Why we’ve changed our approach to AI

As a people-driven enterprise, how do you make the most of what MT and AI have to offer? And how do you do that without devaluing the human experts on whom your business was built?

As a Life Sciences enterprise, how do you communicate the value of using AI to clients who hold people’s health in their hands? How do you offer trusted advice to nervous clients who, despite their best efforts, see only the potential risks?

These are questions we have been wrangling with for a long time; even more urgently since AI and LLMs went mainstream in 2022. Happily, this last year has brought us exponentially closer to answers that align with our values as a company and that provide real-world benefits to all involved.

Our AI evolution

Our approach to MT and AI has evolved a lot in the last 12+ months, since we last publicly took stock. So too has our clients’ openness to the possibilities, as we (and the wider cultural conversation) have offered more context, success stories, and education around the subject.

These evolutions have real-world impacts for our workflows, our linguists, and what we’re now able to do for our clients. So, we thought it was high time we took another look at where we are with MT, shared some exciting news, and ensured our clients were up-to-date and involved in the process.

Note: For more background, and definitions of terms, take a look at our previous insights on MT and LLMs. (We’re not going to go back over old ground here; things are moving too quickly for that! 😊)

And so, with that same urgency in mind, let’s dive right into where we were a year ago; where we are now; what’s changed; and what it all means for our clients.

Where we were in 2023 – A strictly limited scope

Back at the end of 2023, we were deep in AI testing; assessing and incorporating the elements that served our clients best at the time. In our November 2023 round-up, we concluded that “for us – MT can be a useful tool in very specific scenarios with clear parameters”. Those scenarios were primarily gisting and translating content intended for internal reference. As we noted at the time, this is work that “makes up a very small proportion of what we do; most of our projects […] are patient- or authority-facing”. What we didn’t say was that, at the time, our clients were also a major factor in our decision not to go all in on AI. The vast majority of them were not convinced of the value of AI in their work. While it is our job to advise our clients on our areas of expertise, it is not our place to foist technologies on them that they simply don’t want.

What’s changed in 2025 – The scope expands

We’re still testing, assessing, and incorporating elements that serve our clients best. (That part hasn’t changed.) What has changed is what we are able to do with our tested and trained MT just over a year later (a lifetime in the world of AI). The scenarios in which our MT is useful have expanded greatly, far beyond gist and internal translations. The parameters we can set, while no less clearly defined, have become more sophisticated, allowing us to work with more complex, higher-visibility content. At the start of 2025, our MT is in a place where we can process a variety of patient- and authority-facing content with total confidence, meaning the use cases for us have gone from a small proportion of our work to something much bigger.

What this means for our clients – Added value and total transparency

We can now offer MT across a wide range of our clients’ projects, where it will add value. We assess projects and content types on a case-by-case basis and advise with complete transparency as to whether MT is or is not suitable.

Our clients can be as involved or uninvolved as they want in the process. For those clients that are still nervous, we’ve found that getting involved in the detail helps to allay their fears, as they take their first steps into MT/AI with us as their trusted advisor.

Where we were in 2023 – Insignificant savings

Back in the mists of 2023, we found our MT implementation was having a negligible impact on time and cost savings. In our November 2023 round-up, we noted that “it can often take longer to check and fix something that’s not quite there than it does to get it right first time”. And that “working with medical content involves a series of checks and balances that make the benefits MT does bring to the table relatively insignificant”.

Given that our clients were already sceptical about quality, the fact that we couldn’t guarantee immediate time and cost savings meant a wide-ranging AI launch was a non-runner at the time.

What’s changed in 2025 – A considerable contribution

Over the last year, our trained MT has come on leaps and bounds in terms of the first translation produced by the machine. This is due to a combination of factors, including continued training; feedback loops and iteration; and pairing different engines with different content types and language combinations. Not to mention how much AI in general has learned and improved in that time!

In terms of the checks and balances that all but cancelled out savings before, over time, we’ve trained our MT tools to incorporate many of these within custom processes, meaning that all that previously manual work has been automated into the workflow.

What this means for our clients – Custom workflows and significant savings

For our clients, this means more time, cost, and efficiency savings through AI. At the same time, it means knowing that none of this entails a drop in quality or the cutting of procedural corners.

Over the last year, showing our clients the benefits these solutions can bring to their overheads and workflows in real terms has increased their willingness to try these solutions on more and more varied projects.

Where we were in 2023 – Linguist vs Machine

In 2023, we stated that “none of the MT options currently available do as good a job as our human expert linguists”.

What’s changed in 2025 – Symbiosis

That statement was true then and it remains true now. Once more for the people in the back: WE HAVE NO INTENTION OF REPLACING OUR HUMAN EXPERTS.

The expertise, nuance, and accuracy of our linguists is something we will always value above bare efficiencies.

The difference between November 2023 and now is that – through practice, continuous feedback, and collaboration with our linguists – we’ve found a way to combine humans and AI in a way that works in real terms.

The key, we’ve found, is to keep human and machine in a meaningful balance, where the machine does no more and no less than help the human linguist to do their best work.

What this means for our clients – Collaboration and clear advice

With AI as one of many tools in our arsenal, we will continue to produce the same exceptional quality we always have, and work with the very best people in the sector. There will be some cases where we advise against MT, keeping the process purely human. Where MT is implemented, this will always be followed by human expert review.

In summary, where MT works, we will be sure to tell you, and you’ll reap all the benefits that entails. Where it falls short, we will be sure to tell you that too, helping you avoid the pitfalls of indiscriminate adoption.

And now for the most exciting news…

Having tested our MT on a variety of real projects and with a careful selection of clients, we’re going live for all clients and projects as of now!

Unlike those who simply slapped “AI” on the name of an existing service, we didn’t want to launch until we knew we could guarantee more than bells and whistles. We have taken our time to develop a solution that genuinely adds substantial value, and we’ve got receipts – for anyone who may still be nervous.

Currently, our solution can be used for most – not all – languages. It saves time and cost for many – though not every – content type. We’ll assess projects (and even different files within the same project) for suitability and may split projects accordingly. Don’t worry – this will NOT extend timelines or increase complexity – it will only be implemented where it achieves the opposite.

Time for you to try…?

So, if you would like to trial our human-led, actually customized, case-by-case MT, just get in touch as you’ve always done. Maybe you’ll even have new projects you wouldn’t have thought to send before.

If it’s your first foray into AI, you can decide how involved or uninvolved you’d like to be in the process. Let us know and we’ll communicate with you according to your preferences.

And, as we’ve said many times before, our technology solutions are constantly evolving. They will most likely never reach a final, static version. Join us on this journey and we promise to be your most trusted guide through the rapidly changing landscape of Life Sciences, language, and AI.