100×worker · glossary

AI and Work Glossary

This glossary collects the terms you run into when AI starts doing real work in your job, your inbox, and your team's daily tasks. Each definition stands on its own, so you can look up a word without wading through jargon or sales talk.

AI agent

A system that uses AI to complete tasks on its own, taking steps without constant human input.

An AI agent can read an email, draft a reply, and send it after checking your calendar, all without you clicking through each step. You still set the rules and check the outcome, but the agent handles the sequence of actions itself.

prompt

The instruction you give an AI system to tell it what output you want.

A prompt can be a single sentence, like 'summarize this contract in three bullet points', or a longer set of instructions with examples. The quality of the prompt shapes the quality of the answer.

large language model (LLM)

An AI model trained on huge amounts of text to predict and generate humanlike language.

ChatGPT, Copilot, and Claude are all built on large language models. When you type a question, the model predicts the most likely next words based on patterns it learned from text, not by looking up facts in a database.

context window

The amount of text an AI model can read and remember at one time.

If the context window is small, the model forgets the start of a long document by the time it reaches the end. A bigger context window lets you paste a full contract or a whole email thread and get answers based on all of it.

context engineering

Structuring the information you give an AI so it produces accurate, useful output.

Instead of just asking a question, context engineering means adding the right background: a template, past examples, house style, and the goal of the task. A sales team that feeds an AI its best proposals first gets better draft proposals back.

hallucination

Confident AI output that is false, invented, or not supported by any real source.

An AI might state a law that doesn't exist or cite a report that was never published, phrased as if it were certain. This is why any AI-generated fact used in client work needs a human check before it goes out.

harvest map

The resulting map of a job's tasks, sorted into eliminate, automate, delegate, or keep.

You list every task in a role, then sort each one into one of four buckets. The finished harvest map shows at a glance which work should stop, which should run through AI, which should go to someone else, and which stays with you.

eliminate / automate / delegate / keep

The four buckets used to sort any job's tasks by what should happen to them.

Eliminate covers tasks that add no value and can simply stop. Automate covers tasks an AI system can run without a person. Delegate covers tasks that go to someone else, human or AI. Keep covers the tasks that stay with you because they need your judgment.

skill file

A written set of instructions that teaches an AI how to do one specific task.

A skill file for writing client proposals might include the company's tone of voice, a pricing table, and three example proposals. Once saved, any team member can point an AI at that skill file and get a consistent draft.

company brain

A shared collection of a company's documents, decisions, and knowledge that AI can search.

Instead of an AI answering from general internet knowledge, it answers from your contracts, past emails, and internal wikis. A new hire asking about the return policy gets the actual company answer, not a generic guess.

human in the loop

A person who reviews, approves, or corrects AI output before it is used.

An HR team might let AI draft rejection letters, but a human in the loop reads each one before it's sent. This catches errors and keeps someone accountable for the final decision.

AI-first

A way of working where AI produces the first draft, and a person edits it.

In an AI-first workflow, a marketing team asks AI to write the first version of a newsletter, then edits for accuracy and tone. The starting point shifts from a blank page to a draft you refine.

automation score

A per-job estimate of how much of the work could be done by machines or AI.

Older automation scores, like the Frey & Osborne estimate from 2013, were built before large language models existed. Microsoft's 2025 AI applicability score, based on 200,000 real Copilot conversations, found translators at 49% and nurses at 12%, a very different ranking.

generative AI

AI that creates new text, images, or other content, instead of only analyzing existing data.

Generative AI writes an email draft, designs a slide, or drafts code from a description. Adoption is already widespread: 43% of Flemish adults use it monthly, and 32.7% of the EU population aged 16 to 74 used it in 2025.

the 100x worker

The person on a team who runs AI systems and agents on behalf of the group.

A 100x worker doesn't do every task by hand. They set up the skill files, manage the agents, and check the output, so the team's harvest map actually gets automated and delegated, not just planned on paper.

Drafted with AI assistance and editorially reviewed. See the terms in action in the job analyses.