BY RODRIGO FUENTES CORRADI
Language service providers (LSPs), custodians of the language industry, must quickly repurpose people, technology, and processes to establish a long-term role in assisting organizations in creating generative AI (GenAI) roadmaps based on unique internal data. Without a new alignment, future LSP offerings will not meet enterprise AI goals and potentially drive such individuals to search beyond the localization industry for new collaborations and holistic offerings. Furthermore, the localization companies will lose traction with the linguistic subject matter experts (SMEs) community, as other language AI opportunities and opportunity providers will take better advantage of their talents.
Advertisement
The mirror image of an enterprise organization is its data.
As we know, the enterprise comprises various departmental components, yet we need to feed it quality data for AI to function holistically. This data must be gathered carefully and systematically across all corporate stakeholders and their global organizations. This is pertinent when we consider that to achieve global success, they rely on large LSPs to translate and localize their content — the ubiquitous “breaking the language barrier” ambition.
In light of this, it’s critical to weigh the benefits traditional LSPs offer and the possibility that their roles will eventually extend beyond language. LSPs provide scale and attempt to address every possible internal and external language need that their customers require to compete in the global market. However, while these huge client organizations provide multilingual content to global markets, it could also be necessary to step back and re-evaluate their goals from a future AI perspective.
Examining large organizations allows us to fully understand their goals and the links between content challenges and their early and tentative AI objectives. The opportunity for LSPs is to shadow their customers’ future AI roadmaps while managing their global content and uncover synergies between these two activities. Building a broad, more future-proof commercial offering is the ultimate objective with this approach.
Starting with the enterprise organization, we know that each enterprise department likely manages data differently, making data difficult to understand for everyone else. Departments’ data cadences, tools, systems, and processes frequently differ. As a result, unifying the enterprise with a single internal vision of data, communicating with a single internal voice, and later developing with a single outward message to its clients is currently impossible.
As a result, confidence is lost as crucial company decisions are quickly impacted, and the ability to course-correct has no supporting data or explanation.
Only an Al framework can solve this difficulty at scale while serving as a single source of truth and an eventual co-pilot to all enterprise stakeholders, including marketing, human resources, product and services, sales, and operations.
Enterprise organizations require time to settle, undergoing an introspective process to audit current technology, processes, and employee skill sets. This will result in a progressive, fit-for-purpose roadmap for future-proofing their operational and commercial goals, allowing them to develop a comprehensive data and AI capability strategy that includes all internal stakeholders.
Like all other technologies, AI has the potential to be disruptive, but it is actually neutral and not inherently divisive. Everything is contingent upon the deployment plan. It’s actually both possible and likely that it will operate as enterprise’s uniting factor.
At this early stage, the enterprise must ensure it has a data strategy driven by specialists so that data is diverse and representative of the overall organization. Thus, data becomes the currency of trust between enterprise players.
To understand where LSPs and their clients’ AI ambitions converge, it’s important to detail the link between data, information, content, and language.
Recognizing certain fundamental traits of content categories, such as what constitutes “data content,” “informational content,” and “nuanced content,” could be a helpful place to start to understand the broad component-based content that permeates organizations.
It is possible to distinguish data from both information and content. Each has similar characteristics, but how they are used to make business decisions and operations differ.
Data is key to all future business goals. It contributes to the creation and enhancement of our machine-learning models. Data exists in various formats (image, text, audio) and will be structured and unstructured. It can also be multilingual data, owing to the localization industry. It provides firms with a data multiplier to compete globally and ensures that established and emerging markets have trustworthy Al solutions.
In terms of information, it should be precise and actionable for business processes to be successfully followed. On the other hand, business-generated information is essential for maintaining effective operations and supporting their products and services, which are essential for customer satisfaction.
Enterprise content, however, can be characterized as purposefully nuanced. It is utilized both externally and internally. It encourages congruence with the organization’s branding and messaging and fortifies connections amongst internal stakeholders, often by human resources departments following executive messaging.
Each of these content categories is valuable and must be harvested or produced within a process that is relevant, cost-effective, and, most importantly, the foundation for company activities both internally and externally.
As a result, the strategy for achieving a more cohesive enterprise is to consider data as the starting point and content as the outcome, traveling the road via informational content that shapes enterprise business decisions and outcomes. Future enterprise alignment will likely rely heavily on a data and content mapping paradigm developed by interdisciplinary enterprise SMEs to curate the initial source data content in the form of previously mentioned enterprise stakeholders such as marketing, human resources, product and services, sales, operations, and more.
Considering these distinctions and observations, we can move toward the important question of designing the proper workflows within the enterprise that accommodate the value and purpose of data, information, and content. Eventually, we can also properly identify, classify, and label all of these components so they can once more be used as data.
Now, with the fundamentals of potential data, information, and content within the organization understood, we can switch our attention to the evolving role of content localization by LSPs and language specialists.
We typically acknowledge the plethora of enterprise content. Nevertheless, to be honest, not all content is created equal. Some content is more equal than others due to time restrictions or the requirement for rapid actionable outcomes and commercial decisions. However, to address this multilingual content dilemma, we are witnessing an expansion in the field of language toolsets that provide content insight, machine translation quality estimation (MTQE), and machine translation (MT) for machine translation post-editing (MTPE).
As a result, early adopters of intelligent content strategies based on these tools can distinguish between “data content,” “informational content,” and “nuanced content” and then map its value and apply the appropriate localization approach. However, the potential repurposing of these localization methodologies, supported by industry-specific tooling, might allow LSPs to expand their offerings to a broader and concurrent AI, data, and content paradigm.
Traditional LSP technology has focused on moving content from its origins in enterprise to its destination, back to enterprise, and subsequently to that enterprise’s worldwide client base. However, while content is currently at the core of the localization sector, data drives the developing AI paradigm and constitutes a much larger opportunity in our global economy. And although AI-based technology is emerging to boost both output and, more crucially, the value proposition of AI-led tools to the whole content flow, the localization industry offering remains a content-centric workflow geared primarily to increase productivity.
The traditional localization method must be unpacked and repackaged to meet with final business goals of an AI paradigm. Importantly, progress has already been achieved regarding the future function of LSPs in the larger future content narrative, both in terms of human roles and component technologies based on content enabling. As a result, many components are already conceptually in place. Still, localization activities should now include the byproduct of data for emerging large language models (LLMs) like ChatGPT and aggregating component AI-powered tools.
Meanwhile, in the case of MT, human editing behavior and final quality, both technology and humans would regularly review performance. This ensures that technology investments pay out in return on investment (ROI). Of course, the feedback from this performance data will be used to develop AI in the future.
When repurposing enterprise localization platforms for future AI development, the best of technology should combine with human SME experience to provide feedback and calibrate the technology to improve effectiveness and safeguard ROI objectives. Importantly, the ecology should be seamlessly integrated, including providing multivariate data formats that feed AI models with localized text, video, and image. This new language/human and data paradigm will benefit the language ecosystem, including language technology providers, LSPs, translators, and customers eager to maximize data for their AI roadmaps.
Therefore, the next stage may be to expand the amount of AI technology that is both focused on increasing productivity and ready for data collecting, such that data harvesting becomes a mission-important exercise in and of itself.
Technology in the localization industry will continue to erode conventional service-based revenue whilst increasing technology-driven margins over time (but always dependent on the deployment quality). The ability to balance between these shifting dynamics continues to challenge every LSP CEO since MT first became scalable and trustworthy with statistical MT (SMT). At that time, the industry was not yet ready to embrace this technology wholeheartedly, and the choice to move ahead was still siloed depending on language and domain. This meant that customers were yet to fully understand concretely the possibilities or benefit from them.
Then came neural MT (NMT) with its exponential leap in quality, and now LLMs. The issues are burning red hot as localization customers — and importantly, their CEOs — gain awareness and savvy about the increase in productivity tools and push for increased discounts and faster turn-around times. The reason is simple: They, too, must protect their margins and attempt to remain competitive. However, there are other threats to the LSP offering as the latest technology in the form of LLMs sweeps through and captures everyone’s attention.
Taking a quick step back, language technology was first predicated on storing validated translation units in translation memories (TM). These were there to create linguistic consistency but also created savings for customers and reduced timelines by eliminating translating content that had been processed previously. More recently, TM became the building blocks for optimal customized MT that could increase productivity and help retain key linguistic requirements such as style and terminology. Importantly, these advances were underpinned by translation management systems (TMS), bringing the key human and increasing technological components into one ecology. Herein lies another component that is now not just a mechanism for delivery for the LPS organization but morphed into something that fundamentally impacts established revenue for LSPs.
TMS, once a means to an end for moving files, may now shift and become the main point where component technologies such as TM, MT, MTQE, and others compete and combine to provide value to end customers. While creating new steps in the localization process, each of these component technologies will also need best practices and governance in terms of quality, transparency, ROI, and security.
Most LSPs will look, as they have previously, to partners to innovate technology and then create connectivity to the overall TMS workflow platform. Therefore, it will be the multifaceted platform that can bring in the component innovation that will endure. However, beyond the multilingual content challenge lies something more profound and goes to the core of LSP offerings: its future identity and market position. With LLMs now encompassing a varied set of linguistic tasks such as summarization for insight, content insight and quality estimation, and MT, the singularity of the LLM ecosystem could well become the destination point for many component localization requirements that the industry currently provides.
New linguistic AI technologies will increase performance and, through production, create additional performance data that customers will expect to be harvested, understood, and actioned by the LSP to increase performance further and continue to reduce costs. In the long term, this data-producing paradigm will bring increased value to LSP customers and be a firm expectation soon enough. In the last decade, the ubiquitous “breaking the language barrier” goal seemed like the ultimate destination point on the horizon. However, the next decade will require LSPs to break the “language barrier” in tandem with the “data barrier,” thus prioritizing the AI roadmap of customers as much as their content in global markets.
Global brands have increasingly understood the value of localization, leading the way by taking advantage of LSPs’ innovations. Attending localization fairs, they now both participate in and advocate further language technology improvements. This increasingly enlightened vantage point has also seen an increase in autonomy, thus seeing some original localization customers creating their internal localization teams, developing platforms and technologies such as MT, with their need for autonomy, often coupled with the need for security. However, they still often require services and engage with LSPs to discuss innovations, often working on pilots and requiring consultancy. This can come in many forms, including data services, training services for both MT model building and internal translation teams as well as testing of their technologies.
As new generative technology emerges, the journey from customer to competition may also emerge in some industries where leading consumers of localization that have created technology roadmaps and independent internal teams start to create services internally and externally. Although this trend may be slow, large global companies rarely have a unified content or localization strategy, with regions and departments coexisting with siloed processes, external providers, technologies, and, importantly, siloed data.
Thus, increasing numbers of localization customers will become localization-independent and potential competitors. Given that this was already an emerging trend, especially in those sectors where security was key, such as financial institutions that were building their own NMT technology to increase the productivity of their internal localization specialists, the expansion of GenAI will only bring the possibilities of more varied requirements when global brands, which the large LSP regularly target, reach out in future. Therefore, the role of large LSPs may require more agility, flexibility, and a growing expectation of being an SME-led partnership.
However, given the growing AI ambitions of CEOs, this consultancy may be just a staging post as business applications that service large organizations incorporate more LLM co-piloting and the multilingual experience is swallowed up into a larger AI paradigm. This is not just because of the control and security factors, but also ultimate corporate governance. When all the prohibitive factors are erased such as cost and security, what’s more pertinent to LSPs could be the growing independence of their major clients through building their own solutions.
The path from TM technology (savings and consistency) to MT technology (savings and productivity) allowed LSPs to be competitive whilst TMS technology (efficiency and control) executed the go-to-market strategies of global brands. However, GenAI marks a new paradigm that will encourage global brands to invest further in AI. As a byproduct, and due to the huge variety of capabilities, it will make language independence more possible, whether the opportunity is taken or not.
With respect to the purely human aspect, it is interesting to note that labor agencies are increasingly present and specialized in the localization field, both on social media and at localization events. One can only imagine that their commercial imperatives will see recruiting human expertise for LSPs, language-savvy global brands, or creators of GenAI, as dependent on their own commercial goals rather than any type of preference.
Thus, we see the potential for new relationships, not least that of the language supply chain experts such as translators who rely on LSPs to aggregate and distribute work opportunities. Suppose the relationship between LSPs and global brands weakens and global brands look to become further technology independent. How long will they look to come to LSPs as an aggregator and supplier of human expertise?
For now, that human element will need to catch up and more fully embrace the AI paradigm. Organizations will be motivated to reconsider their talent strategy, thanks in part to the expanding role of AI co-piloting. They must audit roles and responsibilities and repurpose and prepare individuals for the future. Additionally, to ensure that investment in technology is serviced with talent that is ready to make good on investments and achieve ROI, the AI paradigm necessitates targeted training and certification.
This last point is especially pertinent if we consider the first incarnation of MT. Due to early concerns about MT quality and translator concerns about job security and earnings, the transition to MT post-edit was slow in the localization industry. Many of these challenges, however, have been mitigated by training and certification directly related to post-editing, and we anticipate seeing similar progress on new and emerging linguistic AI technologies.
However, it should be observed that the sequence of events begins earlier. If you consider where localization, like other businesses, finds talent, it is often in formal education offered by universities.
Universities continue to offer translation courses and develop the traditional translator persona but are increasingly migrating to a more technology-based curriculum. Later, following graduation, localization companies continued to develop translators in technology and customer requirements, producing localization experts. Meanwhile, localization buyers continue to engage competitively with their client base, centered around personalized, multilingual content.
The prevailing wind, however, is data-intensive, holistically functioning LLMs in the form of GenAI, which will eventually affect the kind of language competence necessary and how the future language expert’s journey begins.
Perhaps more important is the recognition that the talent that develops from university, which will become the foundation of the global economy where products and services are traded, can also bring about large-scale delivery of human-curated data. Acceptance of each of these roles as part of this data ecology is critical if the industry and its participants are to remain relevant.
New opportunities for translators and linguists will either come through their existing LSP work providers or other industries with direct data and AI demands — and more profitable and exciting career prospects.
Such an exodus away from LSPs can only be mitigated if LSPs play a larger role in defining GenAI expansion, repurposing their tools, processes, and technology, and introducing a new value proposition payment for linguistic talent that moves away from a pure word rate offer. Linguists will play an important role in this as they embark on continuous learning and growth cycles, particularly when focusing on AI, for which the post-editing experience was likely the first step for many.
At the heart of this is the unavoidable technological transformation and the deliberate steps LSPs take to stay a relevant work provider and maintain a meaningful future relationship with their current talent pool based on mutual benefit. There is an urgent need to discuss whether LSPs can truly capture the future potential of language talent and adjust their solution offerings accordingly.
As previously said, the journey begins at universities, followed by LSPs developing and harnessing professional skill sets. Some LSPs and providers of localization technology have had beneficial interactions. However, given the quick cycles of technology, where the period between each new paradigm occurs in less time, the link between academics and industry must be accelerated. With the AI paradigm growing, this is likely to become a requirement across industries, but especially for LSPs and their need for language experts who frequently begin their careers as translators and then evolve and outgrow that title and position, as they will continue to do.
From the perspective of universities, their focus and impact are still limited to the beginning of a career. For example, providing continuous career help to university graduates would boost their revenue and impact over the course of their employment. On the other hand, universities would gain from gathering pertinent information for their curricula. They could assist in molding students’ realistic future visions, preparing them for long-term success and enhancing their position and relevance in a future education marketplace.
Maintaining contact with students as their language careers progress would benefit all parties involved. It would build a mutually advantageous and trusted collaboration that would last for the working life, not just four or six years.
Customer organizations’ localization departments must keep growing through industry events, related reports, and collateral. But still, direct interaction is the best indicator of who might make a good longer-term AI-driven localization partner.
These frequently occur at quarterly business reviews (QBRs). During reviews, there needs to be a necessary shift in the key performance indicator (KPI) reviews towards innovation segments, in which LSPs and customer-based localization managers outline the roadmap and, crucially, establish meaningful next steps through structured AI pilots. This allows all process and quality questions to be addressed, along with an understanding of the financial implications of the new paradigm and obvious ROIs.
This evolving and more demanding localization manager role is critical to promoting change throughout the supply chain, not only to the companies they serve commercially but also to maintain a credible and relevant role within their organizations. It’s a shift from a conveyor belt of localization requests to a content provider for global audiences and eventually as linguistic SME assets on the road to greater AI usage.
Similarly, whether working for LSPs or as part of a customer localization department, translators and language specialists play an important role in developing this path from university student to professional. It is also critical to understand how AI changes the role of translators in each of these domains and whether we, as a localization industry, believe that we are the best or most appropriate home for their skills and futures.
Furthermore, despite frequently being concerned about MT’s impact, translators became the ultimate advocates for the technology through greater widespread adoption as the technology developed. As more AI features and technology become available in the industry, maintaining this feedback loop and eventual advocacy from the translation community will separate the wheat from the chaff in terms of which LSPs and language technology vendors will be trusted and can grow exponentially thanks to an augmented talent pool.
Localization is now well-established with global businesses. Still, as the commercial and professional landscape in and around localization transforms, it will be interesting to observe who has absorbed and leveraged the new AI paradigm. Essentially, the language sector’s AI credibility will be questioned. Whatever the case, any success will necessitate technological coherence within localization systems capable of incorporating new technologies augmenting human ability beyond language and providing high-end data to global enterprises.
Such a confluence of providential conditions would undoubtedly boost industry investment, elevating the industry’s present valuation — which, according to ZipDo, would reach nearly $57 billion by the end of 2022, climbing to just under $73 billion in five years.
Until now, LSPs have been the custodians of the localization industry, charged with successfully removing linguistic barriers. They have maintained a solid hold on the critical success factors, progressing from a speed, cost, and quality-driven offering to a sustainable and scalable technology-driven one.
However, the rise of GenAI may be a mixed blessing for LSPs if the challenges of incorporating AI surpass LSPs’ ability to provide new and relevant roadmaps. As a result, it may become less evident to outsiders that LSPs are and should continue to be the custodians of the key components of technology, talents, and processes when it comes to language when data is also in the mix.
Future language investors will certainly be interested in how technology and human capacities evolve. They will be wary of whimsy AI applications and instead invest in sectors that can mature and not only prosper as a result of AI, but also grow the AI paradigm by contributing data back to themselves and others. For the time being, the language industry is one such candidate industry that can optimize human talent for future economic growth.
Advertisement
So, to break this down, seven points that need to be considered and matter:
Let’s go back to the beginning. Data will be our enterprise companies’ future mirror image. In this setting, however, seeing is believing, and such organizations must have a foundation of competent technology, financially viable processes, and empowered people capabilities.
Al, when properly deployed, will unite the enterprise by creating a unified framework for strategic agreement. As a result, the enterprise is more cost-effective, adaptable, and coordinated, able to detect structural weaknesses and quickly act, establishing a point of convergence for all main enterprise stakeholders. Rather than being divisive, AI and the data it is built on will become the data well from which everyone will drink.
Crucially, LSPs must adjust their solution offerings before their large enterprise customers reach a tipping point and make irreversible investments in their own private LLMs and GenAI.
Still, just like with NMT, LSPs possess a significant advantage in that they have produced ever smaller and more specialized, cost-effective, and secure models of NMT rather than generic ones and have done so in collaboration with a vast and highly skilled SME talent pool.
So, once enterprises’ AI and data roadmaps have matured, the incumbent language industry must be ready. LSPs are in pole position, having been responsible for expanding the growth of global brands and building trusted partnerships over the last decades. However, the AI commercial possibilities are immense, and they are likely to draw many interested rivals from both within and outside the traditional language industry sector. Taking advantage of this potential will require courage, vision, and the support of its most precious asset: its relationship with linguists.
The outcome is, of course, unknown.
Rodrigo Fuentes Corradi has worked in the language industry for the past 25 years, specializing in machine translation technology and human processes and capabilities.
Advertisement
Related Articles
Awtomated has integrated Microsoft Translator into its TBMS platform, giving Language Service Providers instant access to high-quality machine translation.
→ Continue ReadingmemoQ announces strategic investment in memoQ RFP to revolutionize the RFP response process. memoQ, a leading provider of a Translation Management System (TMS) is proud…
→ Continue ReadingAt the 15th Process Innovation Challenge held at LocWorld50, Gabriel Fairman, CEO of Bureau Works, clinched the esteemed title of "2023 Innovator of the Year…
→ Continue Reading
