B2 · Sector Vocabulary · 10 minAI at work in English, taught brick by brick.
A structured ten-slide lesson for B2 learners. Hear a sentence from a typical status update, learn from a tutor why "KI" is "AI" in English and mostly takes no article, meet twelve fixed word blocks sorted by work situation, then practise them in interactive drills before a closing quiz. The whole arc takes about ten minutes and nothing requires a signup.
- CEFR levelB2 · Upper intermediate
- Time to completeAbout 10 minutes
- SkillsVocabulary, Listening, Writing, Speaking
- Bricks10 blocks
01 / 06
"KI" is always "AI" in English — and usually takes no article
The first block is the most important one, because it turns up in every second email. In German you write "KI" and say "die KI". In English the technology is "AI", in full "artificial intelligence", and an abbreviation "KI" does not exist there; put it into an English email and the reader simply will not follow. What matters more is the grammar behind it. "AI" behaves like an uncountable noun, much like "software" or "information": when you mean the technology itself, no article goes in front. "AI helps us draft emails", "we use AI to summarise calls", "AI has changed the way we write reports". The German "die KI hilft uns" tempts you into "the AI helps us", and that is exactly what sounds off to an English ear as soon as the subject is the technology in general. The moment you mean a specific product, on the other hand, you need a countable word after it: "an AI tool", "an AI assistant", "the new AI feature in our CRM". In casual speech you will hear "the AI" for one particular system, but in a status update or a client email you are better off naming the thing itself: "the tool", "the assistant". That single distinction carries you through the whole lesson, and it is the reason the listening slide says "an AI tool" rather than "the AI".
02 / 06
The tools: generative AI, large language model, prompt, training data, output
The second group describes what the tool is and how it works. "Generative AI" is the umbrella term for systems that produce new content — text, images, code — rather than merely sorting existing content. The adjective sits fixed in front; nobody at work says "generative" on its own. Behind it there is usually "a large language model", a model that has learned from large amounts of text what language looks like. That word is countable: "the model", "a newer model", "which model are we using". What you put in is "a prompt" — an instruction or a question. English turns it into a verb without any detour: "to prompt the model", "I prompted it in German and got an English reply". "Training data" is the material the model learned from, and "data" behaves like a mass noun in business English: "the training data is old", "we need more data", never "a data". What comes out is "output" — never plural, no article in general statements, but "the output" once you mean one specific result: "check the output before you send it", "the output was surprisingly good". If you can place these blocks with confidence, you can explain in a meeting what the tool does without dropping back into German once — and at B2 that is the difference between joining the conversation and merely following it.
03 / 06
Putting AI to work: automate, streamline, roll out, draft, summarise
The third group is the one you work with in a status update. "To automate" means to make a process run without a person doing each step, and the noun is "automation" — not "automatisation", even though the German word points straight in that direction. "We have automated the invoice matching", "the automation saves the team an afternoon a week". "To streamline a workflow" means to slim a process down by taking steps out: "we streamlined the approval workflow from five steps to two". In English a new tool is usually not just introduced but rolled out: "to roll out a tool", and the noun is "the rollout" — "the rollout starts with the sales team". The verb is separable, so a pronoun moves into the middle: "we rolled it out in June". What the tool then actually does you describe with two verbs that English has made out of nouns: "to draft" means to write a first version, "to summarise" means to condense. "The tool drafts the reply, I edit it", "it summarises the call in three lines". If you want to say that a process is supported by AI, reach for "AI-powered" or "AI-assisted": "an AI-assisted first draft", "an AI-powered search". And when the team needs new skills for all of this, that is "to upskill the team"; when it is a whole new role, "to reskill".
04 / 06
Risks and checks: hallucinate, bias, human in the loop, sign off on, data protection
The fourth group is the one you need the moment somebody asks whether all of this is safe. "To hallucinate" means the model invents something and presents it as fact — a source, a figure, a contract clause that never existed. The noun is "a hallucination", and it is countable: "we found two hallucinations in the summary". "Bias" names a systematic slant in the output, for instance when certain groups or phrasings are regularly favoured; the word usually stands without an article: "the output can carry bias". The answer to both is "human oversight", and the fixed phrase for it is "a human in the loop": a person checks before anything goes out. That checking has its own verb, and it is a three-part block: "to sign off on something" — "a manager signs off on every assessment", "who signed off on this?". The last block is the one German companies need most often. "Datenschutz" is "data protection"; when the point of view is the individual whose data it is, "data privacy" is the more usual phrase. The DSGVO is "the GDPR" in English. The typical meeting sentence is "customer data must not go into the prompt" — "data" as a mass noun again, no article.
05 / 06
German traps: KI, the AI, automatisation, data are, KI-gestützt
Almost every error in this word field comes straight out of German, and you can tick them off like a list. First, "KI" itself: in an English text it is "AI", including headings and slide titles. Second, the article: "die KI" becomes "AI" with no article for as long as you mean the technology. Third, "Automatisierung": the English word is "automation"; "automatisation" is not a form in common use. Fourth, the data: in German "Daten" is a plural, which is why many learners write "the data are ready" or even "the datas". In business English "the data is ready" is the usual form; "data are" survives mainly in scientific and very formal writing, and "datas" does not exist. Fifth, "KI-gestützt": the usual adjectives are "AI-powered" and "AI-assisted", hyphenated and with no article before "AI". Sixth, the verb question: German often needs a helper construction, such as "einen Prompt eingeben" or "als Problem markieren", whereas English simply uses the noun as a verb: "to prompt", "to flag", "to draft". Use those verbs and you immediately sound less translated. And finally the verb behind the oversight: not "control" but "check", "review" or "sign off on" — in English "control" leans towards steering rather than checking, so "we control every output" sounds as if you were operating the tool by remote.
06 / 06
Four situations: status update, team meeting, client email, works council
Finally, the four situations in which this vocabulary actually comes up for you. In a status update you describe progress in short blocks: "We rolled out the tool to the support team last week. It drafts the first reply; an agent edits and signs off on it. Two hallucinations so far, both caught." In a team meeting the subject is process, so "automate", "streamline" and "workflow": "Can we automate the routing and streamline the handover?" In a client email restraint counts: you promise no magic, you name the check explicitly — "every AI-assisted draft is reviewed by a member of the team before it reaches you". And in a briefing for the works council you need the governance words, clean and without abbreviations: "human oversight", "data protection", "no personal data in the prompt", "we will upskill the team before the rollout". In all four the same principle holds: AI without an article for the technology, a countable word for the product, "data" in the singular. If the everyday office words are what you lack alongside this, office-vocabulary (A2) is the foundation, and for the verbs of this lesson — roll out, sign off on — phrasal-verbs-office (B1) is the right next step.