Not long ago, discussions about AI in translation sounded like a clash between two camps: some saw it as the profession’s killer, while others saw it as a savior from routine work. Earlier, I wrote that the impact of AI on the industry is still impossible to judge unequivocally, but one thing is clear: the technology will be a major part of the future. Those who survive (and grow) will be the ones who can strike the right balance between AI and human expertise — understanding not only the benefits, but also the limitations.

This article is a continuation. No abstract slogans and no panic. I want to talk about the practical side: which “new companions of the translator” have already settled on the desktop, why among them there are both real friends and “false” ones, and how to set up your workflow so that technology increases quality, not just speed.

The profession hasn’t disappeared — the role has changed

AI has dramatically reduced the cost of a “first draft”. That’s a fact. But clients are still not buying a set of sentences in another language. They are buying a result: accuracy, style, accountability, and predictable quality.

Only recently, a translator was often “someone who translates”. Today, a translator is increasingly someone who:

• sets the task (what exactly needs to be done and in what style),
• creates or receives a draft (including with the help of AI),
• checks and edits it to a level they can confidently vouch for,
• builds a system (terminology, QA, unified rules),

AI does not cancel the profession. It shifts the center of gravity from “typing the first version” to “controlling meaning and quality.”

A Tool Belt: Expanding the Translator’s Capabilities

I like to talk about tools plainly and honestly. Today my work rests on three “friends”: ChatGPT, PyCharm, and my own applications. Each has a strong side. And each can turn into a false friend if you use it without a system.

1) ChatGPT — a fast drafting partner (and a source of confident mistakes)

For me, ChatGPT is a thought accelerator. It doesn’t “translate instead of me”. It helps me get through the first mile faster — the part that usually consumes the most energy: making sense of a messy source text, generating options, finding a natural phrasing, building the structure of a letter or instructions, and adjusting tone.

Typical tasks where it helps:

• offer several variants of the same phrase (neutral / formal / conversational),
• rephrase “bureaucratic Russian” into natural English (or vice versa),
• extract key requirements from a brief or SOW,
• compile a rough glossary from the text and propose consistent terms,
• help with a short summary, an email to the client, or translator’s notes.

But at the same time, ChatGPT is a classic “false friend”. The danger isn’t that it makes mistakes (everyone does). The danger is that it makes mistakes confidently.

Here is where it most often lets you down:

• it “fills in” facts that are not in the source,
• it smooths and simplifies meaning (which is critical in law, engineering, and medicine),
• it invents “plausible-sounding” terms,
• it can preserve style while losing precision.

That’s why I treat it like a very fast intern: helpful, but in need of verification. It saves time, but it doesn’t remove responsibility.

The rule that saves me:
AI produces an option. I make the decision and take responsibility for the meaning.

2) PyCharm — be your own developer

Translators often think that programming is “not for us”: we weren’t taught it, and there’s no time for it. That used to be true, but with the arrival of AI a lot has changed: with a properly defined task, it becomes a guide and a personal assistant in the world of code — it helps sketch a script, explains errors, suggests solution options, and shortens the path from an idea to a working button.

Of course, there are pitfalls: AI can skip code fragments, mix up dependencies, or be confidently wrong, so basic knowledge and the habit of checking the result are a big plus. But you no longer need to spend years to become a professional programmer. Even today you can use AI to build small apps for your own tasks and turn routine work into an automated process.

3) Your own apps — narrow tools beat universal ones

At some point I realized: universal services are good, but they rarely solve a translator’s specific task perfectly. So I built my own applications: for alignment and translation-memory building, for working with spreadsheets, for preparing materials, and for automating recurring scenarios.

The point is not to “reinvent the wheel”. The point is to build a tool that is as “tailored to your own process” as possible.

And in my view, this is one of the key turns of the profession in the AI era: a translator becomes not only a practitioner, but also the architect of their own workflow pipeline.

By the way, you can find examples of my own applications in the Tools section of this website.

What’s next: work, food for thought, and growth

We really don’t yet know what the translation market will look like five years from now. But the main thing is already visible: technology is becoming a mandatory part of the profession. And the question is not whether AI will come. The question is whether the translator will have a system — tools, checks, discipline, and a willingness to learn.

AI can be both a helper and a “false friend”. It becomes a true friend when a translator stops limiting themselves, accepts the new reality as a given, and keeps developing — not out of fear, but out of curiosity.

Because in a world where first drafts have become cheaper, what is especially valued is what has always been rare: accuracy, accountability, a feel for language, and the ability to think.