Ready-to-use prompts, aligned to the Shipley methodology, for every stage of your proposal.
Great prompts are the difference between a generic first draft and a compelling, compliant, customer-focused response. This article walks through how to craft a strong prompt, how to choose the right sources, and gives you copy-paste prompt templates for each stage of the proposal lifecycle — from qualify through to review.
The prompting formula
Every successful prompt follows one of these 2 formulas:
Formula
Structure
Basic
Context + Specific Request
Advanced
Purpose + Topic + Context + Structure + Specific Request
What each part does:
Purpose — the why. What outcome are you after — inform, persuade, decide?
Topic — the what. The main subject, so the AI stays focused and relevant.
Context — the background, constraints, or specifics that frame the question precisely.
Structure — how you want the response organized (list, comparison, numbered steps, etc.).
Specific Request — the exact instruction that pins down what the output needs to include.
Tip: Also assign a role, so the model knows what perspective to draw from — e.g. "You are a proposal analyst…"
Prompting pro tips
Show examples of what good looks like. If you want a particular style or format, paste one or two examples — this is called few-shot prompting, and it beats any amount of description.
Iterate rather than expecting perfection the first time. Treat it like a conversation: get a first draft, then refine — "make the second paragraph punchier," "cut the jargon," "give me 3 alternative openings."
Get the AI to improve your own prompt. Run a rough prompt, note what you wanted vs. what you got, then feed that back and ask the AI to rewrite the prompt itself.
Choosing your sources
Before you prompt, decide where the AI should draw its content from. Source Providers lets you specify this, and can be found on the Home page or your Editor page.
There are 3 source types:
The Knowledge Hub — your organization's library and folders
Datasets — structured data you've selected, with the option to specify columns
The Internet (Web Search) — with the option to include only specific sites, or exclude specific sites
Toggle off "Generate using all Library Documents" or "Generate using all datasets" to narrow generation down to specific folders, files, or datasets rather than your whole library.
Ask AI vs. Research
Ask AI — the most effective way to prompt inside a document. Highlight text in the Editor, then type into the Ask AI prompt box. It draws on the Library, the Internet, or Datasets depending on the task.
Research — for direct questions to the AI from your Library or the Internet, found on the Home page. You can filter to specific documents or websites and ask follow-up questions based on the output.
Context is key — if you don't ask, you don't get. Enter all relevant context in your prompt and avoid ambiguous language.
Saving prompts for reuse
Once you have a prompt that works well, save it so your whole team can reuse it instead of rewriting it each time.
Go to 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 a saved prompt, look for the Saved Prompts icon in the Ask AI or Research prompt box — clicking it shows your full collection.
Selecting a saved prompt auto-populates it into the prompt box, ready to edit or send.
Saved Prompts are accessible from Research/Home and from Ask AI in your projects and drafts, and existing ones can be edited at any time.
Prompt templates by proposal stage
Copy and adapt any of the templates below — swap in the bracketed placeholders for your own details.
Qualify
Competitive analysis
Who are the main competitors of [insert company name] in the following area? What known track history does each competitor have of delivering [insert contract type]?
Useful follow-up research prompts:
What challenges, constraints, and limitations is the client facing?
What are risk considerations for the client in completing this procurement?
What are the client's buying habits according to the documents (e.g. do they always hire the incumbent, go with low cost, etc.)?
What is the competitor's experience working with the client, and what are their likely bidding behaviors and key discriminators?
What are [Customer]'s primary strengths related to this opportunity, and where are the opportunities to neutralize the competition?
Prepare
Extracting requirements
Create a table of all category types of requirements that must be fulfilled to successfully deliver this contract. The first column must display the type of requirement. The second column must have a brief description of the requirement. The third column must convey any format requirements. The fourth column must have the pages and/or page ranges the requirements appear in the document.
Other useful requirement prompts:
List all the requirements in the document related to [insert subject of the question]. Use headings where necessary.
What are the stated deadlines or timeline expectations outlined in the documents?
Which terms or expressions does the client emphasize? What insights can we gain about their reasons for commissioning the project?
Are there any other requirements that the supplier should or should not do relating to [insert subject area]?
What requirements or asks are missing from this list: [insert requirements]?
Generating win themes
Write a compelling win theme for [Client's name]'s proposal. Begin by analyzing their key priorities and challenges to identify their core needs. Then, articulate how [our Company's] distinctive capabilities, proven track record, and quantifiable successes provide superior solutions. Include specific evidence to substantiate claims, such as measurable outcomes or case studies, and explain how these solutions deliver unparalleled value compared to competitors. Ensure the win theme is consistently integrated across all proposal sections to maintain a persuasive and cohesive narrative that resonates with the client's objectives.
Creating a storyboard
Using the [Company's] library, create a storyboard for the following proposal questions: [insert questions]. Each bullet point from the question must form a subheading, and placeholders for relevant content must be placed under each subheading.
To storyboard a whole specification at once, create a variable (e.g. "Structure") with values for the types of content you want pulled through — such as awards won, certifications, case studies, added value, and keywords from the client's specification — then run:
Provide a summary of how [Company name] should answer this question: [insert question]. Include areas on <variables: structure>
Write
Generating a first draft
Using [X] amount of words, explain how [Company's name] can [insert proposal question]. Explain how [insert key information]. Write in full paragraphs and each claim must be backed up with evidence.
Repurposing an old response
This proposal response was for [insert sponsor here]. Now I am writing a proposal for [insert NEW sponsor] answering the same question. My answer needs to be adapted to the specific needs of [insert NEW sponsor] as outlined in the specification in [Company's] library. Look at the relevant parts of the specification related to [insert relevant topic] and tell me which parts of this proposal answer need to change, and how it should be changed to address [insert NEW sponsor]'s requirements.
Review
Reviewing a first draft
Using the criteria from an "Outstanding" evaluation rating scale sourced from the library, review this proposal response [to the initial question] to identify what criteria has been met, not met, or unclear. Present the output as a table. Include recommendations to improve.
For reviewing multiple responses against a spec at once:
Use the requirements and supplied <rating scale> specified in the "[ITT/spec document]" to evaluate responses in the "[responses document]" document for the question [RFP question]. Provide feedback in a table on how each response met, did not meet, or was unclear, and recommendations to improve.
Incorporating feedback
Does this proposal answer the below question convincingly? If it does not, state exactly what text I should remove and what text I should replace it with. Don't just tell me what is wrong — give me new text to improve the answer, and state exactly where in the answer the new text should go. Ensure that the suggested text is drawn from material within [Company's] library — this could be case studies, statistics, track record, step-by-step processes, staff experience, illustrative examples, or more.
Once you have feedback (from a reviewer or from Gamma Review), rewrite against it directly:
Rewrite this text to address this feedback: [insert feedback]. Use illustrative examples, processes, case studies, statistics, and data as appropriate from [insert Company's] library to provide the evidence required.
Key takeaways
Prompts should be clear, unambiguous, and relevant — avoid ambiguous language and always provide context.
Prompts can range from a single word to multiple paragraphs, so give the model as much detail as the task needs.
AutogenAI has different tools and features to prompt with depending on the stage of the proposal writing process — match your prompt template to your proposal stage.
Troubleshooting
My output is too generic
Add more context and a specific request — and check your Source Providers are pointed at your library, not the open internet.
The AI isn't using my own case studies or evidence
Confirm your Source Providers are set to your Knowledge Hub library, and toggle off "Generate using all Library Documents" to point it at the specific files you want used.
I can't find a prompt I saved
Check the Saved Prompts icon in the Ask AI or Research prompt box, or go to Saved Prompts under your library sources in the Knowledge Hub.