{"course":{"slug":"ai-for-small-business","title":"AI for a Small Business","url":"https://addaly.com/learn/ai-for-small-business","level":"beginner","track":"work"},"lesson":{"slug":"the-tools-that-hold-your-files","title":"Four ways to keep a small library, and what each costs","url":"https://addaly.com/learn/ai-for-small-business/the-tools-that-hold-your-files","position":23,"of":88,"minutes":9,"answer":"Notebook products, projects in tools you already pay for, features inside software that already holds your data, and a local setup are four places a small document library can live.","keyIdea":"Notebook products, projects in tools you already pay for, features inside software that already holds your data, and a local setup are four places a small document library can live, and the deciding questions are whether the material is confidential and how you would get it out again."},"body":{"format":"markdown","text":"## The choice is about where the files live\n\nThe capability is much the same everywhere. What differs is who ends up holding your documents, how much work it is, and what happens when you want to leave.\n\n### 1. A notebook product\n\nGoogle's NotebookLM is the best-known, and free at the time of writing. You add up to a few dozen sources (documents, PDFs, pasted text, web pages, a YouTube link) and ask questions across them. Answers carry inline citations back to the source passage, which is the feature that matters.\n\nStrong points: nothing to set up, the citations are good, and it is deliberately confined to your sources rather than answering from general knowledge. It also generates summaries and study material from your own documents, which is more useful for training a new employee than it sounds.\n\nWeak points: your documents sit in a large company's cloud, the free tier's limits change, and it is not designed as a permanent business system. Read the terms if the documents are sensitive.\n\n### 2. A project inside the tool you already use\n\nChatGPT projects, Claude projects, Gemini in Google Workspace, Copilot in Microsoft 365. Attach files to a workspace, and every conversation in it starts with those files available.\n\nStrong points: no new vendor, no new subscription if you already pay, and it sits where you are already working. For a business with fewer than twenty reference documents this is usually the right answer.\n\nWeak points: file limits are modest, and how the files are used varies between products: some read everything, some retrieve passages, and the interface rarely tells you which.\n\n### 3. Inside software that already holds the data\n\nYour helpdesk, your CRM, your shop platform. If your help centre software has an AI answer feature, it is already grounded on your articles, and no document leaves a system you have not already trusted.\n\nThis is underrated for the same reason as before. The data is there, the contract exists, and there is no second copy.\n\n### 4. On your own machine\n\nFree and open: **AnythingLLM**, **Open WebUI**, or **LM Studio**, each paired with a local model through **Ollama**. Point them at a folder, and you have a private document assistant that works offline.\n\nStrong points: nothing leaves the building, no per-question cost, no limits, and it settles most of the privacy conversation in one move.\n\nWeak points: an afternoon of setup, answers a step below the best hosted models, and you are your own support desk. For a clinic, a law practice, an accountant or anyone holding other people's confidential material, that afternoon is well spent.\n\n## Choosing\n\nAsk three questions in order.\n\n**Is the material confidential?** If yes, start at option four or option three, and treat one and two as unavailable until you have read the terms properly.\n\n**Do you already pay for something that does this?** Usually yes, and usually nobody has looked.\n\n**How many documents?** Under twenty, a project is fine. Under a hundred with citations mattering, a notebook product. More than that, or growing, you want something built for it.\n\n## The question to ask before you commit\n\nHow do I get my documents and my answers out?\n\nFiles you uploaded should be downloadable. Answers, summaries and any structure you built are often not, and rebuilding a library in a new tool takes an afternoon you had not budgeted. Keep the master copies of every document in a folder you control, and treat whatever the tool holds as a copy. That single habit makes every one of these four reversible.\n\n## What a small library actually looks like\n\nBusinesses that get value here are not running a document warehouse. A typical working set is six to fifteen files:\n\n- The current price list.\n- Terms and conditions, and the refund policy.\n- The two or three procedures people ask about most.\n- A frequently-asked-questions page built from real customer emails.\n- Supplier lead times and delivery areas.\n- The staff handbook, if you have one.\n\nThat is enough to answer most of what anybody asks in a week, and it is small enough that one person can keep it current in twenty minutes a month. Resist the urge to add the archive. A library grows in usefulness up to a point and then starts working against itself, because every superseded document is a chance to be quoted back at you.\n\n## A practical way to start\n\nTake the last fifty questions your business was asked, by customers or by staff. Write down which document answers each one. The documents that come up repeatedly are your library. Anything that answers nothing goes in the archive folder, outside the tool.\n\nThis takes an hour and produces a better library than a month of adding files as you think of them, because it is built from what people actually ask rather than from what you imagine they need."},"check":{"question":"A physiotherapy clinic wants an assistant that answers staff questions from its clinical protocols and patient-handling procedures. Which route should it consider first, and why?","options":[{"text":"A notebook product, because the citation feature lets staff verify every answer against the source protocol.","correct":false,"tempting":"Picks on the strength of one feature while ignoring that the documents in question describe patient care and would be uploaded to an external service."},{"text":"A local setup with Ollama and a tool such as AnythingLLM, because the material is confidential and this route keeps every document inside the clinic.","correct":true},{"text":"A project inside whichever assistant staff already use, since the protocols are internal documents rather than patient records and the convenience outweighs the setup cost of anything else.","correct":false,"tempting":"Treats internal as equivalent to non-sensitive, when clinical protocols and handling procedures sit close enough to patient care that sector rules commonly apply."},{"text":"Whichever option offers the largest document limit, since protocols accumulate and a system that cannot hold them all will need replacing.","correct":false,"tempting":"Optimises for a capacity constraint a clinic will not reach, and lets a specification detail decide a question that confidentiality should decide."}]},"prev":{"slug":"embeddings-for-non-programmers","title":"Why it finds \"refund\" when the customer typed \"money back\"","url":"https://addaly.com/learn/ai-for-small-business/embeddings-for-non-programmers"},"next":{"slug":"scanning-the-paper-drawer","title":"Turning paper into rows you can check","url":"https://addaly.com/learn/ai-for-small-business/scanning-the-paper-drawer"},"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) 'Four ways to keep a small library, and what each costs', in AI for a Small Business. Edited by M. Mun. Addaly. Available at: https://addaly.com/learn/ai-for-small-business/the-tools-that-hold-your-files (Accessed: 13 October 2026).","licence":{"name":"CC BY-NC-SA 4.0","url":"https://creativecommons.org/licenses/by-nc-sa/4.0/"},"reader":{"schema":1,"title":"Four ways to keep a small library, and what each costs","courseTitle":"AI for a Small Business","module":{"index":3,"slug":"your-own-knowledge","title":"Your own documents, and getting straight answers out of them"},"answer":null,"keyIdea":"Notebook products, projects in tools you already pay for, features inside software that already holds your data, and a local setup are four places a small document library can live, and the deciding questions are whether the material is confidential and how you would get it out again.","hero":{"kind":"plate","image":null,"alt":null,"caption":null},"blocks":[{"t":"h2","id":"the-choice-is-about-where-the-files-live","runs":[{"text":"The choice is about where the files live"}]},{"t":"p","runs":[{"text":"The capability is much the same everywhere. What differs is who ends up holding your documents, how much work it is, and what happens when you want to leave."}]},{"t":"h3","id":"1-a-notebook-product","runs":[{"text":"1. A notebook product"}]},{"t":"p","runs":[{"text":"Google’s NotebookLM is the best-known, and free at the time of writing. You add up to a few dozen sources (documents, PDFs, pasted text, web pages, a YouTube link) and ask questions across them. Answers carry inline citations back to the source passage, which is the feature that matters."}]},{"t":"p","runs":[{"text":"Strong points: nothing to set up, the citations are good, and it is deliberately confined to your sources rather than answering from general knowledge. It also generates summaries and study material from your own documents, which is more useful for training a new employee than it sounds."}]},{"t":"p","runs":[{"text":"Weak points: your documents sit in a large company’s cloud, the free tier’s limits change, and it is not designed as a permanent business system. Read the terms if the documents are sensitive."}]},{"t":"h3","id":"2-a-project-inside-the-tool-you-already-use","runs":[{"text":"2. A project inside the tool you already use"}]},{"t":"p","runs":[{"text":"ChatGPT projects, Claude projects, Gemini in Google Workspace, Copilot in Microsoft 365. Attach files to a workspace, and every conversation in it starts with those files available."}]},{"t":"p","runs":[{"text":"Strong points: no new vendor, no new subscription if you already pay, and it sits where you are already working. For a business with fewer than twenty reference documents this is usually the right answer."}]},{"t":"p","runs":[{"text":"Weak points: file limits are modest, and how the files are used varies between products: some read everything, some retrieve passages, and the interface rarely tells you which."}]},{"t":"h3","id":"3-inside-software-that-already-holds-the-data","runs":[{"text":"3. Inside software that already holds the data"}]},{"t":"p","runs":[{"text":"Your helpdesk, your CRM, your shop platform. If your help centre software has an AI answer feature, it is already grounded on your articles, and no document leaves a system you have not already trusted."}]},{"t":"p","runs":[{"text":"This is underrated for the same reason as before. The data is there, the contract exists, and there is no second copy."}]},{"t":"h3","id":"4-on-your-own-machine","runs":[{"text":"4. On your own machine"}]},{"t":"p","runs":[{"text":"Free and open: "},{"text":"AnythingLLM","b":true},{"text":", "},{"text":"Open WebUI","b":true},{"text":", or "},{"text":"LM Studio","b":true},{"text":", each paired with a local model through "},{"text":"Ollama","b":true},{"text":". Point them at a folder, and you have a private document assistant that works offline."}]},{"t":"p","runs":[{"text":"Strong points: nothing leaves the building, no per-question cost, no limits, and it settles most of the privacy conversation in one move."}]},{"t":"p","runs":[{"text":"Weak points: an afternoon of setup, answers a step below the best hosted models, and you are your own support desk. For a clinic, a law practice, an accountant or anyone holding other people’s confidential material, that afternoon is well spent."}]},{"t":"h2","id":"choosing","runs":[{"text":"Choosing"}]},{"t":"p","runs":[{"text":"Ask three questions in order."}]},{"t":"p","runs":[{"text":"Is the material confidential?","b":true},{"text":" If yes, start at option four or option three, and treat one and two as unavailable until you have read the terms properly."}]},{"t":"p","runs":[{"text":"Do you already pay for something that does this?","b":true},{"text":" Usually yes, and usually nobody has looked."}]},{"t":"p","runs":[{"text":"How many documents?","b":true},{"text":" Under twenty, a project is fine. Under a hundred with citations mattering, a notebook product. More than that, or growing, you want something built for it."}]},{"t":"h2","id":"the-question-to-ask-before-you-commit","runs":[{"text":"The question to ask before you commit"}]},{"t":"p","runs":[{"text":"How do I get my documents and my answers out?"}]},{"t":"p","runs":[{"text":"Files you uploaded should be downloadable. Answers, summaries and any structure you built are often not, and rebuilding a library in a new tool takes an afternoon you had not budgeted. Keep the master copies of every document in a folder you control, and treat whatever the tool holds as a copy. That single habit makes every one of these four reversible."}]},{"t":"h2","id":"what-a-small-library-actually-looks-like","runs":[{"text":"What a small library actually looks like"}]},{"t":"p","runs":[{"text":"Businesses that get value here are not running a document warehouse. A typical working set is six to fifteen files:"}]},{"t":"ul","items":[[{"text":"The current price list."}],[{"text":"Terms and conditions, and the refund policy."}],[{"text":"The two or three procedures people ask about most."}],[{"text":"A frequently-asked-questions page built from real customer emails."}],[{"text":"Supplier lead times and delivery areas."}],[{"text":"The staff handbook, if you have one."}]]},{"t":"p","runs":[{"text":"That is enough to answer most of what anybody asks in a week, and it is small enough that one person can keep it current in twenty minutes a month. Resist the urge to add the archive. A library grows in usefulness up to a point and then starts working against itself, because every superseded document is a chance to be quoted back at you."}]},{"t":"h2","id":"a-practical-way-to-start","runs":[{"text":"A practical way to start"}]},{"t":"p","runs":[{"text":"Take the last fifty questions your business was asked, by customers or by staff. Write down which document answers each one. The documents that come up repeatedly are your library. Anything that answers nothing goes in the archive folder, outside the tool."}]},{"t":"p","runs":[{"text":"This takes an hour and produces a better library than a month of adding files as you think of them, because it is built from what people actually ask rather than from what you imagine they need."}]}],"check":{"question":[{"text":"A physiotherapy clinic wants an assistant that answers staff questions from its clinical protocols and patient-handling procedures. Which route should it consider first, and why?"}],"options":[{"text":[{"text":"A notebook product, because the citation feature lets staff verify every answer against the source protocol."}],"correct":false,"tempting":[{"text":"Picks on the strength of one feature while ignoring that the documents in question describe patient care and would be uploaded to an external service."}]},{"text":[{"text":"A local setup with Ollama and a tool such as AnythingLLM, because the material is confidential and this route keeps every document inside the clinic."}],"correct":true,"tempting":null},{"text":[{"text":"A project inside whichever assistant staff already use, since the protocols are internal documents rather than patient records and the convenience outweighs the setup cost of anything else."}],"correct":false,"tempting":[{"text":"Treats internal as equivalent to non-sensitive, when clinical protocols and handling procedures sit close enough to patient care that sector rules commonly apply."}]},{"text":[{"text":"Whichever option offers the largest document limit, since protocols accumulate and a system that cannot hold them all will need replacing."}],"correct":false,"tempting":[{"text":"Optimises for a capacity constraint a clinic will not reach, and lets a specification detail decide a question that confidentiality should decide."}]}],"order":[3,1,2,0],"why":null},"prev":{"slug":"embeddings-for-non-programmers","title":"Why it finds “refund” when the customer typed “money back”","minutes":10},"next":{"slug":"scanning-the-paper-drawer","title":"Turning paper into rows you can check","minutes":9}}}