaS being dead. What’s your take on that?
It’s an interesting claim but I think it overstates things considerably. What people are really pointing at is that some software products, the ones that are essentially just a basic interface sitting on top of a database, may struggle as AI makes it easier to replicate that kind of thin functionality. That part is probably true.
But enterprise software that has genuine depth, real integrations, domain expertise built into how it works, and compliance and security considerations, doesn’t get replaced just because someone can generate code faster.
There’s also something that gets overlooked in the conversation, which is the connection to the real world.
A CRM relies on people entering data whereas WATS, for instance, connects directly to test machines. Those machines are pushing data in huge volumes, automatically. That means whatever platform you look for when managing test data it has to be able to handle the load.
We do see the temptation to build rather than buy, now more than ever. It’s always been something we’ve seen with someone in the test or IT team convinced they can handle test in Excel for instance.
But now, with AI making building software that much easier, it’s something that comes up very often when speaking to manufacturers. But, the pattern is still the same, with teams building for analytics that solve a problem today, but that can’t be maintained or adjusted as the business’ needs change.
We actually embrace customers building on top of WATS, so we always make their data available for them to use how they like. And, with the introduction of connections via MCP (Model Context Protocol) that’s getting more powerful than ever.
The short answer to the question is that SaaS isn’t dead, it just has to evolve to stay relevant.
Where do you see the line between SaaS and what people are calling AIaaS?
The simplest way I think about it is that SaaS gives you tools to do a job, and AIaaS (AI as a Service) has the potential to do the job on your behalf. In traditional SaaS, a person navigates the software, interprets the data, and decides what to do. With AI agents, that loop can become much more automated. The agent can look at the data, surface what matters, and in some cases may be able to act on it directly.
What's easy to underestimate is how complex the improvement loop is in manufacturing, even when you have good data. If AI surfaces a yield issue, that doesn’t mean you can just fix it there and then. The fix usually involves a significant change involving R&D, test engineers, EMS partners, production lines, and individual machines. So the question isn't only whether AI can find the problem faster, but whether it can support the methodology for actually improving things. That's where deep domain knowledge becomes even more critical.
We aren’t fully there yet, and I think everyone who claims to know exactly how quickly this will develop is probably overstating things. The technology is still developing, manufacturers rightly want to see it proven in the real world first, and the pace of change is faster than most people expected.
How does domain knowledge fit into that picture?
This is something I think about a lot. An AI agent is only as useful as the context it is working with. If you point a general purpose AI at manufacturing test data without any understanding of what that data means, what good yield looks like, what a suspicious pattern actually indicates, you can get answers that are generated in a vacuum.
A simple example of this is if a unit starts failing because test measurements are outside the limits you’ve set, AI might conclude that there’s a process issue or component defect. But – since the quality of the data you input (UUT reports) might be limited – it might actually be an instrument problem or a probe that isn’t connecting properly to the device. Without the domain knowledge to distinguish between those possibilities, an AI might point you in the wrong direction, wasting your time and not helping you solve the problem.
Domain knowledge is what makes the difference between an AI that produces a plausible output and one that produces genuinely useful insight. That’s always been central to how we’ve built WATS. The way I see us evolving is that our AI agents will be built and trained on both our own domain expertise, and crucially, that of our customers as well.
What does the shift from human-to-machine to machine-to-machine mean in practice for manufacturers?
Right now, most of the value in a platform like WATS comes from a person logging in, looking at their data, and making decisions based on what they find. That workflow is already much faster and more reliable than what manufacturers were doing before, but it still requires someone to initiate it.
The next step is platforms that are proactive rather than reactive, where instead of a quality engineer going to look for a problem, the system surfaces an issue and tells them where to look. We’re building toward that, and I think the manufacturers who get comfortable with that shift early will have a meaningful advantage.
So why don’t you just connect and collect ALL the data needed for perfect analytics?
Complete analytics requires complete data, and in most manufacturing environments that’s still a significant challenge. Test reports from production are often missing important data and when you have potentially hundreds of test machines spread across multiple sites and contract manufacturers, grabbing every piece of data out there is virtually impossible. At the very least it isn’t cost effective at the moment.
We think that technologies like MCP will genuinely help here, because they allow much easier integration between different systems than traditional APIs. That’s part of the reason we’re developing partnerships with companies like Yieldwerx and Aegis, so that we can extend what’s possible when data from different parts of the manufacturing chain can be connected more easily.
Is there a risk that the industry moves too fast on this?
Genuinely, yes. There’s a version of this where companies add AI to everything without being clear about what problem they’re solving, and end up with something that feels impressive in a demo but doesn’t actually help anyone on the production floor. The manufacturers we work with don’t have time for that. They need answers they can trust, quickly enough to act on. That’s a higher bar than generating a plausible looking output, and it’s the bar we hold ourselves to.
Where does that leave manufacturers who are trying to figure out where to invest their attention right now?
I’d say focus on the data foundation first. AI agents are only as good as the data they work with. If your test data is siloed, inconsistent, or used to calculate things incorrectly, layering AI on top of that doesn’t fix the problem. One thing worth noting about how WATS pricing is structured is that you pay on data volume, not users or stations, which means it’ll always scale with you rather than against you.
Ultimately, get the data right, get the methodology right, and then the AI has something real to work with. That's what actually works.
About WATS
WATS is a test data management platform built for manufacturers that need complete, accurate visibility into what is happening in their production and test operations. The domain expertise Tom describes in this conversation is built into how the platform works, from the way it calculates True First Pass Yield to the AI-assisted root cause analysis that helps test teams find answers faster.
If the ideas in this conversation are relevant to how your operation works, you're already in the right place.
