{"course":{"slug":"ai-at-work","title":"AI at Work","url":"https://addaly.com/learn/ai-at-work","level":"beginner","track":"work"},"lesson":{"slug":"your-own-documents","title":"Working with your own documents","url":"https://addaly.com/learn/ai-at-work/your-own-documents","position":28,"of":73,"minutes":9,"answer":"Attaching your documents puts the truth in the text, but demand quoted citations, because general knowledge blends in invisibly.","keyIdea":"Attaching your documents puts the truth in the text, but demand quoted citations, because general knowledge blends in invisibly."},"body":{"format":"markdown","text":"Everything so far assumed you paste text into a box. The bigger shift is pointing AI at material *you* own: the policy manual, the last three years of board minutes, a folder of supplier contracts, the 200-page standard your industry runs on.\n\nThis is where AI stops being a clever writing toy and becomes useful for the specific job you actually have. It is also where a new failure mode arrives.\n\n## Why your documents change everything\n\nRemember the sorting rule from lesson one: the model is strong when truth lives in the text you gave it. Uploading your documents *puts the truth in the text.* A question the model would have guessed at (\"how many days of bereavement leave do we give?\") becomes a question it can answer by reading.\n\nThree ways this shows up in tools you already have:\n\n- **Attach a file to a chat.** Simplest. Works for one to a few documents. The model reads the whole thing.\n- **A project or knowledge space.** Most assistants now let you keep a set of documents attached across many conversations, so you stop re-uploading the HR manual every morning.\n- **Search over a large collection** (often called retrieval or RAG). For hundreds or thousands of documents, the tool finds the relevant passages first, then answers from those. Important consequence: it only sees what the search found. If the search misses, the answer is built on an incomplete picture and will not tell you so.\n\n## The failure mode: confident blending\n\nWhen you supply documents, the model does not cleanly separate \"what your document says\" from \"what I generally know about this topic.\" Ask about notice periods with your handbook attached, and you can get an answer that is 80% your handbook and 20% standard employment practice, seamlessly merged, in one paragraph, with no seam visible.\n\nThis is worse than an obvious error because it is mostly right. The fix is structural, not hopeful:\n\n> Answer only from the attached documents. For every claim, give the document name and the quoted sentence it comes from. If the documents do not answer the question, say \"not covered in these documents\"; do not fill the gap from general knowledge.\n\nThen spot-check two or three citations. Citations can be wrong. A quoted sentence that does not exist in the file is the single fastest way to catch a bad answer, and it takes one search to find out.\n\n## What this is good for\n\n- **Finding, not reading.** \"Which of these 40 supplier contracts have a price-escalation clause?\" You then read those four.\n- **Consistency checks.** \"Does the new procedure contradict anything in the existing manual?\" Humans are poor at this across long documents. Machines are decent.\n- **Answering the same question for the tenth time.** Shopkeepers with a returns policy, teachers with an exam board specification, civil servants with a procurement rulebook: the question comes in weekly and the answer is in a document nobody wants to reopen.\n- **Onboarding yourself.** New to a case, a client, a ward, a portfolio? \"Here are the last eight months of notes. What do I need to know before Monday?\"\n\n## Before you upload anything\n\nStop. Whose documents are these, and what did you promise about them? A client contract, a patient file, a personnel record, anything under an NDA: the answer may be that you cannot upload it at all, regardless of how useful it would be. That is lesson six, and it is the one that ends careers. Read it before you upload your first real file."},"check":{"question":"A council officer uploads her department's 90-page procurement manual and asks whether a 40,000-peso purchase needs three quotes. She gets a clear, confident yes with a plausible explanation. What is the most important thing to check?","options":[{"text":"Whether the answer quotes an actual passage from her manual, since general procurement norms can blend into the answer unmarked","correct":true},{"text":"Whether the model has been trained on public-sector procurement rules for her country","correct":false,"tempting":"Focuses on the model's background knowledge when the whole point of uploading was to make background knowledge irrelevant."},{"text":"Whether 90 pages exceeds what the tool can read in one go and part of the manual was skipped","correct":false,"tempting":"A real constraint in some tools, but a truncated read still produces a checkable citation: verifying the quote catches both problems at once."},{"text":"Whether she phrased the threshold question precisely enough, since ambiguous wording produces confident wrong answers","correct":false,"tempting":"Prompt precision helps, but a perfectly phrased question can still be answered from general practice rather than from the uploaded manual."}]},"prev":{"slug":"talking-to-it","title":"Dictation, voice mode and who pays for the errors","url":"https://addaly.com/learn/ai-at-work/talking-to-it"},"next":{"slug":"how-retrieval-actually-works","title":"What \"chat with your documents\" is really doing","url":"https://addaly.com/learn/ai-at-work/how-retrieval-actually-works"},"authorship":"Founder and series editor: Dr Mustafa Mun. Written by AI models to a written standard. Exam keys checked blind by a second model: 3,665 of 3,666 agreed. Academic review invited.","review":{"status":"Draft, open for academic review"},"cite":"Addaly (2026) 'Working with your own documents', in AI at Work. Edited by M. Mun. Addaly. Available at: https://addaly.com/learn/ai-at-work/your-own-documents (Accessed: 31 October 2026).","licence":{"name":"CC BY-NC-SA 4.0","url":"https://creativecommons.org/licenses/by-nc-sa/4.0/"},"reader":{"schema":1,"title":"Working with your own documents","courseTitle":"AI at Work","module":{"index":4,"slug":"working-from-your-own-material","title":"Working from your own material"},"answer":null,"keyIdea":"Attaching your documents puts the truth in the text, but demand quoted citations, because general knowledge blends in invisibly.","hero":{"kind":"plate","image":null,"alt":null,"caption":null},"blocks":[{"t":"p","runs":[{"text":"Everything so far assumed you paste text into a box. The bigger shift is pointing AI at material "},{"text":"you","i":true},{"text":" own: the policy manual, the last three years of board minutes, a folder of supplier contracts, the 200-page standard your industry runs on."}],"lede":true},{"t":"p","runs":[{"text":"This is where AI stops being a clever writing toy and becomes useful for the specific job you actually have. It is also where a new failure mode arrives."}]},{"t":"h2","id":"why-your-documents-change-everything","runs":[{"text":"Why your documents change everything"}]},{"t":"p","runs":[{"text":"Remember the sorting rule from lesson one: the model is strong when truth lives in the text you gave it. Uploading your documents "},{"text":"puts the truth in the text.","i":true},{"text":" A question the model would have guessed at (“how many days of bereavement leave do we give?”) becomes a question it can answer by reading."}]},{"t":"p","runs":[{"text":"Three ways this shows up in tools you already have:"}]},{"t":"ul","items":[[{"text":"Attach a file to a chat.","b":true},{"text":" Simplest. Works for one to a few documents. The model reads the whole thing."}],[{"text":"A project or knowledge space.","b":true},{"text":" Most assistants now let you keep a set of documents attached across many conversations, so you stop re-uploading the HR manual every morning."}],[{"text":"Search over a large collection","b":true},{"text":" (often called retrieval or RAG). For hundreds or thousands of documents, the tool finds the relevant passages first, then answers from those. Important consequence: it only sees what the search found. If the search misses, the answer is built on an incomplete picture and will not tell you so."}]]},{"t":"h2","id":"the-failure-mode-confident-blending","runs":[{"text":"The failure mode: confident blending"}]},{"t":"p","runs":[{"text":"When you supply documents, the model does not cleanly separate “what your document says” from “what I generally know about this topic.” Ask about notice periods with your handbook attached, and you can get an answer that is 80% your handbook and 20% standard employment practice, seamlessly merged, in one paragraph, with no seam visible."}]},{"t":"p","runs":[{"text":"This is worse than an obvious error because it is mostly right. The fix is structural, not hopeful:"}]},{"t":"quote","runs":[{"text":"Answer only from the attached documents. For every claim, give the document name and the quoted sentence it comes from. If the documents do not answer the question, say “not covered in these documents”; do not fill the gap from general knowledge."}]},{"t":"p","runs":[{"text":"Then spot-check two or three citations. Citations can be wrong. A quoted sentence that does not exist in the file is the single fastest way to catch a bad answer, and it takes one search to find out."}]},{"t":"h2","id":"what-this-is-good-for","runs":[{"text":"What this is good for"}]},{"t":"ul","items":[[{"text":"Finding, not reading.","b":true},{"text":" “Which of these 40 supplier contracts have a price-escalation clause?” You then read those four."}],[{"text":"Consistency checks.","b":true},{"text":" “Does the new procedure contradict anything in the existing manual?” Humans are poor at this across long documents. Machines are decent."}],[{"text":"Answering the same question for the tenth time.","b":true},{"text":" Shopkeepers with a returns policy, teachers with an exam board specification, civil servants with a procurement rulebook: the question comes in weekly and the answer is in a document nobody wants to reopen."}],[{"text":"Onboarding yourself.","b":true},{"text":" New to a case, a client, a ward, a portfolio? “Here are the last eight months of notes. What do I need to know before Monday?”"}]]},{"t":"h2","id":"before-you-upload-anything","runs":[{"text":"Before you upload anything"}]},{"t":"p","runs":[{"text":"Stop. Whose documents are these, and what did you promise about them? A client contract, a patient file, a personnel record, anything under an NDA: the answer may be that you cannot upload it at all, regardless of how useful it would be. That is lesson six, and it is the one that ends careers. Read it before you upload your first real file."}]}],"check":{"question":[{"text":"A council officer uploads her department’s 90-page procurement manual and asks whether a 40,000-peso purchase needs three quotes. She gets a clear, confident yes with a plausible explanation. What is the most important thing to check?"}],"options":[{"text":[{"text":"Whether the answer quotes an actual passage from her manual, since general procurement norms can blend into the answer unmarked"}],"correct":true,"tempting":null},{"text":[{"text":"Whether the model has been trained on public-sector procurement rules for her country"}],"correct":false,"tempting":[{"text":"Focuses on the model’s background knowledge when the whole point of uploading was to make background knowledge irrelevant."}]},{"text":[{"text":"Whether 90 pages exceeds what the tool can read in one go and part of the manual was skipped"}],"correct":false,"tempting":[{"text":"A real constraint in some tools, but a truncated read still produces a checkable citation: verifying the quote catches both problems at once."}]},{"text":[{"text":"Whether she phrased the threshold question precisely enough, since ambiguous wording produces confident wrong answers"}],"correct":false,"tempting":[{"text":"Prompt precision helps, but a perfectly phrased question can still be answered from general practice rather than from the uploaded manual."}]}],"order":[2,0,3,1],"why":null},"prev":{"slug":"talking-to-it","title":"Dictation, voice mode and who pays for the errors","minutes":8},"next":{"slug":"how-retrieval-actually-works","title":"What “chat with your documents” is really doing","minutes":10}}}