Better Context. Better Questions. Better AI.
AI can be an incredibly useful assistant in government contracting, but the quality of what you get back depends heavily on what you give it. A vague question usually produces a vague answer.
Anatomy of a Good GovCon Prompt
A well-constructed prompt gives AI:
- A role
- The right context
- A specific job
- Clear boundaries
- An expected output
- Instructions for verification
This framework is designed specifically to encourage responsible AI use in government contracting.
Use AI to Think Better. Not to Stop Thinking.
- Analyze
- Organize
- Compare
- Brainstorm
- Question
- Draft
- Edit
- Summarize
- Structure
- The solicitation
- Your solution
- Your pricing
- Your contractual obligations
- Your company's capabilities
- The promises you're making to the government
You remain responsible for the work.
Let's Start With a Bad Prompt.
What's Wrong With This Prompt?
R → C → T → C → O → V
R — ROLE
Who are you asking AI to help you think like?
Giving AI a role provides perspective and helps establish the kind of analysis you want.
"Review this."
"Act as an experienced federal proposal manager reviewing this response for compliance."
- Federal Capture Manager
- Proposal Manager
- Technical Reviewer
- Red Team Reviewer
- Pricing Analyst
- Contract Administrator
- Business Development Strategist
- Federal Market Research Analyst
Assigning AI a role does NOT make it an actual credentialed professional.
"Act as an attorney" does not make AI legal counsel. "Act as a CPA" does not make AI an accountant.
Use roles to establish analytical perspective, not professional authority.
C — CONTEXT
What are we working on, and why?
AI performs better when it understands the situation.
"Analyze this."
"We are a small facilities management contractor evaluating whether to pursue this Department of Veterans Affairs opportunity. We are currently in the pre-solicitation capture stage."
- Company type
- Company size
- Capabilities
- Customer
- Agency
- Opportunity
- Capture stage
- Proposal stage
- Prime vs. subcontractor
- Business objective
- Relevant experience
- Known constraints
Provide the information that matters to the task.
T — TASK
What exactly do you want AI to do?
Give one clearly defined task at a time when possible.
Specific tasks generally produce more useful outputs than broad requests.
"Help me with this RFP."
"Review Section L and identify every instruction affecting the Technical Volume."
- Identify requirements
- Compare documents
- Develop questions
- Analyze weaknesses
- Challenge assumptions
- Create an outline
- Evaluate a draft
- Identify unsupported claims
- Summarize changes
- Develop potential win themes
- Explain terminology
- Create meeting questions
C — CONSTRAINTS
What should AI NOT do?
Constraints are particularly important in government contracting.
- Use only the documents I provided.
- Do not invent requirements.
- Do not assume customer preferences.
- Do not invent competitor information.
- Do not create past performance metrics that I haven't provided.
- Do not assume an attachment says something unless I provided the attachment.
- Do not interpret missing information as confirmation.
- Clearly distinguish fact from inference.
- If information is missing, identify it as unknown.
- Do not rewrite contractual language in a way that changes its meaning.
This is one of the most important habits in responsible GovCon prompting.
O — OUTPUT
What should the answer look like?
Don't make AI guess how you want the information organized.
The right output format makes AI results easier to review and verify.
- Table
- Checklist
- Executive summary
- Compliance matrix
- Outline
- Action plan
- Questions list
- Risk register
- Comparison
- Draft narrative
- Meeting agenda
- Bullet list
Create a table containing: Requirement, Solicitation Reference, Page, Required Response, Proposal Location, Owner, Status, Questions / Ambiguities.
V — VERIFY
How will we know whether the answer is supported?
Verification is especially important in government contracting. Tell AI how to show its work.
- Cite the solicitation section and page for every requirement.
- Identify which uploaded document supports each conclusion.
- Clearly label assumptions.
- Mark anything that cannot be verified.
- Quote only enough source language to identify the requirement.
- Do not state an inference as a fact.
- Flag conflicting requirements.
- Tell me when you don't know.
If AI can't tell you where it got the answer, you should know that before you rely on it.
Build the prompt step by step
Read this RFP and tell me what to do.
Look at the Difference.
"Read this RFP and tell me what to do."
ROLE: You are assisting me as an experienced federal proposal manager. CONTEXT: We are preparing a proposal for a federal facilities management solicitation. I am currently developing the Technical Volume outline. TASK: Analyze Section L and identify every instruction affecting the Technical Volume. CONSTRAINTS: Use only the solicitation documents I provided. Do not invent requirements. Do not assume referenced attachments contain requirements unless those attachments were provided. Clearly distinguish explicit requirements from interpretation. If something cannot be determined, mark it Unable to Verify. OUTPUT: Create a table containing: Requirement, RFP Section, Page, Required Response, Page Limit, Formatting Requirement, Questions / Ambiguities. VERIFY: For every requirement, cite the source section and page. After completing the table, identify anything you believe requires human verification.
The second prompt does not make AI infallible. It makes:
- The task clearer
- The assumptions narrower
- The output easier to review
- The source easier to verify
- The human more capable of evaluating the result
Don't Ask AI to Swallow the Whole RFP and Hope for the Best.
"Read this 300-page solicitation and tell me everything I need to know."
Use a structured workflow.
- 1. Identify the solicitation structure.
- 2. Review Section L.
- 3. Review Section M.
- 4. Review PWS / SOW / SOO.
- 5. Review required forms and attachments.
- 6. Review amendments.
- 7. Crosswalk L + M + technical requirements.
- 8. Build compliance matrix.
- 9. Conduct human review against the source.
Why chunking matters
- Hundreds of pages
- Tables
- Attachments
- Referenced documents
- Forms
- Conflicting instructions
- Amendments
- Q&A
- Pricing schedules
- Technical exhibits
- Information that supersedes earlier information
- Miss information
- Misread tables
- Lose context
- Overlook attachments
- Confuse versions
- Fail to recognize that an amendment changed an earlier requirement
Therefore: break complex work into manageable analytical tasks.
Build Your Own GovCon Prompt
Who should AI help you think like?
What are you working on?
What exactly should AI do?
What should AI not assume or do?
How should the answer be structured?
How should AI demonstrate that its answer is supported?
GovCon prompt starters
These examples exist to teach the pattern. Adapt them to your company, your customer, and your documents.
AI Is an Assistant. Not Your Proposal Team.
- Outlines
- Messages
- Questions
- Analysis
- Compliance reviews
- Drafts
- Graphics concepts
- Editing
- Comparisons
- What you're proposing
- How you'll perform it
- How you'll staff it
- What it costs
- What risks you're accepting
- What contractual promises you're making
If your proposal could result in your company accepting hundreds of thousands, or millions, of dollars in contractual obligations: someone on your team needs to understand and own every promise you're making.
"AI wrote it" does not change your obligation to perform.
Could You Explain It Without the Proposal in Front of You?
- Could your Project Manager explain the technical approach?
- Could your proposed key personnel explain the solution?
- Could your pricing team explain the assumptions?
- Could leadership explain the staffing model?
- Could your team explain why your solution works?
- Could you defend your past performance claims?
- Could you answer evaluator follow-up questions?
- Oral presentations
- Demonstrations
- Clarifications
- Discussions
- Final Proposal Revisions
- Interviews or other exchanges where permitted by the procurement
The specific acquisition determines what applies. Regardless of how the government evaluates the proposal, your company should be prepared to explain and perform what it submitted.
STOP. What Information Are You About to Share?
- CUI
- Classified information
- Personally Identifiable Information
- Proprietary company information
- Third-party proprietary information
- Pricing
- Employee information
- Government-controlled information
- Source-selection information
- Procurement-sensitive information
- Export-controlled information
- Contractually restricted information
- Their contractual obligations
- Customer requirements
- Cybersecurity requirements
- Company AI policies
- Approved information systems
- Data handling restrictions
Not every AI system is appropriate for every category of information. When uncertain, do not upload sensitive information until appropriate authorization and controls are confirmed.
The Human Verification Check
Where did the answer come from?
Does the original source actually support it?
Does it apply to this procurement and situation?
What could the AI have missed or misunderstood?
Does someone qualified understand and agree with the conclusion?
R-C-T-C-O-V GovCon Prompting Cheat Sheet
One page covering Role, Context, Task, Constraints, Output, and Verify, plus the Human Verification Check. "AI can accelerate the work. You still own the answer."
A Prompt Is Not a Workflow.
A strong prompt can help you analyze a problem. Complex GovCon work often requires:
- Multiple documents
- Structured data
- Version history
- Requirement tracking
- Saved company information
- Cross-document analysis
- Repeated reviews
- Human verification
That's why GovCon Learning Pro tools are built as structured workflows, not simply prompts.
- Capture Planner
- Solicitation Decoder
- Proposal Compliance Workspace
- Evaluator Lens
- Pricing & Cost Readiness
Use AI to Think Better. Not to Stop Thinking.
Good prompting isn't about finding magic words. It's about giving AI:
- The right role.
- The right context.
- A clear task.
- Boundaries.
- A useful output.
- And a way to verify the answer.
Then bring your own judgment to the result.