Use the same questions for up to three proposals. The summaries count what you recorded; they do not verify an answer, recommend an agency, or predict a project outcome.
Scroll horizontally to review every proposal.
Evidence question
Proposal 1
Proposal 2
Proposal 3
Job and outcomeDoes the proposal name the workflow, users, desired outcome, and non-AI alternatives?
Delivery teamAre the people, roles, availability, and subcontracting expectations named?
Related workIs comparable work supported by an inspectable artifact or contactable reference?
Data readinessAre source systems, access, quality limits, privacy, and migration work explicit?
Testing and acceptanceAre evaluation data, thresholds, failure cases, and acceptance decisions defined?
Security and governanceAre permissions, approvals, logging, incident handling, and accountable owners named?
Full economicsAre implementation, usage, support, maintenance, and change costs separated?
Handoff and supportAre documentation, access, monitoring, maintenance, and exit responsibilities explicit?
Coverage summary
See the gaps before choosing.
Do not add these counts into a quality score. A documented claim can still be irrelevant, stale, or wrong; a gap may simply require a follow-up question.
Proposal 1
0 documented or testable · 0 answer-only · 8 not addressed
Proposal 2
0 documented or testable · 0 answer-only · 8 not addressed
Proposal 3
0 documented or testable · 0 answer-only · 8 not addressed
Reference request
Ask for one comparable example.
Request a project with similar users, data, constraints, and operating stakes. Ask what shipped, what failed, how success was measured, what the client maintains, and whether the reference can confirm those facts.
