In this article:
- Overview
- Before you begin
- Building your knowledge hub step by step
- Organizing your knowledge hub
- Upload limits and supported file types
- Maintaining your knowledge hub
- Using your knowledge hub across AutogenAI
- Best practices
- Summary
Overview
Your knowledge hub is the content library AutogenAI draws on to write with your evidence, in your voice. Build it before your first project, and every project starts with source material.
It holds three kinds of reusable content:
- Libraries and folders: documents such as case studies, past responses, policies, and CVs
- Datasets: structured data in rows and columns, such as CVs or company facts
- Saved prompts: prompts your team has refined and wants to reuse
When you write, source providers let you choose which of these AutogenAI uses. When you create a project, you can connect existing libraries so the AI has context from the start.
Why build it first
- More of your evidence in every draft. AutogenAI can only draw on what you give it.
- Faster project setup. Connecting an existing library is one step, instead of uploading files each time.
- Reuse across every proposal. Build once, use on every project and document.
This guide walks you through building your knowledge hub from scratch.
Before you begin
A little preparation up front saves rework later. Work through this checklist before you add anything.
- Check your access. You only see the libraries and folders you have access to. Uploading and organizing needs document manager permissions (see below).
- Gather your best content. Pull together the documents you want AutogenAI to draw on.
- Clear out the clutter. Leave out outdated, duplicate, or draft versions.
- Decide how you'll organize it. Follow how your team looks for content, for example by topic, sector, or client.
The document manager role
Your organization assigns document Managers. They hold the permissions for uploading and organizing the knowledge hub, and they keep it well-structured and easy to navigate.
Some organizations appoint designated staff to keep things uniform. Others let all team members upload and organize.
- No document manager permissions? Contact your project lead and your Customer Success Manager.
- A library or folder is missing? Ask your administrator.
What to upload
Upload anything a proposal writer in your organization would refer to during research. This typically includes:
- Previous proposals
- Standard operating procedures
- Specifications and case studies
- Marketing collateral
- Annual reports
- Government reports
Choose by relevance and accuracy, not polish:
- Include a poor-scoring proposal that contains useful information.
- Include a badly written handbook or SOP that contains useful information.
- Leave out a beautifully written document if its information is out of date.
Rule of thumb: the only requirement for including a document is that it contains useful information.
Documents or datasets?
Not all content belongs in the same place. Use this table to decide.
| Content type | Examples | Where it goes |
|---|---|---|
| Narrative evidence | Case studies, past responses, policies, method statements | Library documents, organized in folders |
| People information | CVs and biographies | A CV dataset |
| Repeated facts | Company revenue by year, standard answers to common questions | A free form or answer bank dataset |
Tip: If a document holds many separate facts that you ask about again and again, a dataset usually returns more consistent answers than the document alone.
Building your knowledge hub step by step
Building the knowledge hub takes six steps. Steps 1 to 4 set up your document library. Steps 5 and 6 add structured data and reusable prompts. They are optional, but worth doing.
Step 1: Open the knowledge hub
Select Knowledge Hub from the left-hand navigation. You will see the libraries you have access to. Each library is a collection of documents, folders, datasets, and saved prompts.
Step 2: Choose or create your library
- Open the library you want to build in, or create a new one if your team is starting from scratch.
- Choose a name your whole team will recognize at a glance, such as the name of your team or business unit.
Step 3: Add your documents
Upload the source documents you gathered. These become the evidence AutogenAI draws on when you generate content. You need Document Manager permissions.
- Select your library, then upload files or folders.
- Add your strongest, most current content first. You can add more later.
- Use clear file names. "Case study - NHS digital transformation" is easier to find than "Final v3".
- Check your files against the upload limits and supported file types below.
Step 4: Organize with folders
Group your documents into folders that follow how your team searches for content, for example by topic, sector, or client.
Folders matter beyond tidiness. When you generate content, you can choose specific folders or files as your source instead of the whole library. A clear structure lets you point the AI at exactly the right material. See Organizing your knowledge hub for structure options.
Step 5: Add datasets for structured data
Datasets turn repeated facts, such as CVs, revenue figures, or standard answers, into searchable rows and columns. To create one:
- In your library, open the Datasets tab.
- Select +Dataset. Create a blank dataset, or upload an existing Excel or CSV file.
- Enter a name and a short description. Both are required, and they help AutogenAI understand what the dataset contains.
- Choose a Dataset Type (see the table below).
- Name each column clearly and specifically, for example "Company revenue (GBP)" rather than "Details".
To extract data from files already in your library, select the three dots in the top right and choose Populate Dataset from Files.
| Type | Behavior | Use it for |
|---|---|---|
| CV | One row per source file | Files that each represent a single record, such as one CV per file |
| Free Form | Can extract multiple rows per file | Files that legitimately contain several records |
| Answer Bank | Multiple rows per file | Question-and-answer content with many pairs per source |
If you upload a spreadsheet, keep it clean:
- One header row
- No merged cells
- No multi-tab workbooks. AutogenAI reads only the first tab of an Excel file.
Tip: Column names act as instructions to the AI. Run a few test questions after setup to confirm answers come from the right column. For a full walkthrough, see Datasets.
Step 6: Save the prompts your team reuses
Once a prompt works well, save it so your whole team can reuse it.
- Open Saved Prompts, found under your library sources in the knowledge hub.
- Select New Prompt, give it a name, and paste in your prompt text.
- To use it later, select the Saved Prompts icon in the Ask AI or research prompt box. The prompt fills in, ready to edit or send.
Quick reference
| Step | What you do | Required? |
|---|---|---|
| 1. Open the knowledge hub | Go to your libraries | Yes |
| 2. Choose or create your library | Pick where your content will live | Yes |
| 3. Add your documents | Upload your source content | Yes |
| 4. Organize with folders | Group content by topic, sector, or client | Recommended |
| 5. Add datasets | Create structured data for repeated facts | Optional |
| 6. Save your prompts | Store reusable prompts for the team | Optional |
Organizing your knowledge hub
Structure your knowledge hub in the way that suits your business. The goal is to make it easy for anyone on your team to select folders and groups of documents, and to read a folder's file path at a glance.
Common ways to structure it
| Approach | Example |
|---|---|
| Thematically | Social value, data governance |
| By product line | Lawnmowers, vacuum cleaners, garden furniture |
| By business vertical | Healthcare, construction |
| By organizational structure | Policies and procedures, marketing materials |
| By mirroring your file-sharing platform | SharePoint or Google Drive |
Keep bid-specific content separate
Where possible, create a private library for the proposal you are working on. Include:
- Tender documentation and clarifications
- PowerPoints from the commissioner
- Capture material and transcripts
This stops bid-specific content polluting the rest of your knowledge hub.
Upload limits and supported file types
| Limit | Value |
|---|---|
| Maximum file size | 2GB per file (around 1,000 pages of A4) |
| Files per library | 20,000 (standard limit) |
| SharePoint sync | 1TB total synced content |
| Approved file types | pdf, doc, docx, xls, xlsx, xlsm, ppt, pptx, csv, txt, msg, odt, ods, odp, gif, jpg, jpeg, png, webp |
Maintaining your knowledge hub
A knowledge hub is only as useful as it is current. A few habits keep it that way.
- Replace, don't pile up. When you update a case study or policy, swap out the old version so AutogenAI does not draw on outdated content.
- Review datasets after changes. AutogenAI does not automatically refresh existing rows when you add columns or update source files.
- Keep datasets focused. Each dataset should cover one topic, client, or subject area.
- Name consistently. Agree on a pattern for file and folder names so anyone can find content quickly.
- Check access. If a colleague cannot find content they need, review their access with your administrator.
- Name an owner. Assign someone to review the knowledge hub regularly, so upkeep does not depend on whoever notices a gap.
Deleting files
- In the knowledge hub, open the library.
- Select the file or folder so it is highlighted.
- Select Delete in the toolbar.
Uploading a duplicate
The knowledge hub keeps the latest version of a document. If you upload a file with an identical file name, it replaces the old file instead of creating a duplicate. To update a document, upload the new version under the same file name.
Tip: For help with knowledge management or optimizing your proposal library, contact your Customer Success Manager or Implementation Consultant.
Using your knowledge hub across AutogenAI
Once your knowledge hub is built, it feeds your work in four places.
| Where | How your knowledge hub is used |
|---|---|
| Creating a project | In the project setup step for your project library, connect existing libraries from the knowledge hub, or upload files directly. You can also skip this and add sources later. |
| Source providers | On the homepage or editor page, choose the knowledge hub, datasets, or the Internet (web search) as the sources AutogenAI draws on. |
| Ask AI and the assistant | AutogenAI generates content or answers questions from the sources you have selected. |
| Saved prompts | Select the saved prompts icon in the prompt box to reuse a team prompt. |
Narrowing your sources
By default, AutogenAI can draw on everything you have access to. To focus the selection:
- Go to Settings
- Click Source Selection.
- Toggle off Generate using all library documents.
Choose the libraries, folders, or files you want. Every library and folder you have access to appears in the menu.
Tip: Use a narrower selection when the answer should come from specific evidence, such as one case study or one set of CVs. Use the full library for broader questions.
Connecting your library to a project
When you are ready to start a proposal, the next article, Creating a new project, takes you through project setup. At the project library step, select the libraries you just built so AutogenAI has context for your storyboard from the start.
Best practices
- Start small and strong. A focused library of your best, most current content beats a large archive of everything you have ever written.
- Choose by usefulness, not polish. Include accurate, useful content even if it is badly written. Leave out polished content that is out of date.
- Match the content to the container. Use documents for narrative evidence and datasets for repeated facts.
- Name things for the people who will search. Clear file, folder, and column names help both your team and AutogenAI.
- Keep bid-specific material in a private library. It stays out of your reusable content.
- Test before you rely on it. After setting up a dataset, run a few sample questions to confirm answers pull from the right column.
- Build before your first project. It makes project setup quick and gives AutogenAI context from day one.
Summary
Building your knowledge hub gives every future project a head start. The path is short:
- Prepare. Check your access, gather your best content, and decide how to organize it.
- Build. Add documents to your library and group them into folders.
- Structure. Add datasets for repeated facts, and save the prompts your team reuses.
- Maintain. Replace outdated files and review the hub regularly.
- Connect. Use your knowledge hub through source providers, and connect it to each new project.
For questions or advice about structuring your folders, contact support@autogenai.com or speak to your Implementation Consultant or Customer Success Manager.