Translation is often treated as a final production step: write content in one language, send it through a separate workflow and wait for a reviewed version to return. DeepL argues that this model is becoming an infrastructure problem as international companies produce more material, serve more markets and expect conversations to happen instantly.
Its research finds that content volumes have risen 50% since 2023, yet 35% of international businesses still manage translation entirely by hand. Another 33% combine traditional automation with systematic human review. Only 17% use next-generation language AI, leaving most organisations dependent on processes that do not scale with demand.
Language technology is therefore moving closer to the systems where work begins. Instead of operating as a detached tool, it can support meetings, customer conversations, documents and software workflows as they happen.
Language becomes part of daily operations
Companies are investing for concrete business reasons. DeepL identifies international expansion as a driver for 33% of respondents, followed by sales and marketing at 26%, customer support at 23%, and legal and finance work at 22%.
Each area creates different requirements. Marketing needs consistent tone. Support teams need speed and conversational context. Legal and financial users need precision, controls and a reliable record. A generic translation box cannot easily satisfy all of them.
Voice illustrates the shift particularly clearly. While 32% of executives regard real-time voice translation as essential today, 54% expect it to be essential during 2026. Expectations also vary sharply by market: 48% in the United Kingdom, 33% in France and 11% in Japan currently see it that way.
Those differences make a uniform rollout difficult. Language infrastructure must account for local adoption, quality expectations and the circumstances in which people will still require expert review. Automation can remove delay without pretending every communication carries the same risk.
Agents extend translation into complete workflows
DeepL serves 200,000 business customers across 228 markets. It says 2,000 customers are deploying agents for tasks including report analysis, sales targeting and legal review. Its DeepL Agent, made generally available in November 2025, can navigate tools such as CRM systems, email, calendars and project software without complex integrations.
That changes the unit of automation. The task is no longer merely translating a paragraph. An agent might locate a report, interpret it in another language, extract relevant findings and prepare a follow-up inside the application where a team already works.
The benefit is less switching between systems and fewer manual handoffs. The risk is that an error can travel further. A mistranslated sentence in a standalone document is one problem; the same mistake used to update a customer record or inform a legal decision is another.
DeepL highlights ISO 27001, SOC 2 Type II and GDPR alignment, along with bring-your-own-key encryption that lets customers withdraw access. Such controls matter when language systems process contracts, internal messages and customer histories rather than public text.
Modernisation depends on trust and workflow design
DeepL reports that 71% of leaders prioritise workflow transformation, even though only 17% have modernised language operations. The gap suggests language remains an overlooked dependency inside broader AI programmes.
Organisations can close it by identifying where multilingual work creates delay, duplication or inconsistent service. They should then decide which steps can run automatically, which require review and how quality will be measured for each language and domain.
Human expertise does not disappear. It moves toward high-consequence decisions, cultural nuance and exception handling while machines deal with repetitive volume. That division only works when users can see what the system did and correct it before an error propagates.
DeepL’s infrastructure argument is compelling because language sits across nearly every international process. A company cannot claim to have transformed customer service, sales or collaboration if multilingual communication remains trapped in a slow side channel.
The next stage of language AI will be judged less by an impressive isolated translation and more by whether it helps people complete work securely across markets. When embedded thoughtfully, it can shorten the distance between a question in one language and useful action in another. That makes language capability not an accessory to enterprise software, but part of the connective tissue of a global organisation.
