Dave Ruane is Global Director Client Solutions at Lion People Global, and runs the Process Innovation Challenge (PIC), the pitch competition that has become one of the industry's clearest windows into what companies are actually building.
We asked him where language technology stands, what the AI opportunity looks like beyond speed and cost, and what he's watching for over the next two years.
How long do you have? I think you have to look at how the space evolved to understand where it is now.
That's where the platform part of language technology comes from, what we now call the translation management system (TMS), with the business management side and various connected pieces sitting around it.
So why am I giving you a history lesson you didn't ask for? Because with any new blue ocean space, there's room to fill in terms of matching your position to what customers need. Companies like XTM did very well there. There was natural growth. And to some degree that's become less of a growth curve, maybe saturated in certain segments. There are still sectors where companies like XTM are bringing on net new customers, so it's not a decline story.
But the question now is how you move those systems into the next sphere. End customers are being pushed to deal with everything repetitive through some kind of AI system, and the platforms they subscribe to need to fulfill that.
So how do you move a legacy piece of software, which it just is, into the AI space?
How do you make it AI?
That's the challenge for anyone in TMS land, and anyone in language tech. You either AI your system or you build a new one on top of an AI stack.
I'll give you an example. There's a SaaS business you may have heard of. Up until about three years ago they were growing laterally, around 11% year on year, which isn't good for a pure SaaS play. And that was a decreasing curve over the previous five or six years. They were still growing, but the rate had dropped.
They decided they had to turn it around and get back toward the 30% a year they'd seen earlier in their growth trajectory. So they brought the founder back in as CEO, and he was allowed to look at things and come up with a plan.
That meant a lot of the legacy stuff they did, they had to cannibalize it, drop it, and make genuinely hard decisions about how they ran the business. In a lot of respects they've created a template for other SaaS companies. If you take the technology adoption life cycle bell curve, they're on the left hand side, in the innovators group. Anyone who hasn't followed their lead is somewhere to the right of them (on that bell curve).
I still think there's time for companies in particular in the language and multilingual content space to assess how far into becoming an AI company they need to go in their own niche. The question for CEOs, is how relevant it is to their market. In TMS land, I'm sure every CEO is asking themselves the same thing.
That's very much where the game is at for anyone in the localization business, which is becoming more like the multilingual AI business.
Some people call it governance. Some call it AI quality control. A term will stick soon, because we're still in that incubation phase where everyone is trying to work out what to call this thing. But it comes down to how you put layers of checking in place on what the AI is doing, and then how a service side ecosystem manages that, and how the technology side helps support it. Enterprises clearly need a way to outsource the management of it.
It's less about what we're translating, because the translation piece may well be handled by AI the enterprise has brought in-house. It's now about how quality is gated, managed and controlled.
I use the word quality carefully, because it means something quite specific in the language space. So we need to determine what leveling up that quality layer really means, and then what managing it means.
A lot of this is becoming more closely linked to Key Business Objectives that an enterprise would have. Where someone was previously doing linguistic quality control, now the questions are what is the impact of a quality issues, how we stop it happening again, and what is it worth to do that.
It's bringing the humans up a layer and enabling them to move them into specific quality control jobs that look different to what has come before. Sometimes you have a bit of time to test this, figure it out, and build the processes around it. I think that's what the industry as a whole is going through right now.
We get around 30-40 submissions each cycle. We run a preliminary round with the vast majority of them, they pitch the idea, and the dragons and organizers decide who reaches the final. So it's a big cut, 30 down to six this time. Then we spend a couple of weeks coaching the finalists. Pitches at the final are three and a half minutes in length, which I set because I get bored listening to anyone for longer than that, myself included.
I'd say a lot of what has come to these stages from established companies has been iterative rather than invention level. That's shifting. People are going to want to see more genuinely innovative ideas, and I think we're starting to see that requirement now. In the latest version, which will take place at LocWorld56 in Vancouver in October 2026, there is a good mix of ideas pushing into genuinely new territory, from rethinking human roles and commercial models in AI-assisted localization to agent-driven services, AI quality evaluation, voice workflows and new approaches to AI transparency.
Sara Basile's winning pitch for XTM was a good exponent of solving a real problem, and it got people's attention. It dealt with intelligent triage, delivered through XTM Go, and the strength of it was that it took something enterprises are quietly quite worried about and gave them a way to get in front of it. It dealt with shadow AI.
The interesting part is that it's a problem AI itself created. The tools got good enough and accessible enough that people started using them on their own, ahead of the controls. So the answer isn't to ban it, it's to give enterprises a way to see it, route it, and bring it back under management. That's a live issue for a lot of companies right now, and it's exactly the kind of thing stages like the PIC should be surfacing.
There's a mix of things going on. You have incumbents having to pivot, making wholesale efforts to bring in AI. Then you have fresh companies with no legacy incumbency, whose hands aren't tied by previous stacks or approaches. They live in an AI world already, and they're working out how to connect into this industry.
In my day-to-day work with M&A, I see a lot of this thinking and activity. I expect we'll see more of it through next year, from both technology and service companies that have the revenue streams, cash and capability to do it.
Sometimes you try to plug in the next AI technology and then realize you're actually plugging into them. That's fine, as long as you end up with the best-of-breed, relevant technology stack supporting real customer needs.
The signals aren't fully set yet. I'm seeing this under the hood and it's starting to appear at the PIC. We run two cycles a year, and I expect we'll see more client-side AI innovation and more pivoted language technology companies emerging over the next few events.
And that brings me back to the point I made earlier.
If the part of the value chain you traditionally got paid for can increasingly be done by AI, simply doing that same thing more efficiently isn't enough. You have to work out where the value moves next and build the company around that.
The companies we'll increasingly see on the PIC stage will be evidence of that shift. And over the next couple of years, I think the dividing line in localization won't be between technology companies and service companies. It will be between companies that have genuinely become AI companies and those that have simply added AI to what they were already doing.