Ben Faes is CEO of RWS, a global AI solutions company.
Enterprise AI has delivered on its promise of speed. Content that once took hours now takes minutes, and teams that once struggled to keep pace with channel and market demands are no longer constrained by creation capacity.
Somewhere between the input and the outcome, however, something’s going wrong.
Our survey of 200 enterprise leaders revealed a contradiction that should give every C-suite pause: The same AI investment that’s accelerating content creation is slowing the localization process because so much of what AI produces needs to be fixed before it can actually be used globally.
Solving The Wrong Part Of The Problem
Most AI investment targets creation. Remove that bottleneck and, the logic goes, scale follows. In practice, the opposite is emerging. As the front end accelerates, the downstream workflows responsible for adapting and validating that content are buckling under the pressure.
AI is being applied where the problem used to be, not where it now sits. Without the infrastructure to match it, faster creation just means more content queued up and waiting.
As a result, one in every five dollars (21%) of enterprise localization budgets is now spent on rework, inconsistency and content that was never designed for global markets, including correcting AI-generated content before it can be deployed globally.
That cost compounds with every market you enter. A product description or support article must function across languages, cultural contexts, regulatory requirements and formats. AI can produce language that reads well in isolation. Making sure meaning, tone and brand voice hold up when that content crosses into a different market is harder.
AI: Not Organizing The Attic But Filling It Faster
Most enterprises are running on years, sometimes decades, of accumulated material (marketing campaigns, product documentation, support content, internal knowledge) that’s been created, stored and adapted over time, often without a unified structure.
The result can resemble something less like a system and more like an attic. Content piles up faster than it can be classified or reused, and AI only accelerates that. As creation gets easier and volume grows, fragmentation deepens rather than resolves. The question stops being how much content you have and becomes how much you can actually use.
The numbers bear this out. Only 14% of enterprise leaders have centralized content management in place. The rest are scaling AI on top of disconnected systems, with governance (versioning, approval, brand voice control) struggling to keep pace and wondering why the output doesn’t hold together.
A confidence gap compounds this. Many leaders believe their organizations will cope without fundamental change, even as teams report being stretched. When the reality is that AI enhances what’s already in place, it amplifies fragmentation.
None of this is a failure of the technology itself. They’re failures to plan for what happens after AI produces the content (who checks it, who owns it, who accounts for the market it’s about to land in). Most enterprises already recognize at least one of these problems. The real question is why so few have done anything about it.
What Holds Buyers Back
I hear the hesitation constantly, and it rarely comes from people who doubt the problem. The enterprises I talk to have seen their own rework numbers. What stops them is calculation, and not always the right one.
Cost comes up first almost every time, and it’s usually miscalculated rather than genuinely prohibitive. Cultural intelligence gets priced as a new line item on top of an AI budget that already feels stretched instead of being weighed against the rework spend it’s meant to prevent. Put the two numbers side by side, and the investment looks less like an extra cost and more like a correction to one already being paid.
Security is the concern I push back on least because it’s the one that’s actually well-founded. Handing sensitive material (customer data, regulated documentation, product details a competitor would want) to any outside layer, human or machine, deserves scrutiny. The organizations that get this right can say precisely who touches what, where it lives and who answers for it. If a vendor can’t answer that clearly, the caution is doing its job.
What surprises people most is how small the implementation actually is. They come in expecting a rebuild (new vendors, new systems, a road map on hold while it all beds in). What works in practice is smaller: cultural and contextual intelligence built into review points that already exist, not stood up as a new department.
The barrier I can’t solve with a contract is who owns the decision. Content, localization, IT and legal usually report to different people with different budgets, and this intelligence layer runs across all four. When a decision belongs to everyone, it tends to belong to no one. That’s more often what stalls things than any of the concerns above.
None of that is a reason to wait. It’s a reason to be specific about what’s actually in the way.
The Cost Of Doing Nothing
AI has transformed the economics of content creation. What was once scarce is now abundant. However, abundance doesn’t guarantee content that works. Cultural nuance, situational awareness and audience expectations all shape how a message lands, and none of that transfers automatically across markets.
The enterprises that need it most are the ones already paying that 21% rework tax and quietly accepting it as the cost of doing business. For them, the choice is simple: Fix it once before it ships, or keep paying the tax market by market.
Call it what it actually is: a business cost. Most enterprises are already paying it whether they’ve noticed or not.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
