AI can't give you value, if you've lost control.

AI can’t give you value, if you’ve lost control.

December 2, 2025

Why can't every large translation buyer access the same AI?

Too many companies have lost control of their translation. Control isn't just downloading TMs – AI that actually creates concrete value in the real world requires more than that.

Every week, folks knock on our door here at ModelFront, wanting to get started with AI to check and fix AI translations, but we have to give them the bad news: they have nothing.

The good news is if you take back control now, you can be ready to go in as soon as six months, and it does not require building in-house.

— Adam Bittlingmayer, co-founder, ModelFront, December 2025

Outsourcing translation is understandably tempting – it rarely makes sense to build everything in-house, translation is not core to most companies and volume fluctuates.

But if you want to roll out AI that creates concrete value, and you want that value to go to your company, you need to control key bits.

  • Is your translation management system controlled by you? Or by a manual human translation service?

  • Can you get your full workflow data for every project and file out at scale (e.g. via API)? Or only TMs and only by manual downloads?

  • Can you change or add AI translation or other AI unilaterally? Or do you have to ask permission from the translation management system or manual human translation service?

You'd be surprised how many companies fail this basic test, even in shiny sectors like tech or gaming.

(So if your company does have full control, your translation team deserves praise for their street smarts and integrity.)

How loss of control is blocking AI

Translation AI that actually creates concrete value in the real world requires proper assessment, data, customization, integration and monitoring.

If you don’t have control, you can’t even do any of this. For example, just for an assessment, many translation teams find that it’s frustratingly hard to get out the data that they have been paying for.

  • In some cases, data is stuck inside the translation management system, or auto-deleted. (Even if the translation management system runs on the buyer’s own infra.)

  • In other cases, data is stuck inside the agency. (This is especially complicated if the agency has been pretending that it did not use machine translation, or just outsources the work to other agencies.)

  • And in the worst cases, buyers have even lost control of their own team – the buyer-side translation leadership is practically working for the mega-agency, against the interests of the buyer.

Mega-agencies selling human translation inherently lack the DNA or motivation to support AI that actually creates serious efficiency for buyers, let alone research and launch it. It’d be like taxi companies inventing self-driving cars.

Why “the industry” is selling lock-in

You never hear from “the industry” that buyers should have get control or transparency.

Undifferentiated mega-agencies know that buyers can get the same services from dozens of others. Mega-agencies mostly just outsource to other agencies anyway. It may even go to the same human translator at the end of the supply chain. The middlemen are commoditized.

So for the mega-agencies, the business model is lock-in, behind an opaque wall. Their refrain is basically “Trust in me, just in me”, like the snake in The Jungle Book.

Snake from The Jungle Book

Trust in me, just in me
Shut your eyes and trust in me
You can sleep safe and sound
Knowing I am around
Slip into silent slumber
Sail on a silver mist
Slowly and surely your senses
Will cease to resist
Trust in me, just in me
Shut your eyes and trust in me

At this point, your survival instinct should kick in. Wake up, Mowgli! The fact that they want to control your whole translation operation so badly, is precisely why you should keep control with you!

Of course, in this jungle, mega-agencies evolved to market lock-in and opacity as a convenient, efficient, one-stop shop that magically checks all the right boxes – AI, quality estimation, automatic post-editing, agentic… (Even if they haven't yet started trying to build those, or are already failing for years.)

They are even pitching that they can replace the buyer’s core translation team. Just give it all to them. Hypnotizing.

And those mega-agencies sponsor the main industry events, the steak dinners, the Forbes articles… They make a lot of noise.

The most efficient large buyers avoid this trap. But they’re busy, not selling anything, so you won’t hear much from them, unless you ask.

How buyers keep control

Keeping control does not require building everything in-house. Many of the most efficient large buyers buy every single piece, and pull it all together.

The key is that large buyers should never, ever buy technology and manual human translation services from the same vendor.

It’s too deep a conflict of interest – technology can automate part of manual human translation, which is the business that makes all the money.

Efficient large buyers get work done transparently inside translation management systems they control, whether that work is done by in-house translators or freelancers or through agencies.

Translation management systemTechnologyBuild or buy
AI translationTechnologyBuild or buy
AI to check and fix AI translationTechnologyBuild or buy
Manual human translationServicesIn-house, freelancers or agencies

In practice, the most efficient large buyers go beyond buying technology and manual human translation services from separate providers. They buy each piece from the best provider for that piece.

So they use translation management systems that make it easy to plug in third-party providers for all the other pieces, just like iOS and Android lets you use apps that aren’t made by Apple and Google.

And needless to say, large buyers should never let a manual human translation service vendor replace the people in their core translation team. You need people on your side to evaluate and police the technology vendors and manual human service vendors.

To be clear, this recommended setup is for large end buyers – those spending more than $1M a year on translation, in a relatively centralized way. For small or medium buyers with more money than time, like decacorns, or one-off projects, or law firms, a full-service one-stop shop can make more sense.

Escaping lock-in

So if you want to start using translation AI that actually creates concrete value for your company, you first need to take back control of translation.

Many companies have avoided or escaped lock-in. So there are plenty of other translation buyers, private communities, open communities and paid consultancies that are happy to help you.