Change(s) ahead
What does the future hold for localization? A reasonable response would be something like, “We don’t know all the details yet, but AI is going to change everything.” That may prove to be the case, but when I look at what’s being asked of the localization team in my organization, and when I speak to my colleagues in software localization at other organizations, I get the impression that a more dramatic reorientation is needed than the one AI appears to have in store for us.
In my view, the situation has an ironic element to it because AI, at least the way it’s currently being deployed, is acting as a stalwart against the kind of change that we, as localizers, will need to stay relevant. The risk of obsolescence stems from the fact that the tools of e-commerce and product growth have become razor-sharp over the past 15 years, and their asks of localization teams are shifting to suit new strategies, tactics, and ways of working. If we stay the course, even with all the promised innovations of AI, we risk becoming obsolete.
My experience of the shifting needs at Canva can serve as an example. The big challenges the team addressed just five years ago were focused on scale. We quickly launched in 100 languages, took on a project producing hundreds of thousands of localized templates, and translated tens of thousands of words each week. But in 2024, our goals are not to launch in 100 more languages (though we’d like to…), produce a million localized templates, or translate hundreds of thousands of words each week. On the contrary, our team has narrowed our scope to a few areas of the product. But the depth and meaningfulness of our localization work has increased exponentially. For the recent Canva Create event and associated product launches, the team produced copy that had been meticulously crafted and undergone user testing and internal vetting. They provided visual media to product and marketing teams which utilized localized brand assets, photography, and talent. And they provided localized onboarding materials with bespoke screen recordings and guided exercises for customers to try the new features being launched.
Two eras of software localization
Part of this shift is surely just the natural evolution of a localization team and the broader organization it’s a part of. But some observations stand out to me: our reliance on the typical machinery of “the localization industry” (i.e., LSPs and TMSes) shrunk during this evolution, our stakeholders are happier for the change, and AI hasn’t been central to the innovations we’ve seen.
This all seems at odds with the mooted “AI content explosion.” What’s driving this trajectory from the localization industry and its preferred model and tools? From my vantage point, it seems that LSPs and localization software have remained fixated on meeting the needs of a couple of project archetypes: (1) The simultaneous launch of software products in many languages and (2) the reproductions of voluminous content marketing and product support documentation.
Faced with challenges of these sorts for 40-odd years, the industry got really good at producing “direct” translations quickly and cheaply. The way it achieved that, roughly speaking, was by minimizing the role of slow and expensive human creativity. The Localization Industry is a goliath supply chain of LSPs providing (ideally interchangeable) linguists and a handful of TMS players offering the same set of technologies (Computer Assisted Translation, translation memory, glossaries, and post-edited machine translation), which are all elaborations of this same trajectory. And so with GenAI, which aims to further compartmentalize the surface area of human creativity.
Meanwhile, the growth functions of software (and probably other) industries have not been static at all. Twenty-odd years ago, and just a few hundred meters from Canva’s headquarters in Sydney, Australia, Atlassian pioneered “product-led growth” (or PLG) as a means of selling software. In a nutshell, product-led growth stands in contrast to a traditional sales growth function. So, where once software businesses grew revenue by hiring salespeople to do cold calls, follow up on leads, etc., many software businesses today allow potential customers to use the software for free and provide some pathway for them to upgrade to a subscription. Salespeople are often still involved in the process, but the point is that a customer’s decision to purchase (or not) now happens within the product.
More or more meaningful?
That should be a seismic change for localization because, in the sales-led model, the bit that needs to be fine-tuned for each market is the salesperson. Salespeople are great at this — their job is to understand their potential customers, anticipate their needs, and funnel product feedback to the developer teams. But in the product-led model, the bit that needs to be fine-tuned for each market is the product itself and, as a consequence, we’ve seen for many years now an explosion in the attention paid to onboarding, checkout, and customer retention surfaces, which are scrutinized and iterated upon by product designers, UX researchers, and product managers as incredibly valuable and impactful surfaces.
This is the crux of a decade’s change in my view: the value of software product surfaces has increased with the sophistication of product growth teams, and with that, the incentives have shifted from producing a lot of multilingual content quickly and cheaply to producing very high-value content that creates measurable impact in target markets. The required function of localization in this new model is the exact opposite of what the localization industry has evolved to serve and which AI is currently accelerating; the challenge is not to remove slow and expensive human creativity from the process but to incorporate more of it.
Risk and opportunity
If I’m right about all this, the future of localization is about understanding customers and creating experiences that move them through the customer journey toward purchasing. If the seller side of the localization industry will be more than a small part of this future, it will need to adapt.
What might that look like? On the software side, localization will need tooling that operates further upstream in the tech stack to enable far more complex content variants. The target should no longer be duplicating strings of copy in multiple languages but allowing for divergent blocks of design content: CTA buttons, modals, banners, and so on. On the services side, the expertise that LSPs provide will need to become deeper and broader. Linguists still have a (substantial) place, but the required expertise goes beyond this into UI design, UX research, and content strategy. These skills are much harder to deliver at the arm’s length of an agency partner but also provide much more scope for services, which help bridge the stubborn divide between localization services and measurable impact — the elusive “ROI” of localization.
These few comments are admittedly far from being a roadmap to more successful localization teams and more suitable language services and software. I want to convey what my experience at Canva suggests about the future of client-side localizers and the opportunity that creates on the seller side. My experience at Canva tells me that client-side localizers can’t let their impact be molded by what the industry wants to sell us. Nonetheless, the localization landscape will continue to demand fractional rather than permanent talent — very few companies operate at the scale necessary to be able to hire in-house teams delivering multilingual UX research, visual design, and product expertise — and this gap isn’t currently being filled by AI or other service niches. The story is similar on the software side, where tooling is too tightly oriented towards strings of copy instead of the larger blocks of design that drive impact in PLG companies. I’m confident that someone will bridge this gap. I’d like the opportunity to add to rather than take from our industry.
