Skip to main content

Building Your Knowledge Base in Awarded AI

Awarded AI starts every chat with a clean slate. Your Knowledge Base is how you give it lasting memory of your company, your past performance, and the way you like to work. Here's how to set it up so the AI shows up ready on day one.

Written by Chelsey Hernandez

In this guide

  1. Why the Knowledge Base matters

  2. First things first: key terms

  3. How nodes and sub-nodes work

  4. The semicolon rule (important for efficiency!)

  5. Populating your Company Profile

  6. Populating your Past Performances

  7. Creating custom Knowledge Categories

  8. Turning it on: the Sources Panel and default configuration

  9. If you own the Knowledge Base

Why the Knowledge Base matters

Awarded AI works a little differently from the public AI tools you may be used to, like ChatGPT, Claude, or Gemini. The most important thing to understand is this: Awarded AI is a non-training model. For security and functional reasons, it does not retain or remember information across sessions. Every new chat begins with a clean slate, with no company or user data carried over from your previous chats.

That is a good thing for security, but it means the AI does not automatically know who you are, what you have won, or how you like your proposals written. Your Knowledge Base is the workaround. It is a repository of information you build once so you can feed it to the AI quickly, at the start of any session, without pasting the same details over and over.

📌 The big idea: The Knowledge Base is your company's long-term memory. Anything you find yourself repeating to the AI, whether it is your CAGE code, a past project, or a writing rule, belongs here so you only have to say it once.


First things first: key terms

You will run into these words all over Awarded AI, so here is what each one actually means.

  • Knowledge Base (Edit Mode). The full repository where you add and edit your company information. Think of it as the workshop where you build everything out. Find it under Data Hub > Knowledge Base.

  • Sources Panel (Configure Mode). A condensed version of that same information, right on the AI Chat screen. This is where you toggle knowledge on or off for the session you are working in. Same information as the Knowledge Base, just a different place that does a different job. (See the 'Turning it on: the Sources Panel and default configuration' section below for more information on configurations)

  • Knowledge Category. A top-level bucket of related information, like Company Profile, Past Performances, or a category you create yourself such as Preferences, Boilerplate Info, Resumes, or Teaming Partners.

  • Node and sub-node. The individual pieces of information inside a category, and the detail that sits beneath them. More on these next.


How nodes and sub-nodes work

Inside a Knowledge Category, information is organized into nodes. A category can hold as many nodes as you want, and each node can have its own sub-nodes beneath it. It is a simple parent-and-child structure:

  • A main node is a heading or a grouping, for example a single project name.

  • A sub-node sits underneath it and holds the detail, for example the description of that project.

You get to decide how deep to go. For past performance, some people keep it flat with one node per project. Others create a main node per project and place the write-up as a sub-node. Both are fine. Choose whatever will be easiest for you and your team to scan later.


The semicolon rule (important for efficiency!)

This is the single most useful thing to know about building nodes quickly, and it catches everyone off guard the first time. In Awarded AI, the semicolon is a reserved character. Its only job (as far as the Knowledge Base goes) is to separate one node from the next.

When you type text into a node field and then click anywhere else on the page, Awarded AI splits your text at every semicolon and turns each chunk into its own separate node. Type A;B;C into a single field, click away, and you instantly get three nodes: A, B, and C. **You can easily test this out to ensure you understand the concept!

💡 Why this is powerful: If you already have your information somewhere else, like a spreadsheet, you can have the AI format all of it as one block with a semicolon between each entry. Then you paste the whole thing into a single node field once, and every entry auto-splits into its own node. No copying and pasting entries one at a time.

Because the semicolon is reserved for splitting, there is one rule you have to follow so it doesn't backfire:

⚠️ Important: Never use a semicolon inside a node's own text as punctuation. If you do, the AI will split that node in the wrong place. When you need a semicolon grammatically within an entry, use a pipe | or other delimiter instead.

Semicolons create new nodes.


Populating your Company Profile

Your Company Profile is the first default category (**the other default category that's editable: Past Performance), and it should already be mostly filled out when you first log in, pulled from publicly available sources about your company. Your job is to review it for accuracy and fill in anything missing. This is an incredibly crucial step because a comprehensive, current Company Profile directly feeds Awarded AI's content creation and its automated opportunity fit scoring (via the Bid Search).

What to include

Add anything you would want the AI to know about your company, including:

Company details

Codes and qualifications

Industry partnerships

CAGE code and other relevant industry codes

List of agencies you have worked with

Certifications

Services and products

Set-asides

Mailing and physical addresses

Contract vehicles

Authorized signer information

Custom fields (add your own)

If something you need is not covered by the existing fields, scroll to the bottom of the profile and add a custom field. The more complete the profile, the better the AI performs.


Populating your Past Performances

Past Performances is the other default editable category worth building out early, and here is the key reason: along with your Company Profile, Past Performances is one of only two categories that feed the automated opportunity fit scoring in Bid Search. Other categories you create will not. So even if you are not focused on the pipeline yet, getting this one in good shape pays off later.

Unlike the Company Profile, this category is not pre-filled, so you will populate it yourself. The good news is the AI can help avoid some of that manual labor! We created a prompt that does most of the heavy lifting.

🔎 You can find this prompt in the Crawl Training Companion Word Document, which you automatically receive if you have attended a live Crawl training (run live as a webinar the first Tuesday of each month - register here!); or you can ask your CSM for it, and they can either send it to you or add it to your Prompts library directly in Awarded.

Using the "Summarize Past Proposals" prompt

  1. Add the Summarize Past Proposals into PP Entries prompt to your Prompt Library. Open Prompts > Library > New Prompt, paste it in, give it a title, fill in your client name where the prompt asks for it, and save. Toggle it to Shared if you want teammates to be able to use it too.

  2. Start a new chat, apply the prompt, and attach your past proposal files (or a spreadsheet that already holds the details).

  3. Send it. The prompt extracts the salient, proposal-ready details from each source and formats them into a clean, reusable record: contract overview, project description, key contacts, capability highlights, and performance results, with pipes | used inside each record so nothing accidentally splits.

  4. Copy the output, go to your Past Performances category, and paste it into a node field. Because each record is separated by a semicolon, every past performance drops into its own node automatically.

💡 Tip: The default prompt is set to process proposals one at a time and pause for your review. That is a safety feature, not a limitation. If you would rather batch several at once, you can edit the prompt to say so. And always read the output before you save it, to make sure it captured everything you want the AI to know.

💡 Another Tip: If you want to edit the prompt to reflect your unique dataset/situation, give the AI the original prompt, tell it you want to edit it, and tell it HOW you want to edit it. Then ask it to revise the prompt for you!

💡 Already have a spreadsheet where you have all your past performance data/details saved? Even easier. If your past performance already lives in a spreadsheet, you can adapt the prompt to extract every row verbatim, label each piece with its column name, and separate each row with a semicolon. Attach the spreadsheet, run it, then paste the result straight into your category. Every row becomes its own node in one paste. It will not look pretty in the field, and that's fine (though admittedly difficult for us Type-As). The AI only needs the raw data, not tidy formatting.


Creating custom Knowledge Categories

Once your two defaults are solid, create new categories for anything else the AI should know. A good rule of thumb: create a category for information you find yourself repeating. Specifically, make a new category for anything you:

  • find yourself attaching as a file over and over,

  • want all of your users to communicate consistently (a style guide or boilerplate, for example),

  • could benefit from distilling and simplifying, or

  • could use regularly to streamline a workflow.

Categories worth creating

Category

Why it helps

Boilerplate Company Info

Reusable language you drop into most responses.

Templates

Executive summary starters, section shells, and formats.

Company Style Guide and Rules

How you want the AI to write, including your preferences and things to avoid.

Previous Proposals

Your strongest past submissions for reference and voice matching.

Resumes

Key personnel bios ready to tailor.

Teaming Partners

Partner capabilities so the AI can suggest who fills a gap.

Competitor Data

Intel to sharpen your positioning.

Capabilities Statements and Product Info

What you offer, in the AI's reach for every draft.

📌 The Preferences category is a favorite. A "Preferences" or "Rules" category is one of the highest value things you can build. It is where you tell the AI your writing rules once and reuse them forever; for example, a rule to avoid the em dash, a rule to write in formal narrative prose rather than bullet lists, or a rule to always cite the source file when pulling facts and to never make anything up or hallucinate. The Companion document includes ready-to-use examples you can adapt.


Turning it on: the Sources Panel and default configuration

Building your Knowledge Base is only half the job. The AI still needs to know which knowledge to use in a given session. That is what the Sources Panel does. A configuration is simply the combination of categories and nodes you have toggled on. You can toggle a whole category on or off and use the Configure button to select or deselect individual nodes within it.

The one you will use most is your default configuration, the set of toggles that turns on automatically every time you log in and start a new session. Setting it up means the AI always opens with a baseline of knowledge about your company.

  1. On the AI Chat screen, open the Sources Panel and toggle on the categories and nodes you want active by default.

  2. When you make a change, a yellow prompt appears. Review your selections, then click Update Config to save them as your default.

  3. You can also save other configurations for different kinds of work, but your default is the important one to get right.

⚠️ Heads up, this is per user. There is no admin who can set one default configuration for everyone. Each user sets up their own. If you want your whole team working from the same baseline, align on it internally and have everyone toggle on the same categories.


If you own the Knowledge Base

If you are the person responsible for your organization's Knowledge Base, here are a couple of tips to help you build it out and maintain it over time.

Begin with the end in mind

Before you create categories, think about how your users will look for information later, then structure your categories so they can find it efficiently. A little planning up front saves a lot of hunting down the road.

Designate a Knowledge Base Manager

Have one person own keeping the Knowledge Base accurate and current. Review it on a regular cadence to remove anything outdated and add what is new. A simple way to get started: hold an internal meeting, align on what belongs in the Knowledge Base, assign categories to different people to populate, then come back together to review.

💡 You have PSci support available to you! Your CSM is in Awarded AI every day and is happy to help you complete your profile, build out categories, and refine your prompts. You can also use the live chat inside the app anytime, or email [email protected].

Procurement Sciences · Save Time. Deliver Faster. Win More.

Did this answer your question?