The LangOps Paradigm
Perceptions of machine translation
within the translation industry

By Riteba McCallum

Introduction and background

After attending GALA 2022 in April, I began to wonder if there was a gap in the way different stakeholder groups within the translation industry perceive machine translation (MT) and its usefulness. At the conference — which brought together the leaders of language service providers (LSPs) from around the world — I noticed that many of the presentations were on the topic of MT and artificial intelligence (AI), and the transformational opportunities enabled by this technology.

In one presentation, titled “LangOps – A New Paradigm Beyond Transactional Translation,” Britta Aagaard, chief business officer at Semantix, and Jochen Hummel, founder of TRADOS, claimed that “Neural machine translation (NMT) is approaching human parity for many domains and language pairs,” and that to fully capitalize on this breakthrough, language service companies need to adopt a new, multidisciplinary business model beyond transactional translation, which they call LangOps (a term coined by Unbabel co-founder João Graça).

Aargaard and Hummel argued that, on the Venn diagram of “What AI can do” and “What humans can do,” everyone is focused on where the two overlap, but what we should really be focused on is all the things AI and humans can do only when they work together, things which neither of them can do alone. They claimed that in the LangOps paradigm there would be new roles for language experts that are more multidisciplinary, more stimulating, and better paid than translation.

The enthusiastic tone of the speakers at GALA struck me as very different from the one that translators use when discussing MT. The majority of the social media posts that I see from translators on this topic are either complaining about clients forcing them to post-edit MT output (a task they consider unstimulating, frustrating, and underpaid), laughing at MT mistranslations, warning potential clients about the risks of using MT instead of a human translator, and claiming that MT will never reach human parity.

This discrepancy worried me. Is it possible, I wondered, that the business people running translation agencies think MT is amazing and plan to shift their entire business model around it, while the translators, on whose linguistic expertise this new business model depends, think MT is overrated and hurting the industry more than helping it?


To answer this question, I ran a survey with a mix of multiple choice and open-ended questions. I shared it mainly on Facebook and LinkedIn.

The survey starts with demographic questions to establish the respondent’s professional role, their language pairs if they are a translator, what type of company they work for, the continent where they are located, and their age group. I thought it would be interesting to see whether older stakeholders are more reticent about MT than younger ones.

Next, I asked how often the respondent uses MT and how they use it. I was curious to see whether the people who are most against MT are those who have not given it a fair shot, or, on the contrary, whether those who are most in favor of it are project managers and executives who do not have to use it themselves.

The next three questions ask for the respondent’s opinion about how close MT is to reaching human parity. I had to word this carefully, as “human parity” can be interpreted and defined in different ways. According to a paper published by market research firm CSA Research, the classic method of assessing human parity is the Turing Test:“Whether a bilingual human evaluator can reliably distinguish between translations produced by a human and those produced by a machine” (Lommel and Pielmeier, 2021). However, the authors point out that this is not straightforward, as it is necessary to define the scope of the test and to determine who is evaluating the translation, what criteria they are using, and how qualified they are to apply these criteria.

Another issue is that when we question whether MT produces the same quality as a human, many people think of a best-case scenario for the human: a highly skilled professional translating a text in their field of expertise, with as much time as they need to do it. However, this is not the reality in which most translations are produced. Most translations are done by average translators working with very tight deadlines and minimal context. Therefore, if you change the question about human parity to whether MT can match the quality that is typically produced by an average translator working under typical conditions, you may get a very different answer. With this in mind, I decided to ask, “In your opinion, when do you think MT will reach the same quality as what is typically produced by a human translator working under typical conditions?”

My last question was about the respondent’s level of interest in roles related to MT. If few translators are interested in becoming the human in the loop, the industry could face a major labor shortage.



The survey received 58 responses. Although there are many stakeholder groups within the translation industry, there are three main groups who responded to the survey:

  • Managers/directors: 12 people
  • Translators who are employed by an LSP: 20 people
  • Freelance translators: 26 people

Some respondents fit into more than one group (e.g. a translator who is employed by an LSP but also freelances on the side). I counted them in each of the groups they identified with.

Nine respondents identified as project managers, one as a lecturer/professor, and zero as students. I did not make an effort to reach the lecturer/professor and student stakeholder groups. This would be an interesting area for further research, to see whether views of MT in the education sphere align with those in the corporate world.

The age demographics of respondents were evenly distributed, with roughly half of respondents over the age of 40 and half 39 or younger. Responses did not vary significantly between age groups.

In response to the question, “What is your level of interest in roles related to MT,” translators who are employed by an LSP showed significantly more interest than freelance translators. The LSP translators were less likely to feel forced into these roles (10% compared to 31%) and more likely to express interest in learning about these roles (35% compared to 15%). Fewer LSP translators expressed having no interest in learning about or working with MT (11% compared to 19%). Overall, 70% of LSP translators gave positive answers (either interested in learning more or already doing it and enjoying it), compared to only 38.5% of freelance translators.

LSP translators thought human parity (as defined by the question) was more imminent than freelance translators did. For example, 55% of LSP translators thought MT has reached parity or would reach it in the next five years, compared to only 38% of freelance translators.

On the whole, managers/directors were more likely than translators to believe in the imminence of human parity, with 50% thinking MT was already as good as typical human translators or would be in the next five years. Meanwhile, 41% of translators shared this opinion. However, as shown above, LSP translators were actually more likely than their managers to believe parity would be reached now or in the next five years (55%).

Translators versus managers: level of interest

Overall, more than half of all translators (52%) expressed that they are either interested in learning more about MT and gaining new skills or that they already work in MT and enjoy it. A handful of translators expressed mixed feelings or indifference (“My interest in MT is purely as a tool to help me do my work more quickly,” responded one person. “It’s a bit existentially troublesome,” commented another). And 41.3% of translators responded either that they had no interest or that they only worked with MT because they felt they had no choice.

In comparison, 60% of managers responded that they either already work with MT and enjoy it or are interested in learning more, while 20% responded that they only work with MT because they feel they have to. None responded, “I have no interest in learning about or working with MT.” While the managers seemed less optimistic and enthusiastic than all translators together, their enthusiasm was surpassed by translators employed by an LSP (70% of whom responded positively to this question).

Qualitative results

The majority of the comments received in response to the open-ended survey questions show that, in general, the respondents’ understanding of MT is aligned with what industry experts are saying. The majority agree that MT is already quite good for some types of content but that it is not going to fully replace humans — that it is a tool for humans to use to boost their productivity and unlock new possibilities.

In the article Translation Economics of the 2020s published in MultiLingual magazine, the founder and director of TAUS, Jaap van der Meer, argued that LSPs need to completely revolutionize their business model if they don’t want to become obsolete. He also wrote that by 2030 we will have reached the singularity, at which point human translators will not be needed in the translation process. In response to this claim, Dr. Alan Melby, who holds a PhD in computational linguistics, and Dr. Christopher Kurtz, head of translation management at ENERCON, argued that there is no use worrying about what our industry would look like under the singularity or planning for such an event. Under the singularity, AI would be able to perform the tasks of every professional in our society, which would be literally unimaginably disruptive. They argue that, until we reach such a point, “There is definitely a future for professional human translators,” and that this is what we should focus on.

In its 2022 Market Guide for AI-Enabled Translation Services, Gartner predicts that “the use of post-editing of machine translated content will increase. As this occurs, the linguist’s role will shift toward editing machine-translated sections of documents that have been flagged as ‘needing attention,’ and toward more complex translation work.”
These expert opinions are substantially similar to those shared by survey respondents in the comments.


Based on the results of the survey, if we were to rank the three main stakeholder groups by level of optimism and enthusiasm about MT, LSP translators would come first, followed closely by managers, and freelance translators last.

Interestingly, the groups that believe MT is closer to human parity are also the groups who are most enthusiastic and optimistic about working with MT and learning more about it. Rather than despairing that MT is going to steal their jobs, they seem excited for the new human-in-the-loop paradigm. Meanwhile, the group that has the least faith in MT’s usefulness is the least interested in working with it and the most resentful of clients who force them to do so.

More research is required to determine why freelance translators are less keen on MT than LSP translators, but I have some hypotheses.

The first is that freelance translators may prioritize quality above other factors like quantity, time and cost, seeing themselves as individual artisans/creators. Meanwhile, translators who work for an LSP may be more inclined to consider other factors like business goals and may have more of a team-player approach, taking less individual pride over the quality of their work. They may naturally have more affinity with the human-in-the-loop model, which is predicated on the idea that translations need to be fit for purpose, not perfect, and which positions the human as an expert collaborator within a complex system.

Another possibility is that many translators do not have a problem with MT technology itself, but with its misuses and abuses. Several translators who responded to the survey were concerned about LSPs and direct clients taking advantage of MT as an excuse to devalue their work and cut their rates. Some complained about clients not paying as much as they used to, imposing tighter deadlines than is reasonable, not sharing clear quality expectations for MTPE, imposing MTPE for all projects indiscriminately, and treating MTPE the same as revision. Freelance translators are more vulnerable to these abuses than those who work for an LSP. LSP translators get paid the same regardless of how long a task takes, have more direct lines of communication with the decision makers in their organization, and have more job security. A freelance translator may worry that if they make any fuss the LSP will send their projects to a more compliant freelancer next time, whereas an LSP translator does not need to worry that they will be fired if they politely question whether the expectations for a given MTPE project are reasonable.

If this later hypothesis is the case, then LSPs would benefit greatly from communicating more transparently about MT with their freelancers and establishing more open dialogue. The majority of LSP owners do not intend to exploit freelancers and do in fact value their expertise: “The human post-editors and revisers remain absolutely necessary […]. We now only work with a core team of the most talented revisers,” commented one manager. “Machine translation is an excellent tool that helps translators improve productivity, consistency and managing tight deadlines much like robots now help surgeons perform surgeries. The brains behind the operation is still the surgeon though,” wrote another.

In a context of major change and disruptions, leaders need to foster communication, trust and engagement for their organizations to be able to move forward (Donald, 2019). To improve trust with translators, LSPs need to clearly communicate their expectations for MTPE, show that they are eager to receive freelancers’ feedback on MT projects, and be open to exploring new payment models where all stakeholders feel that compensation is fair. They need to stop considering MT a “taboo topic,” as one respondent put it. More open, two-way communication around MT is the solution for reconciling stakeholders’ disparate perspectives, adopting new paradigms, and seizing the potential of MT for the benefit of all stakeholders

Riteba McCallum is a localization and communications professional and director of marketing and communications at OXO Innovation.



The Ultimate Top 3 of MT

By MultiLingual Staff

With the democratization of AI, an engineer can build an MT model in one day. However, beating Google or DeepL in quality, performance, and market…

→ Continue Reading


Subscribe to stay updated between magazine issues.

MultiLingual Media LLC