Introduction
Welcome to the instructional guide for utilizing the advanced capabilities of Awarded.AI's semantic searching tools. This guide is designed to help you understand and effectively use semantic searching, vectorization, and embeddings to streamline and optimize your data retrieval tasks.
By rapidly analyzing and generating text, Awarded.AI can achieve up to 80% efficiency improvements. However, it's crucial to manually review the remaining 20% of the output to ensure accuracy, as the AI may occasionally create incorrect information.
Overview
Awarded.AI leverages Generative AI, a cutting-edge form of Artificial Intelligence, to streamline and optimize traditionally labor-intensive tasks. By rapidly analyzing and generating text, the system can achieve up to 80% efficiency improvements. However, it's crucial to manually review the remaining 20% of the output to ensure accuracy, as the AI, despite its sophistication, may occasionally create incorrect information.
Key Terminology
Prompting: Instructing the AI in a chat format. Clarity and detail yield better responses.
Example: "Act as an expert proposal writer. Using a persuasive and inspiring tone, assist me in crafting an executive summary for my government proposal."
Context Window: The AI retains memory for each session of about 400-800 pages. For optimal results, paste no more than 15 pages into the chat at once.
Conversational: The AI retains memory within the ongoing session, allowing it to recall prior text in the conversation.
Example: After a generated response, you can say, "Summarize that into bullet points."
Semantic Searching: Upload and direct the AI to single or multiple files, or even whole folders. Pose a question, and the AI scans your specified files for answers. This enables swift information retrieval and response formulation. However, the system might struggle to focus or extract extensive data from the files. When laser focus is required, copy and paste is the best method for information injection into the chat window (*note – this is called direct injection).
High-Level Capabilities for Semantic Searching
Text Creation: Craft content tailored to any topic.
Text Enhancement: Adjust for sentiment, enhance quality, and refine readability.
Text Condensation: Summarize by word or paragraph count.
Text Insights: Extract qualitative data, recognize named entities, and analyze sentiment.
Competitor Insight: Delve into market trends, pinpoint weaknesses, and conduct SWOT analysis.
Agency Understanding: Grasp mission/values and identify core pain-points.
Strategic Vision: Employ SWOT analysis, highlight differentiators, and plot growth.
Innovation: Brainstorm solutions, ideate differentiating factors, and suggest novel features.
Proposal Evaluation: Produce section overviews, identify potential risks, and craft pertinent questions.
Draft Review: Ensure compliance, examine readability, and suggest scoring enhancements.
Past Performance Reference: Retrieve relevant past performance data and address associated requirements.
Understanding Vectorization and Embeddings
Vectorization and embeddings are key to semantic searching. They transform text into high-dimensional vectors that capture the semantic meaning of the content.
Vectorization
Definition: Converting text into vectors to capture semantic meaning.
Process: Text is converted into vectors, which are then used to find conceptually similar documents.
Example: When the AI processes a document, it breaks down the text into smaller passages and converts them into vectors. These vectors are like GPS coordinates that map out the semantic meaning of the text.
This image shows a grid of pixelated data points on the left being transformed into smooth, continuous lines on the right, with arrows indicating the transformation process.
Embeddings
Definition: Representing text as vectors in a high-dimensional space.
Process: Text is broken down into smaller passages, which are then indexed and stored.
Example: The AI uses embeddings to understand the context and meaning behind the text, allowing it to retrieve relevant documents even if they don't contain the exact search terms.
Improving Prompts and Data Sets
To optimize search results, it's important to improve the quality of prompts and data sets.
Crafting Effective Prompts
Be Specific: Provide clear and specific instructions.
Use Natural Language: Write prompts in natural language for better comprehension.
Optimizing Data Sets
Clean Data: Ensure data is clean and well-organized.
Structured Data: Use headers, subheaders, and bullet points to structure data logically.
Common Use-Cases
Understanding Your Company's Strengths
Using AI Chat: Create a new session, copy & paste in your capabilities statement (aka direct injection prompting), and use the following prompts to explore your strengths.
Prompts:
"Create me a SWOT analysis of our business based on our capabilities."
"Create me a Market analysis around our business competencies."
"Recommend additional business verticals my company could expand into for new revenue streams or contract opportunities."
"Based on my role as a {ROLE}, what do you recommend I do to better myself and advance my career? Ask me relevant questions until you have a strong enough understanding of my background, company, and role."
Past Performance Searching
Using AI Chat: Create a new session, activate the “Our Files” switch, and select your past performance library folder.
Prompts:
"What have we done with {INSERT_SERVICE} before and with what agencies?"
"Respond to the following requirements based on the services we have done before with different agencies. {PASTE_IN_PAST_PERFORMANCE_REQUIREMENTS}"
"Turn our past performance around {INSERT_SERVICE} into a narrative showing we are a strong vendor/partner selection for a government agency."
Teaming Partner Search
Using AI Chat: Create a new session, activate the “Our Files” switch, and select your teaming partners library folder.
Prompts:
"Which company has experience performing {INSERT_SERVICE} with the {INSERT_AGENCY}?"
"Create me a list of companies with experience working with the {INSERT_AGENCY}."
Build a Compliance Matrix
Using AI Chat: Create a new session, select the ‘Prompt Library,’ and in the featured list, select the compliance matrix prompt.
Instructions:
Copy/Paste the requirements/task area of a solicitation into the {Insert_requirements} section.
After the matrix is completed, you can copy and paste it into Excel. (The right-click option will allow you to keep the formatting when pasting.)
Analyze Agency Pain-Points
Using AI Chat: Create a new session.
Prompts:
"What is the primary mission and values of the {INSERT_AGENCY}?"
"Create me a list of 10 pain-points this agency likely faces."
"Suggest 15 words I should include in my proposal to resonate with this agency."
Solicitation Review
Using AI Chat: Create a new session, activate the “Our Files” switch, and select the uploaded file.
Prompts:
"What are the different sections/task areas?"\
"In section {INSERT_SECTION}, create me a list of the top key points, a list of key risks, a list of 5 questions I should consider asking the agency for clarification, and a list of 5 oral presentation questions that might be asked to my team."
"Create me a win strategy and proposal plan for this solicitation."
Proposal Drafting
Using AI Draft: Create a new session, configure the information about your company, the agency, and the system or service.
Instructions:
Once you have created an outline that includes all proposal requirements, click “send to draft,” navigate to your draft in the history pane (left-hand side of the screen), and select AI Draft Copilot.
If you have successfully synced your files and have those accessible, using the Files tab in this view, select your similar past proposals, past performances, reference files, or similar information to supplement the AI.
Complete the About You section in this view. Fill in key information about your company, your project, and any additional desired guidelines you want the AI to reference during this draft session (i.e., a specific tone or voice, “be as brief as possible,” always refer to our company’s nickname after the first reference, etc.).
Select the Write tab in this pane, highlight the requirement, and select the “Respond to Requirement” option.
Once the response is created, you can manually edit the generated tex,t or you can highlight the generated text and select a text editing option such as summarize or make more detailed.
In some cases, you can highlight the entire document (under 30k characters) or an entire section and have the AI create multiple responses at once.
*Note: The information should be verified by SMEs for accuracy and is designed to get you rapidly to a first draft, not a blind submission.
Review Proposal Draft
Using AI Review: Enter your requirement/task area and then enter your draft response. Select ‘Run AI Review.’
Instructions:
The insights generated will be custom to the requirement/response pair.
You can launch a second window and open AI Chat or AI Draft.
Feed the original requirement into the chat and then paste in the desired feedback to have the AI rewrite a new response variation.
Thank you for using Awarded.AI! For further assistance, please contact your Customer Success Manager.