Escaping False Polarizations in the AI Narrative

Creative Destruction

Joseph Alois Schumpeter, the 20th-century Austrian economist, coined the notion of “creative destruction” to explain technological and economic development. This process, which has become increasingly evident since the first industrial revolution, is a dynamic and relentless feature of capitalist economies in which new innovations destroy old economic structures and replace them with more efficient ones. In the short term, this destruction causes economic pain, but in the medium and long term, it creates systems that are able to offer higher wages. That’s the entire foundation of economic growth in all industrialized countries. The interim period that includes the destruction phase is painful but nonetheless necessary so that less-efficient companies can be replaced by more-efficient ones. The question that arises is, “What degree of destruction (in qualitative and quantitative terms) is AI already imposing on our industry?” Is this process entirely similar to Schumpeterian creative destruction, or is there something unprecedented happening related to the particular nature of AI? Will the destructive phase be followed by an actual creative phase capable of improving people’s wage conditions? In other words, is AI an innovative technology like any other, or is it something radically different never seen in recent history? Are we going through a normal process, or are we in the midst of a paradigm shift imposed by the exceptionalism of this new technology? The trouble is that, being in the middle of this transformation, we can’t easily understand much of what’s really happening. However, we can analyze trends in the language services sector and try to draw inferences about AI’s role. We can ask, “Is AI already having measurable impacts on language industry trends? And what are these impacts, if any?”

A Look at the Data

Let’s start with the facts. In 2024, the global language services market reached USD $71.7 billion (up 5.6% year-over-year). For the first time, two companies exceeded a billion dollars in revenue. Just below them, other players sit in the $800–900 million range. Yet the industry’s structure remains highly fragmented, with the top 100 companies covering only approximately 19.7% of the total market value and the top 10 just 9.6%; meanwhile, 85% of the market is still in the hands of small- and medium-sized enterprises (SMEs). The point isn’t just how much the market is growing — it’s how it’s growing. A significant portion of recent momentum is inorganic. The 2024–2025 period saw an acceleration of consolidation, from Keywords’ privatization (a multi-billion-dollar deal) to Propio’s leapfrog growth through acquisitions of ULG, Akorbi, and ASL Services. Where like-for-like data can be isolated, growth appears more modest; RWS, for example, reported essentially flat revenue on a constant perimeter basis and a slight decline in reported figures. In other words, the aggregate pie is expanding partly because many pieces of the pie are being consolidated under the same brands. Here, we return to the question of AI’s effect. The quantitative signals are already there. In Europe, 78% of LSPs say they plan to increase their use of machine translation post editing (MTPE). We’re seeing cost reductions of 30–50% and editing time cuts of up to 63% with large language model (LLM)-supported workflows. RWS attributes more than 25% of its revenue to AI-enabled products and services; TransPerfect has explicitly stated that automation savings have cooled growth in traditional services because they’ve been passed on to clients. In other words, AI is compressing unit margins and shifting the revenue mix. Mergers and acquisitions (M&A), meanwhile, support aggregate revenue volume and recalibrate market shares. In sum, the data tell us:

• Global demand for multilingual content is growing faster than AI’s capacity to “crush” its costs; the market isn’t retreating.
• AI isn’t eliminating work, but rather recomposing it: less pure human translation; more MTPE; more orchestration (of processes, quality, and data governance); and more integration into client workflows.
• The “visible” growth is partly accounting-driven (acquisitions), while “organic” growth (volumes/prices at constant perimeter) is more subtle and depends on the ability to offer value beyond the “word” (such as service-level agreements (SLAs), security, compliance, integrations, and data).

In short, the “AI effect” on the language sector is already measurable: lower price per word, higher productivity, more polarization between those who truly integrate (data, automation, or regulated verticals) and those stuck with the piecework model. It’s not the end of the market — it’s a structural reorganization in which innovation lowers unit margins but raises the competitive bar and shifts value creation upstream and downstream from the translated sentence. These data demonstrate how AI is already imposing a fundamental shift on the sector, with the real variable being not whether AI will “destroy” language work, but who will occupy the new links in the chain: data and model readiness, quality operations, security and compliance, and application integration. Everything else is inertia — the fatal comfort zone that blinds us even as new scenarios unfold before our eyes.

It’s Time to Take the Red Pill

“You take the blue pill — the story ends, you wake up in your bed and believe whatever you want to believe. You take the red pill — you stay in Wonderland and I show you how deep the rabbit hole goes.” So says Morpheus to Neo in the movie The Matrix. And so I say to you, colleagues and friends in the language industry. The red pill is the recognition of the radical and irreversible change happening in our sector, of which we’re experiencing only the early signs today. We may be in the early Schumpeterian destructive phase and don’t know what form the sector will take once things mature. Despite the unknowns, the worst thing we can do is fail to objectively analyze the reality in front of us — not only aggregate data, but also what we see in our own companies. Machine translation (MT) is now used in almost every scenario, and automatic post-editing tools based on LLMs are spreading, promising to produce post-edited texts of quality that is nearly on par with what a human can guarantee — or at least texts that can be edited by a human in less time. Denying this reality and spreading excessively pessimistic visions or, conversely, indulging in excessive optimism doesn’t help anyone. So, let’s not tell ourselves stories. No, it’s not true that there will be room for everyone. No, it’s not true that we don’t have or won’t have an excess of freelance linguists. In many sectors, AI already has the effect of favoring the highly skilled and disadvantaging low-skilled or entry-level profiles. But by definition, the highly skilled are, in a Gaussian distribution of competencies, very few. And those few will have productivity sufficient to replace several low-skilled or entry-level people, who represent a good portion of all potential workers. Many say that language work will be replaced by consulting work from professionals like “workflow automation consultants.” That is certainly the case, but the number of those professionals will be far fewer than the current number of workers in the language sector. If we think about it, this is exactly what Schumpeter said: With new, more efficient technologies, the higher-productivity worker will earn more, and those expelled from the sector will have to change jobs. Is all this really such a bad thing? No, it’s not. What’s bad is continuing to take the blue pill and saying that things won’t change and that there will be room for everyone. The only way to help people, especially young people in training, is to tell them the truth; only then can they make informed decisions about their professional future.