Artificial intelligence (AI) is everywhere and talked about by everyone. What AI is and what it is not has already been exploited by countless authors. Also much has been written about the impact of AI on people in general and in specific fields. I do not want to presume to judge whether all this already is or will be true in the future.
However, I would like to allow myself a personal assessment in an area in which I have been working for over 20 years and where I provide strategic advice to customers.
Impact of AI on language operating models
Currently, I am reminded of my dark past, when Machine Translation came ‘out of the closet’ around 2012 in the form of Neural Machine Translation (NMT) – thanks to AI. All of a sudden, this technology, previously ridiculed as a pure research topic, rapidly became accepted by the general public and in the industry, the latter taking quite a while. Now, the industry relies on NMT in all efficient translation pipelines, whether via service providers or internally, with self-trained engines.
The question of whether well-trained translators like to check machine translations all day at high speed for less money is one that everyone can answer for themselves.
One can go back even further: When translation memory was introduced in the 1970s, the software was initially labeled as ‘the devil’s work’ and rejected by translators. Today, there simply exists no efficient translation process without TM and NMT.
Enters AI
The advent of generative AI (Artificial Intelligence, AI), i.e. ChatGPT-like AI that ‘seems to know everything’ and ‘can master all language and image-based tasks’, is supposed to make our lives even easier. First and foremost, however, AI raises questions for our customers:
- What does AI mean for our language operating models?
- What will AI change for us in the near future?
- What do we need to prepare ourselves and our language operating models for?
But first: What is a language operating model?
When I talk about language operating (short LangOps) models I am referring to the way our customers, i.e. industrial companies, produce translations for their internal and external customers.
The spectrum ranges from ‘everything internal’, i.e. within the company, to ‘everything external’, i.e. with the support of one or more service providers. ‘Everything’ refers to all systems, data, and services needed to create translations for a company. Hybrid operating models are often present, with varying degrees of externalization. Many companies have recognized the value of their multilingual data at an early stage and have taken care of internal repositories, e.g. in the form of translation memories. Today, the actual translation activity is largely carried out by external service providers.
So, when you look at the question of the language operating model of the future, you look at it from the perspective of your company’s current position and strategic challenges.
If you have already externalized everything, for example, you could sit back and say: I will continue to give everything to a fitting language service provider (short: LSP), who will take care of all this AI stuff for me.
This raises new questions
What does ‘fitting’ mean in the context of AI and LSPs? What do I expect from my LSPs? What do my stakeholders expect from me? How does the overall service offering evolve and which LSPs will have the upper hand in the ‘AI game’ in the future? In a constantly changing translation service market, it is already very difficult to choose the right external partner. What happens to my valuable data? And: Is it wise to hand everything over to a single service provider (keyword ‘vendor lock’)?
If you already have a ‘hybrid’ approach, e.g. you manage your translation data internally with a controlled level of quality, have possibly set up an end-to-end translation management process with TM and MT, and have achieved a high degree of automation, you are in a completely different starting position.
Adapting the language operating model to the future of AI also depends heavily on the company’s AI strategy and what the stakeholders (will) need. Are editorial processes already planned with AI? Will they ‘generate’ more in foreign languages with AI in the future? How important is reliable quality for certain types of information? Do they perhaps need foreign-language experts instead of translators for testing and quality assurance of data and processes? Do these have to be internal experts or could this be a future service by LSPs?
LangOps as strategic player
We are currently seeing a lot of AI co-creation in the language sector, between departments with service providers, and other partners. Data and knowledge silos are being torn down a lot faster than was the case just recently. Best practices do not (really) exist yet, but with this multitude of future-oriented pilots and projects new language processes and visions are crafted.
Personally, this development gives me hope: until now, LangOps has been more of an entrepreneurial stepchild, required but unloved. Now LangOps becomes an important player for AI projects in the company, as data provider, as well as process consultant and expertise supplier. This is an ideal base to build-upon for more visibility inside the company. Some tips for more visibility for language topics in the company can be found here.
Feed your own machine
Finally, an appeal: The TM may indeed die at some point in the future (…there is life in the old dog yet…) but TM data will continue to live and good multilingual data is needed badly. AI is already ‘starving’ and methods of synthetic data generation are being worked on. Currently, it looks as if the AI machine is devouring itself.
Whatever you do, take care of your valuable multilingual data, especially when it comes to terminology. After all, how can companies still stand out in these times of ‘generic automated text creation with AI’? With unique corporate language, corporate style, and multilingual terminology.
I look forward to your comments and suggestions! And if you would like to talk about your language operating model of the future, I would be delighted to hear from you. Feel free to contact me.
