Three voices from Learning Technologies 2026 on edtech localisation, the multilingual workforce, and where human translation still earns its place.
The Learning Technologies (and the co-located HR Technologies) show took over ExCeL London a few weeks ago, allegedly attracting 15,000+ L&D, HR and training professionals from dozens of countries… and I was one of the marketers in attendance, nosily exploring the many and varied tech platforms on display.
Edtech is a pretty fruitful niche for the ICS team given the scale at which we work and the specialism my colleagues have in translation for technology, games and interactive media.
Entertainingly, one of the main things I noticed was how busy the speaking areas stayed from open to close - a lot of people genuinely interested in learning, at a conference about learning. Possibly related to this, I found people outstandingly open to chit-chatting with a stranger (i.e. me), and it's from those chats that this panel blog has emerged, with new friends lending their time to answer our questions.
So - here's the panel:
On the topic of machine versus human translation, as you'll see they agree more often than not. Where they don't, there's food for thought. Let's take a look -
"If the goal is information, let the machine handle it. If the goal is influence, you need a human." - Nieves Cohen, edX
The question: What international marketing strategies do you expect learning and edtech brands to be pursuing in 2026?
Nieves was blunt about how far things have moved. "AI is no longer a differentiator; it is the operational baseline," she said. Her three pillars: compliance as a selling point rather than a checkbox, with "the EU AI Act setting a high bar compared to more permissive markets"; the platform as "infrastructure" rather than a bolt-on tool; and what she called an "efficacy reckoning", where "decision-makers are moving away from surface-level engagement metrics in favour of verified learning outcomes." Her verdict: "the real international marketing battleground has shifted to proving trust, scalability, and impact."
Dan, understandably given our focus on the power of human translation to minimise regulatory and business risk and maximise opportunity, came at the same question differently and landed somewhere complementary.
"Gone are the days when you could launch in English-speaking markets, translate this into your target markets and supplement this by running paid acquisition campaigns," he said. "Organisations are increasingly looking to build local trust signals, as opposed to global brand authority, and stand out via local expertise, authenticity, cultural relevance and domain expertise."
He also flagged community-led growth - local webinars, customer advocates, regional thought leadership - as buyers lean harder on peer recommendation to, once again, minimise risk.
Lola kept the focus on substance, expecting "a focus on skills development… that address local employability and skills gaps" alongside "the promotion of customisable learning paths that cater to diverse learning styles and needs in different regions."
Three lenses there - infrastructure, local trust, learning design - and marketers have plenty to chew on, because generic "global" brand-building is steadily losing ground to proof and local relevance. On the show floor that showed up in the marketing itself: brands working hard to demonstrate expertise - named experts, research, case studies - rather than simply assert reach, which is the same instinct behind Nieves's efficacy point.
The question: What do international customers consider when looking to evaluate edtech and learning platforms?
This is the question I was most curious to ask - what actually makes an international buyer say yes - and it's where the three of them harmonised most.
Nieves described a move from "content-first" to "capability over content", with budgets tighter than they were five years ago. "Buyers have moved past vanity metrics like 'time-on-site' or 'number of course completions'," she said, in favour of "Speed-to-Competency" and a direct line between a learning path and a real business problem. They also want platforms that slot into existing CRM, HRIS and performance workflows rather than adding what she nicely termed "digital clutter".
Dan boiled buyer thinking down to three questions: "Will employees engage? Will learners complete courses? What will the adoption rate look like?"
A platform stacked with features but low on adoption, he noted, "ultimately isn't serving its purpose." The deeper shift he identified runs "from 'did they finish the course?' to 'what's changed in terms of skills acquisition and performance?'" - with integration, AI governance and language accessibility as the other big determiners.
Lola anchored her answer in fundamentals: "at the core, quality content and compliance to regulations and standards," then "user experience, support, ease of navigation and accessibility, interfaces in multiple languages, aligned with local cultures," and a visible "track record and testimonials."
Nieves and Dan, from very different backgrounds, landed in much the same place: buyers want ecosystems rather than standalone tools, and both raised AI governance without being asked.
The question: Which translation and localisation processes most depend on human expertise, versus machine translation?
Nieves reckons we've "moved past the 'human versus machine' debate" and put a number on it: "roughly 80% of standard content - product documentation, high-volume support articles, internal comms - now runs through AI and machine translation," freeing human experts for the critical 20% she sees as shaping brand impact. She split that 20% into cultural transcreation (where AI handles accuracy but "struggles with the 'persuasion gap'"), high-stakes risk management in regulated sectors, and what she called "strategic brand voice architecture" - humans "defining glossaries and tone-of-voice parameters rather than translating line by line."
Dan reached the same place by content type.
"Machine translation for scale, human expertise for judgement." - Daniel Rennie, ICS-translate
Machine for the informational, repetitive and high-volume; human for marketing, learning content, assessment materials and compliance. His point on learning content is worth mulling over, in edtech and well beyond it: learning content "involves engagement, comprehension and motivation," and "awkward wording or phrasing isn't the main risk - reduced effectiveness is." Given how many people use learning software begrudgingly, a clunky or flat translation can quietly lose the learner.
Lola reinforced the human side, naming her top three: "ensuring content resonates with target market; creative content - in particular focusing on psychological impact"; and anything involving "legalities and regulatory compliance."
What's striking is that a translation buyer and a translation supplier, coming at it from opposite ends, ended up with the same rule. The human-versus-machine debate has settled into a division of labour, and both put compliance and persuasion firmly in the human column.
There's a part of that human column none of the three quite got to, though, and I can't ignore how people actually experience translated learning content, because so much of it now lives on the screen. A lot of what I saw at ExCeL isn't really "words on a page" at all - it's copy squeezed into microlearning apps, mobile-first courseware, gamified training, branching scenarios and chirpy little AI tutors.
Translated copy stubbornly expands or shrinks, scripts run the other way, character limits bite, and someone suddenly has to rebuild the slide so the new text both fits and reads like a human wrote it.
One small word can wreck the UX… "Save" might mean store a file or save your progress in a game, and a machine with no idea what's on screen will happily pick the wrong one. Localising e-learning content, it turns out, is as much about layout, UX and regional pedagogy as it is about language.
This issue is more acute in learning tech than almost anywhere else because, let's be honest, very few people work through a compliance module on a wet Tuesday for the sheer joy of it - they do it because the job says so.
There's precisely zero patience among business users sitting through mandatory training when it comes to clumsy localisation or ‘accurate’ translation that makes the UX more of a challenge to engage with.
The question: How does multilingual functionality in career, HR and learning software help companies onboard and develop teams?
This drew the most varied answers of the four.
"Multilingual functionality… promotes cultural sensitivity, minimises unnecessary frustrations and helps content meet target audience expectations." - Dr Lola-Peach Martins, Fir & Myrtle
That was Lola's framing - inclusion and cultural fit first.
Nieves pushed past inclusion language toward productivity. Onboarding people in their first language, she said, removes the cognitive "tax" of translation, lifting comprehension and retention so new hires get productive faster.
She also made a standardisation argument - multilingual platforms as a "single source of truth", so "a team in Manchester, a team in Tokyo and a team in São Paulo are all operating under the same set of expectations" - before landing on a neat closing line: "the 'language gap' is increasingly being bridged by technology, but the 'cultural alignment' is what organisations are now chasing."
Dan went furthest, and gently contradicted the inclusivity framing head-on. "The key outcome here isn't necessarily increased inclusivity, but more increased operational effectiveness," he said, listing smoother onboarding, better compliance, stronger engagement, internal mobility and retention.
Three readings of the same feature, then - Lola as cultural sensitivity, Nieves as productivity, Dan as operational effectiveness - which is really one effect viewed from three seats.
Step back from the four questions and the same movement shows up each time: across marketing, evaluation, translation and onboarding, the panel kept describing a shift away from volume, features and surface metrics, and toward proof, relevance and judgement.
AI's assistive role is largely assumed now. What's doing the work is human subject-matter expertise - plus a genuine affinity for the customer - to land the engagement metrics and learning outcomes edtech brands need to make the business case. Which means someone still has to judge which content carries risk, which carries persuasion, and which markets need more than a literal translation.
That maps neatly onto how good localisation already works: machine speed for scale, human expertise for the parts that decide whether a message lands. It's the model Nieves and Dan both described, and roughly how things run at ICS-translate day to day - machine translation post-editing to handle the high-volume material at pace, then transcreation, multilingual desktop publishing and in-situ review for the learner-facing content, where a linguist checks the words in place on the actual screen.
That last step is how we keep turning up engagement fixes in content other suppliers had translated perfectly correctly and left utterly flat. The ISO 17100:2015 certification behind it all was earned in iGaming, about as regulated and unforgiving a proving ground for accuracy as the tech industry offers. Which, on the evidence of this panel, is the whole game in 2026 - knowing where to spend the human judgement, and where to let the machine run.
Thinking about the show as a whole, two things stood out across the stands and sessions:
As you would expect - AI was everywhere, but arguably (and as the panellists allude) the more useful conversations were about what happens once AI stops being the novelty and becomes a component part of how translation initiatives take shape in future… with the safeguard of human subject-matter expertise.
With thanks to Dr Lola-Peach Martins (Fir & Myrtle), Nieves Cohen (edX) and Daniel Rennie (ICS-translate) for taking part.
Related reading from ICS-translate: MTPE in 2026: When It Works, When It Doesn't, and What It Costs, which includes a section on full post-editing for screen-based learning content, and When to Switch MTPE Agency: A Practical Guide and Red Flags.
