7 Ways to Use AI When English Isn't Your First Language

Using AI when English isn't your first language gets sold as one thing: a translate button. That's the smallest version of it. The larger version is that AI can take over the parts of English that cost you time and confidence without touching the part that makes you worth reading, which is your thinking.

Here's what the generic listicles miss. English at work isn't one skill, it's four: producing it, understanding it, sounding right in it, and being heard in it. Most advice fixes only the first, and it fixes it by swapping your sentences for correct, bland, forgettable ones. You're also nowhere near alone in this. Ethnologue counts roughly 1.5 billion English speakers worldwide, and about 1.1 billion of them learned it as a second language. Close to three in four people writing English at work today are doing it in their second, third, or fourth language. Nobody's English is the default English. Certainly not the model's.

Every idea below is new to this column, and each one ends with something you can do this afternoon.

1. Ask for a Diff, Not a Rewrite

Marek estimates commercial flooring jobs outside Chicago. For a year he ran every bid through an AI assistant with three words, "fix my English," and sent back whatever came out. The bids were spotless. They also read like a bank statement, and a longtime client eventually asked him why his emails had gotten cold.

That's what "fix my English" buys you. The model rewrites, because rewriting is easier for it than surgery, and the rewrite is always its house style rather than yours. So ask for the other thing. Tell it to correct only what's grammatically wrong, keep your sentence structure and word choices intact, and then list every change with a one-line reason.

Two things happen. Your writing survives with your fingerprints on it. And the list teaches you the pattern behind your own mistakes, which is almost always a small handful of things, articles or prepositions or tense agreement, repeating forever. After a month of reading those lists you stop making half of them.

Start today: On your next English message, prompt: "Correct only what is grammatically wrong. Keep my sentence structure and my word choices. Then list each change you made with a one-sentence reason."

2. Decode What English Speakers Mean, Not Just What They Wrote

Everyone treats this as an output problem. The harder half is input. "Let's take this offline." "I have some concerns." "Interesting." Every word in those sentences is easy, which is exactly why no dictionary helps. The meaning lives in what the speaker chose not to say.

AI reads that layer well. Paste the message and ask three things: what is this person asking me to do in plain terms, how serious is it on a scale of one to five, and what are they implying that they didn't write. The seriousness rating is the part people underestimate. English at work runs on softened requests, and reading a five as a two is how deadlines slip.

Start today: Take the last message you weren't sure how to read and ask: "What is this person asking me to do, how urgent is it on a scale of 1 to 5, and what are they implying that they didn't say directly?"

3. Build a Glossary of Your Workplace's English From Your Own Inbox

Every company runs on a private dialect. Acronyms nobody expands, product nicknames, a verb used in a way no dictionary lists, three words that mean the same thing depending on which team says them. Native speakers absorb this by sitting near people for a year, then forget they ever had to learn it, which is why nobody thinks to hand you the list.

You can extract it in an afternoon. Take thirty or fifty recent internal messages, tickets, or docs, and ask AI to pull out every acronym, internal name, and industry phrase, define each one from how it's used in that context, and flag any whose meaning differs from ordinary English. That last flag is the valuable column. Those are the words that quietly mislead you.

The judgment stays yours. Strip anything confidential before you paste, and keep the glossary somewhere you can search in five seconds.

Start today: Copy your last thirty internal messages into an AI assistant and ask: "List every acronym, internal product name, and industry phrase here. Define each from how it's used, and mark any that mean something different from ordinary English."

4. Draft Recurring Documents From the English You Already Got Right

You have a corpus and you've never used it. The status report your director called clear. The proposal that won. The support reply that never got escalated. That's a body of English proven to work in your workplace, written by you, and it's sitting in a sent folder doing nothing.

Most people start each week's version from a blank page plus an AI's factory settings, which is why the output keeps sounding like a stranger. Reverse the inputs. The approved past versions become the model of how you write; this week's facts become the only new material. Handled once as a document workflow that drafts from your own past versions, the recurring documents stop being a weekly language exercise and start arriving in the English you already own.

Start today: Find three past documents of the same type that landed well. That set is the pattern. Then hand the pattern off once instead of rebuilding it every Friday.

5. Get Pronunciation Coaching on the Twenty Words You Say Every Day

Language apps drill generic vocabulary. You don't need generic. You need your company's name, your job title, the product, the five terms that come up in every standup, and the client surname you've been avoiding out loud for six months.

Ask AI to build that list for you: a plain respelling rather than only phonetic symbols, the syllable that carries the stress, and which sound a speaker of your first language most often gets misheard on. Stress matters more than people expect. English listeners parse rhythm before they parse individual sounds, and a word with the beat in the wrong place lands as a different word entirely. For anything you're unsure of, Cambridge Dictionary has recorded audio in both US and UK versions.

The practice itself has no AI in it at all, and that's worth saying plainly. You say the word out loud twenty times in your kitchen. What AI removes is the guesswork about which twenty words were worth the effort.

Start today: Write down the twenty words you say most at work. Then ask: "For each of these, give me a simple respelling, mark the stressed syllable, and tell me which sound a [your first language] speaker most often gets misheard on."

6. Put the Work You Already Wrote in Your First Language in Front of a New Audience

Duyen spent six years writing sharp supply chain analyses in Vietnamese. Her English-speaking industry has read none of it. She was never short on ideas; she was short on the weekend it took to move each piece into English, so the archive stayed where it was.

This is the reach item, and it's the one that changes what people think you're capable of. You likely have the same archive: posts, internal analyses, conference talks, teaching notes. Moving it across takes more than swapping words. References and examples have to change too, because a comparison that lands instantly in Hanoi may mean nothing in Ohio, while your argument and structure stay exactly as you built them. A content workflow that repurposes what you already wrote makes the second, third, and tenth piece cost almost nothing.

Guard against flattening. Run the diff rule from idea one over whatever comes back, and rewrite the swapped examples yourself. Your unusual angle is the reason anyone will read it.

Start today: Pick one piece you're proud of. Ask for an English version that keeps your argument and section order, plus a list of which examples won't land for a US or UK reader. Rewrite those examples in your own words.

7. Speak It First, Type It Second

Almost everyone is more fluent out loud than at a keyboard, and the gap is widest in a second language. Typing English forces you to compose and edit in the same motion, so the sentence gets shorter and safer and less like what you meant. Talking doesn't do that.

Dictation crossed the usability line a while ago. Current speech models handle strong accents far better than the phone dictation of five years ago, and they cope with code-switching, dropping a word from your first language mid-sentence, without collapsing. Google Docs has voice typing built in and free, and most AI assistants now take voice input directly.

Then chain it to idea one. Talk for two minutes in whatever mix of languages comes fastest, hand over the transcript, and ask for the correction list rather than a rewrite. What arrives is your thinking at full speed, in correct English, in your own order.

Start today: Open voice typing or your assistant's microphone and talk for two minutes about the email you've been avoiding. Then: "Turn this into a clear English email. Keep my points and my order. Fix only what's wrong."

Pick One

Seven is too many for one week. Pick the one that named something you've been quietly working around for years. The diff rule costs you nothing and pays off tomorrow. The back catalog is slower and changes what your industry knows you can do.

Sunday, All Things Agentic reports what changed in AI that week. Wednesday, The Loop builds one workflow from end to end. This column hands you the ideas in between, and the agent workflow directory is where an idea you like stops depending on you remembering it.

Frequently Asked Questions

• Will using AI make my English sound like everyone else's?+
  Only if you ask it to rewrite. A rewrite request returns the model's default business English, which is correct and forgettable. Ask instead for corrections plus a list of what changed and why. Your sentence structure survives, and you learn the pattern behind each fix rather than outsourcing it again next week.
• Which AI tool is best when English is your second language?+
  Every major AI assistant handles all seven ideas on this list. What you ask for matters far more than which tool you open. Dedicated language apps drill vocabulary; a general assistant works on your real documents, your real inbox, and the words your job uses. When one of these habits turns into a weekly chore, automating the repetitive version is the next step.
• Is it dishonest to use AI to write in English at work?+
  No. The test is whether the ideas, the judgment, and the commitments are yours. Nobody has ever called a dictionary or a spell checker cheating. It turns dishonest only when you're claiming a language skill you don't have in a setting that depends on it, such as a proficiency exam or a role whose core duty is unaided writing.
• How do I keep up in fast English meetings?+
  Record with everyone's permission, then work from the transcript afterward. Ask for the summary in your first language so comprehension costs you nothing, and ask for action items and quotes kept in the original English so you can reply using the exact words the room used. Turning that into a standing meeting notes workflow means you stop doing it by hand.