The assumption that quietly breaks global internal comms#
Most multilingual communication problems do not start with a translation failure. They start with the assumption that English is sufficient.
The numbers make that assumption harder to defend than it looks. English is spoken by roughly 1.35 billion people worldwide out of about 8.1 billion, and only around 380 million of those are native speakers. The remaining 950 million or so speak it as a second language at widely varying levels of proficiency. Inside a global company, that variance is invisible on an org chart. Everyone in the meeting nods.
Organizations that get caught out tend to hold three specific assumptions, and each one fails in a different way:
The translated materials guarantee understanding. Providing a translated handbook is not the same as someone comprehending it. Translation delivers words. Comprehension requires nuance, sequence, and context, especially for procedural or safety content where the order of steps carries meaning.
That employees will adapt over time. Language proficiency does improve with exposure, but "eventually" is not a communications strategy. The gap matters most in the first year, which is precisely when onboarding, safety training, and policy comprehension are most critical.
That conversational ability is enough. Someone can be fluent in a lunchroom conversation and still miss the conditional clause in a compliance policy. Real workplace communication is fast, situational, and often high-stakes, and that is where the gap between conversational and functional proficiency becomes visible.
The problem in most global companies is not awareness that the workforce is multilingual. It is the assumption layer sitting on top of that awareness.
Where language gaps actually cost you#
Language barriers tend to get filed under training or DEI rather than risk. In practice, the damage concentrates in three areas, and all three are measurable.
Safety and legal exposure. Safety instructions require understanding of nuance and sequence, not just literal word substitution. When that understanding is incomplete, incident risk rises. The US Equal Employment Opportunity Commission has emphasized that communication failures connected to national origin and language can create legal exposure for employers, which means this is not only an operational problem but a compliance one. In regulated or high-hazard environments, an employee who understood 80 percent of a procedure is a genuine liability.
Supervisor effectiveness. Frontline supervisors are usually promoted for technical skill rather than for their ability to lead a multilingual team. When a supervisor lacks the tools to explain expectations or give feedback clearly across a language gap, they often avoid the conversation entirely. Standards go unenforced, feedback goes ungiven, and accountability erodes quietly. This compounds because supervisors are simultaneously the most relied-upon communication channel in most organizations and the least supported one.
Training and knowledge transfer. Companies invest heavily in onboarding, safety training, and skills development, then deliver it in a language a portion of the audience cannot fully process. The spend happens either way. The return does not. This is the least visible of the three, because a training completion rate looks identical whether people understood the material or clicked through it.
Notice what these have in common: none of them show up in a communications dashboard. They show up as incidents, turnover, and repeat questions, which is why language rarely gets attributed as the cause.
The shadow translation problem#
Here is the part that most guides on this topic miss entirely, and it is the one worth acting on first.
If you are not translating your internal communications, your employees are doing it themselves. They are pasting policy documents, HR directives, executive memos, and restructuring announcements into free web translation tools, because they need to understand the content and you have not given them another way to.
That has a consequence beyond quality. An audit of internal communication at a major European bank found employees using unsecured internet portals for ad-hoc translation, which meant the specific language describing internal policies was exposed to the open web. This is not theoretical. In 2017, sensitive documents including resignation letters, employment contracts, and details of corporate mergers were discovered to be publicly accessible through a free translation service, because translations were stored in a way that let search engines index them. The Oslo Stock Exchange was among the organizations that responded by blocking access to the service.
The general principle is uncomfortable but simple: content entered into a free public translation service can end up in the public domain. For a routine team update, that may not matter. For an unannounced reorganization, a merger, a product launch, or anything material and non-public, it matters a great deal.
So the practical question for most global organizations is not whether to translate internal communications. It is whether translation happens inside a system you control or outside one you do not. You are already paying the cost either way.
Translation is not the same as landing#
Even when translation is sanctioned and secure, there is a second failure mode: the message technically translates but does not carry the same weight.
Tone and formality register do not map cleanly between languages. The level of directness that reads as clear and confident in American English can read as blunt in Japanese organizational language or oddly informal in German business communication. A policy that sounds firm in English and vague in its translated version effectively creates two different organizations: one where people know exactly what is expected, and one where they are guessing.
This is why serious corporate translation work starts with a style guide and a terminology glossary rather than a word count. Your company has terms that mean something specific internally, and those need to be handled consistently across every language, every time, or your policies drift apart from each other with each release.
The industry standard for handling this at scale is machine translation post-editing, where an engine produces a first draft and a human reviews, corrects, and applies subject-matter and cultural judgment. That workflow is now normal practice for high-volume, mid-stakes content. What machine translation alone still cannot reliably deliver, even now, is cultural adaptation, legal precision, regulatory compliance, and accountability for the result.
Tier your content by risk#
The most common mistake is treating all internal content with the same level of scrutiny. That either drains budget on low-stakes material or under-protects the high-stakes material. A risk tiering approach fixes this, and it is the single most useful thing to set up first.
Tier 1: low risk, machine translation is fine. Team updates, event announcements, social and culture content, internal FAQs, knowledge base articles. An error causes mild confusion and gets corrected in the next message. Route these through a secure machine translation workflow and do not over-invest.
Tier 2: medium risk, machine translation with human review. Company-wide announcements, strategy explainers, onboarding material, general training content, manager talking points. An error here causes real misunderstanding and requires re-communication, but is not a legal exposure. This is where post-editing belongs: engine drafts, a fluent speaker reviews before publication.
Tier 3: high risk, full human expert review. Safety procedures, compliance and regulatory notices, HR policy with legal implications, disciplinary and contractual language, anything financial or material. An error here creates injury, liability, or regulatory exposure. Never route these through an unreviewed automated pipeline, regardless of how good the output looks.
Two rules make this work in practice. First, assign the tier when the content is created, not when someone remembers to translate it, because retrofitting is where things slip. Second, make Tier 1 genuinely easy and secure. If your sanctioned process is slower than pasting text into a free web tool, employees will use the free web tool, and you are back to the shadow translation problem.
Why audio changes the multilingual equation#
Written translation solves for people who read comfortably in the target language. That is a narrower group than most comms teams assume, and it is where audio does something written material cannot.
Literacy levels vary within any language, including people's native one. Translating a dense policy document into someone's first language helps considerably, but a dense document is still a dense document. Frontline, field, and production employees, who are frequently the most linguistically diverse group in a global company, are also the group least likely to sit down and read a translated PDF. They also receive the least strategy and change communication of any employee group, so the gap compounds. We cover that audience in full in reaching deskless and frontline workers.
Audio in someone's own language addresses several of these at once. It works during physical work when reading is impossible. It carries tone, emphasis, and pacing, which is exactly the layer that written translation most often loses. And for anyone who processes spoken information more readily than written information, which includes many neurodivergent colleagues, it is a genuine accessibility improvement rather than a nice extra.
The honest limitation: audio is worse than text for anything people need to reference precisely. Nobody should be checking a compliance threshold or a contractual term by scrubbing back through an audio file. The right pattern for high-stakes multilingual content is an audio version for comprehension paired with a written version of record, in the same language, for reference.
Where Sprep fits, and where it does not#
This is our product, so read the next two paragraphs as an interested party talking rather than neutral guidance.
The multilingual problem in internal comms has always been production economics. Producing one polished explainer in one language is manageable. Producing it in nine is where global comms teams stop, and most teams have fewer than six people in a communications role regardless of company size.
Sprep converts documents you already have into two-host conversational podcasts, and the relevant part for multilingual work is the sequence. A person reviews, edits, and approves the master script before any audio is generated. That approved master script then translates into more than 70 languages with one click on Team plans and above. The order matters: you are approving the content once, in a language your team actually reads, and then propagating an already-vetted script rather than reviewing nine separately generated outputs and hoping they say the same thing. It is Swiss-hosted, which is the direct answer to the shadow translation and data residency problem described above. Dätwyler, a global manufacturer, uses it for onboarding and internal communications.
Where this is the wrong tool: it does not replace certified human translation for Tier 3 content. If you are distributing a regulatory notice or contractual language, that needs qualified human translators with accountability for the output, and no automated pipeline changes that. It also will not fix a message that was unclear in the source language. Translation faithfully reproduces ambiguity.
A practical starting sequence#
- Find out what languages your workforce actually uses at home. Not what they can work in. HR usually has this or can collect it, and most comms teams are surprised by the answer.
- Ask how people currently handle content they do not fully understand. This is the shadow translation audit, and it is a five-question survey. Expect free web tools, expect a bilingual colleague doing uncompensated translation work, and expect some people who simply skip the content.
- Tier your recurring content. Take your last quarter of internal communication and sort it into the three tiers above. Most teams find the majority sits in Tier 2.
- Fix Tier 3 first, then Tier 2. Tier 3 is your liability. Tier 2 is your volume and where a repeatable workflow pays off most.
- Make the sanctioned path faster than the unsanctioned one. If it is not, this will not stick.
- Add a second format for the content that is not landing. Where a translated written channel is measurably being ignored, particularly with frontline teams, test audio in the target language rather than rewriting the document again.
- Measure comprehension, not distribution. Track whether people can answer a question about the content, not whether it was sent in nine languages. This connects back to the wider point in our internal communications strategy guide: most comms functions measure output when they should be measuring outcome.
FAQ#
What is multilingual internal communication? It is the practice of delivering internal company communications, such as policies, announcements, training, and onboarding, in the languages employees actually understand rather than in a single corporate language. It covers translation, but also format, tone, and channel choices that determine whether the content is comprehended rather than just distributed.
Isn't English enough for a global company? Usually not. English is spoken by roughly 1.35 billion people globally, but only around 380 million natively, with the rest at widely varying proficiency levels. Conversational fluency also does not guarantee comprehension of procedural or legal content. The gap tends to surface in safety incidents, training that does not transfer, and policies that are followed inconsistently across regions.
Is it safe to use free online translation tools for internal documents? No, not for anything confidential. Content entered into free public translation services can end up indexed and publicly accessible. In one well-documented case, resignation letters, employment contracts, and merger details became discoverable online through a free service. If you have not given employees a sanctioned option, assume they are using an unsanctioned one.
How do you decide what to translate? Tier your content by risk. Low-risk material such as team updates and internal FAQs can go through secure machine translation. Medium-risk material such as company announcements, onboarding, and general training should use machine translation with human review. High-risk material such as safety procedures, compliance notices, and contractual language requires qualified human translation with accountability.
Is machine translation good enough for internal communications? For low and medium-risk content, yes, provided a human reviews medium-risk output before publication. Machine translation post-editing, where an engine drafts and a person edits, is now standard practice for high-volume work. What automation still does not reliably deliver is cultural adaptation, legal precision, regulatory compliance, and accountability, which is why high-risk content stays with human experts.
Why do translated internal communications still fail to land? Because tone and formality register do not map directly between languages. A policy that reads as firm in English can read as vague or as needlessly harsh in translation, which creates inconsistent expectations across regions. This is why professional corporate translation begins with a style guide and terminology glossary rather than a word count.
How do you reach frontline and deskless workers in multiple languages? Written translation reaches them poorly, since they are rarely on screen and are often the most linguistically diverse group in the company. Audio in the target language works during physical work, carries tone and emphasis that written translation loses, and does not require sustained reading. Pair it with a written version of the record for anything people need to reference precisely.
How much does multilingual internal communication cost? The main variable is workflow intensity, not language count, which is why risk tiering matters so much. Over-investing in low-risk content is the most common budget mistake. The costs that rarely get counted are the ones you are already paying: uncompensated translation work by bilingual employees, repeat questions, and training that does not transfer.
Can you produce internal audio in multiple languages without re-recording? Yes. The efficient pattern is to approve one master script in a language your team can properly review, then generate the language versions from that approved script rather than producing each one independently. This keeps the message consistent across languages and means you are reviewing content once rather than nine times.
How do you measure whether multilingual communication is working? Measure comprehension rather than distribution. Sending an announcement in nine languages proves reach, not understanding. Better signals include whether questions about the content drop after publication, whether completion rates for training hold up across language groups, and whether incident or error rates differ between regions receiving different language treatment.
See what this looks like for your team#
If you are producing internal communications for a workforce that spans several languages, we can walk you through how other global comms teams run it: approving one master script, generating the language versions from it, and keeping the whole workflow inside Swiss-hosted infrastructure rather than public translation tools.
See it in action