Best Practices for Prompting the AI Project Analyzer
The Project Analysis feature is a powerful tool that helps you screen new projects by asking critical questions about the documents you upload. By crafting effective prompts, you can get more accurate and insightful answers from the AI, allowing you to identify potential red flags and make better go/no-go decisions.
This guide provides best practices for structuring your questions to get the most out of the Project Analyzer. Each analysis runs a pre-configured set of static questions, so designing clear and targeted questions is the key to success.
💡 Pro-Tip: How Pre-configured Questions Work Together
When you run a Project Analysis, all of your questions and their contexts from all categories are sent to the AI as a single, combined request. This gives the AI a holistic view of your priorities. Consider making your questions across different categories complementary to reinforce what's most important for your analysis.
Be Clear and Specific
Vague questions lead to vague answers. The more specific you are, the more useful the AI's response will be. Use the New Question field for your main query and the Additional Context field to guide the AI on what to focus on.
Example 1: Analyzing Risk
New Question:
Identify all clauses related to data security and privacy.Additional Context:
Highlight any requirements in these clauses that deviate from standard GDPR, such as data residency rules or breach notification timelines.
Example 2: Understanding Payment Terms
New Question:
What are the payment terms and schedule?Additional Context:
Extract all details related to the payment schedule, including invoicing frequency, net payment terms (e.g., Net 30, Net 60), and any penalties for late payment.
Use the Context Field to Provide Your Business Rules
The Additional Context field is perfect for telling the AI why you are asking the question. Give it your business rules or "deal-breakers" so it can analyze the document from your specific point of view.
Example 1: Checking for On-Premise Requirements
New Question:
Does this document contain any requirements for on-premise deployment or customer-hosted solutions?Additional Context:
We are a SaaS company and do not offer on-premise installations. Identify any language that would make us non-compliant with this delivery model.
Example 2: Checking for Subcontracting Constraints
New Question:
Are there any restrictions on using subcontractors to deliver the work?Additional Context:
Our delivery model relies on using certified third-party subcontractors for implementation. Identify any clauses that forbid or require prior approval for subcontracting.
Assign a Persona
You can assign a persona in the Additional Context field to instruct the AI to adopt a specific point of view for its analysis. This focuses the response on the aspects a particular expert would care about.
Example 1: Legal Review
New Question:
Identify any clauses related to liability and indemnification.Additional Context:
Act as a legal counsel. Analyze these clauses to see if they expose our company to unlimited liability and extract the exact wording.
Example 2: Financial Analyst Review
New Question:
Summarize the pricing model requested by the customer.Additional Context:
Act as a financial analyst. Identify if the request is for a fixed-fee, time and materials, or other pricing model. Note any requirements for volume discounts or specific currency.
Break Down Complex Questions
Instead of asking one very broad question, use multiple question slots to break down a complex topic into smaller, focused queries. This ensures you get detailed and accurate information for each distinct area.
Instead of one question like: "Analyze the entire RFP and tell me everything that's important."
Create multiple, specific questions:
Question 1:
What are the mandatory requirements for this RFP?Question 2:
What is the evaluation criteria for vendor selection?Question 3:
Are there any specific formatting requirements for the response?Question 4:
What is the timeline for submission and vendor selection?
Ask for Specific Formats
Instruct the AI to format its response in a particular way. This makes the information easier to digest and act upon.
Example 1: Formatting Reference Requirements
New Question:
What are the customer reference requirements?Additional Context:
Extract all requirements for customer references and format the output as a table with three columns: 'Customer Name', 'Industry', and 'Date of Project Completion'.
Example 2: Formatting Key Dates
New Question:
What are all the key dates and deadlines mentioned in the document?Additional Context:
Extract all dates and present them as a bulleted list with the format: 'Event/Milestone: YYYY-MM-DD'.
Using 'Add Question' for Deeper Analysis 💬
After your initial analysis runs, you may have new questions. The "Add question about the documents" feature lets you ask ad-hoc, follow-up questions. Think of it as starting a new, direct conversation with the source documents. This feature does not remember your previous questions or answers, so each new query is a fresh start.
⚠️ Important: Ad-Hoc Questions are not remembered
Each time you use the "Add question about the documents" feature, your query is sent to the AI without any memory of previous questions. The AI does not have context from your initial analysis or any other ad-hoc questions you've asked. For best results, ensure every follow-up question is self-contained and includes all the context needed for the AI to find the right answer.
Here are some best practices for using it effectively:
1. Clarify Ambiguities
If an initial answer was vague because the source text was vague, use a follow-up question to search the entire document for clarifying details.
Scenario: The initial analysis for "Liability" returned a clause mentioning the liability cap is tied to fees paid in "the preceding period."
Good Follow-Up Question:
Search the entire document for a definition of "preceding period" or any related terms that clarify how the liability cap is calculated.
2. Reference Specific Sections
To get a precise answer, guide the AI to the right part of the document. This helps narrow the focus and yields better results.
Scenario: The initial analysis summarized the Service Level Agreement (SLA). You want to know more.
Good Follow-Up Question:
In the 'Service Level Agreement' section, what is the exact definition of "System Uptime" and how is it measured?
3. Extract Specific Wording
Use this feature to pull the exact, word-for-word text you might need for a proposal or a legal review.
Scenario: You know the document contains termination clauses, but you need the precise language.
Good Follow-Up Question:
What is the exact, word-for-word clause related to contract termination for convenience?