Slug: compare-two-proposals-evidence-table
Tags: Small Business, Critical Thinking, decision making
Meta description: Compare two proposals using a shared evidence table, source locations and explicit uncertainties before making a business decision.
Two proposals arrive for the same project. One looks polished and concise. The other is longer and full of detail. Comparing them by impression alone can hide a more important problem: they may be offering different work under the same project name.
Before asking AI which proposal is better, build a shared evidence table. It should show what each document actually promises, where the promise appears and which questions remain unanswered. This gives the person making the decision a clearer basis for comparison.
Write the decision brief first
Name the outcome you need, the essential requirements and the constraints you already know. For a fictional website project, those might include the pages required, content responsibilities, accessibility expectations, approval stages, ongoing support and a deadline.
Keep essential requirements distinct from preferences. A style you like may be negotiable. A required delivery date may not be. Decide how you will use the comparison before producing a table full of attractive but irrelevant features.
Use the same questions for both documents
Read each proposal against a common set of fields. Start with deliverables, exclusions, timeline, responsibilities, review rounds, support and stated charges. Add fields that genuinely matter to the project.
Where a proposal does not answer a question, record “not stated”. Do not treat an omission as either a promise or a refusal. An important ambiguity deserves written clarification from the supplier.
| Comparison field | Proposal A | Proposal B | Question to resolve |
|---|---|---|---|
| Content preparation | Client supplies final text | One editing round included | Who prepares the initial text? |
| Post-launch support | Not stated | Defined support period | What issues and response times are covered? |
| Approval stages | One final review | Two named review points | What happens if a change is requested after approval? |
These entries are invented examples. They show how a table can reveal different assumptions without declaring either supplier superior. In a real comparison, include a source location for every entry.
Ask AI to extract evidence, not invent completeness
Compare the two supplied proposals against the fields below. Use only the documents. For each entry, identify the document and section or page. Mark missing information as “not stated”, and separate an explicit promise from an inference. Do not select a winner. Finish with the questions that would most affect the decision.
This original prompt makes the output easier to inspect. If your tool cannot read a document reliably, use readable text or a manual table. Do not assume that an uploaded file has been fully processed because the system returned a confident answer.
Where Insight AI 360 fits
The existing MaryChuks Insight AI 360 profile describes document summaries and questions about PDFs among its document-intelligence functions. Explore that profile if locating information in a proposal is your immediate problem.
Check current application availability, access terms and relevant functions before choosing it for a project. The comparison method here is a reader workflow; it does not claim that a particular application automatically performs every stage.
Verify the rows that matter most
Open the original section for any entry that could change the decision. Check numbers, dates, exclusions and responsibilities directly. A model may merge statements from different sections or produce a neat summary that loses a qualification.
Preserve the original documents and the date of the comparison. If a supplier sends a revision, record which version you are evaluating. An old table can be accurate for an earlier proposal and still mislead the current decision.
Separate the evidence table from your judgement
Once the table is checked, apply the requirements in your decision brief. You might reject an option that does not meet a non-negotiable condition, seek clarification or revise the project scope. The table organises evidence; the decision still includes priorities and trade-offs.
If you use scores, explain the criteria and weights. A number can make a subjective choice look more objective than it is. Test whether a modest change in the weights would reverse the result, and investigate why.
NIST’s risk-management guidance emphasises understanding an AI use in context and making responsibilities clear. Here, that means keeping the business requirement, source evidence and accountable decision owner together. It does not make a proposal comparison a legal review.
The useful output is a better conversation
Take the unresolved questions to the suppliers. Ask for concrete answers that can be included in a revised document. “What does support cover?” is useful; “our AI says your proposal is worse” adds conflict without explaining the requirement.
Business: compare the work you will receive. Psychology: reduce the influence of presentation by making the evidence visible. AI: use extraction and organisation to support a decision you can explain.
Next step: open the Insight AI 360 profile, then build one comparison with the original proposals beside it. For checking summaries, read our AI summary reading audit.
References: MaryChuks Insight AI 360 profile; NIST AI RMF Playbook: Map.
Featured image: conceptual illustration created with AI for MaryChuks.com.
Discover more from Marychuks.com AI, Psychology, Business & CreativeVerse
Subscribe to get the latest posts sent to your email.
