BETA PRODUCT
AI Proposal Assessment
Getting pitched AI solutions but don’t know how to evaluate the proposals? These structured prompts will help you review proposals with rigor and confidence. Use them to better understand what's being pitched and whether it meets your standards for responsible AI.
About the Tool
AI can exponentially scale social impact – but only if done right. Adapted from The Philanthropist's Guide to Nonprofit AI Investments, these prompts are designed to surface what needs further discussion in a proposal.
The prompts work across all major AI assistants and are designed for non-technical reviewers. By being thoughtful and intentional with funding, you play a leading role in ensuring AI is a genuine force for good.
This tool does not summarize what's there, make a funding recommendation, or replace reading the proposal yourself. Most importantly, engage with curiosity. AI is new for most nonprofits. They may not have perfect answers — yet.
Instructions for Use
Before starting: Remove any personally identifying information from the proposal if your organization's policies require it.
Step one: Open your AI assistant: Gemini, Claude, ChatGPT, or any other.
Step two: Upload or paste the grant proposal you're reviewing.
Step three: Pick a prompt below for the question you want answered, copy it, and paste it into your assistant.
Step four: Use the structured output to guide your review and follow-up questions.
Step five: Return to the grantee with any follow-up questions.
Impact & Purpose
Select a question below for a ready-to-use AI prompt.
What problem is the nonprofit addressing with AI?
Use this prompt to clearly define the social or environmental issue that the AI solution aims to solve.
From the attached grant proposal, write questions to ask the applicant to better understand the evidence behind the problem and why AI is the right intervention. If the problem statement is vague or conflated with mission, suggest what to explore in conversation. Address the applicant directly using 'you.' Avoid summarizing what the proposal says — the funder has already read it. Respond in 2-3 sentences only. No headers, no lists, no preamble.
How will the AI solution create positive change?
Use this prompt to identify the measurable outcomes and impact the nonprofit expects to achieve through AI.
From the attached grant proposal, write questions to ask the applicant to better understand the measurability of outcomes and the specific role AI plays in achieving them. Where outcomes are broad or the AI connection is unclear, frame this as something worth exploring in conversation. Address the applicant directly using 'you.' Avoid summarizing what the proposal says — the funder has already read it. Respond in 2-3 sentences only. No headers, no lists, no preamble.
Who are the beneficiaries?
Use this prompt to ensure the solution benefits the intended community.
From the attached grant proposal, write questions to ask the applicant to better understand how the AI solution creates value for each beneficiary group. Distinguish between direct beneficiaries (who uses the tool) and ultimate beneficiaries (who benefits downstream) — and probe whether the causal pathway between the two is clearly explained or assumed. Address the applicant directly using 'you.' Avoid summarizing what the proposal says — the funder has already read it. Respond in 2-3 sentences only. No headers, no lists, no preamble.
What is the potential for scalability and sustainability?
Use this prompt to assess whether the solution can be expanded to reach a wider audience and continue beyond the grant period.
From the attached grant proposal, write questions to ask the applicant to better understand the organization's plan for scale and sustainability beyond the grant period. If continuation depends on funding that isn't yet described, frame that as worth exploring. Address the applicant directly using 'you.' Avoid summarizing what the proposal says — the funder has already read it. Respond in 2-3 sentences only. No headers, no lists, no preamble.
How will the nonprofit communicate learnings with the wider community?
Sharing knowledge and lessons learned can benefit the AI for Humanity field. Use this prompt to decipher whether there is an opportunity to improve and expand large-scale open data initiatives and infrastructure.
From the attached grant proposal, write questions to ask the applicant to better understand the organization's plan for sharing data or learnings. Suggest questions about venues, timelines, and long-term governance of shared assets. Address the applicant directly using 'you.' Avoid summarizing what the proposal says — the funder has already read it. Respond in 2-3 sentences only. No headers, no lists, no preamble.
Capacity & Resources
Select a question below for a ready-to-use AI prompt.
Does the nonprofit have the resources to implement effectively?
Use this prompt to assess the organization's existing capabilities and its plan for acquiring any necessary expertise or resources.
From the attached grant proposal, write 1-2 questions to ask the applicant to better understand the team's capacity to execute the described AI work. Probe both general software/engineering capacity and AI-specific expertise (e.g., ML experience, model development, evaluation) — proposals often name staff without clarifying whether anyone on the team has hands-on AI development experience. Address the applicant directly using 'you.' Where limited detail may reflect format constraints rather than absent planning, acknowledge that context. Frame this as conversation prep, not a verdict on the proposal. Respond in 2-3 sentences only. No headers, no lists, no preamble.
What stage is the AI solution?
Use this prompt to determine what stage is the AI solution currently in (e.g., concept, alpha, pilot, fully deployed). Understanding the maturity helps assess feasibility and timelines.
From the attached grant proposal, write 1-2 questions to ask the applicant to better understand the solution's current stage and the evidence behind it. Pay attention to the gap between what exists today and what this grant will build — proposals often blur this distinction. Look for whether evidence of the current stage is concrete (deployments, users, published results) or primarily asserted. Address the applicant directly using 'you.' Where detail is limited, frame this as something worth exploring in conversation rather than a gap. Avoid summarizing what the proposal says — the funder has already read it. Respond in 2-3 sentences only. No headers, no lists, no preamble.
How robust and mature is the data that powers this solution?
Use this prompt to determine whether they have sufficient, high-quality data to train and sustain the AI model effectively.
From the attached grant proposal, write 1-2 questions to ask the applicant to better understand the organization's approach to data robustness and maturity. Pay particular attention to what happens after initial development — how the dataset will be updated, governed, or quality-controlled over time, and whether the AI model has a plan for maintenance or retraining. These post-launch questions are often absent from proposals and worth surfacing. Address the applicant directly using 'you.' Where limited detail may reflect format constraints rather than absent planning, acknowledge that context. Frame this as conversation prep, not a verdict on the proposal. Respond in 2-3 sentences only. No headers, no lists, no preamble.
Is the nonprofit collaborating with partners in the AI field?
Use this prompt to evaluate how partnerships can enhance the solution's quality and impact.
From the attached grant proposal, write 1-2 questions to ask the applicant to better understand the nature and depth of the partnerships listed. Focus on whether key partnerships are formally confirmed or still aspirational, and whether each partner's specific role in the AI work is clearly defined — proposals often name organizations without clarifying what they've actually committed to. Address the applicant directly using 'you.' Avoid summarizing what the proposal says — the funder has already read it. Respond in 2-3 sentences only. No headers, no lists, no preamble.
Is the nonprofit building an AI tool or using an off-the-shelf product?
Use this prompt to better understand the nonprofit's use of AI. This impacts building and managing effectively.
From the attached grant proposal, if the solution is custom-built or hybrid, write questions to ask the applicant to better understand the technical choices. If 'custom' may mean prompt engineering or configuration rather than original development, frame that distinction as worth exploring. Address the applicant directly using 'you.' Avoid summarizing what the proposal says — the funder has already read it. Respond in 2-3 sentences only. No headers, no lists, no preamble.
Is the nonprofit using or creating open source technology?
Use this prompt to discern whether the nonprofit has considered leveraging open source tools or contributing to open source communities. This can reduce costs and expand impact.
From the attached grant proposal, write questions to ask the applicant to better understand the organization's relationship to open source and open data. If 'open access' may refer to a proprietary platform rather than genuine contribution to the commons, frame that distinction as worth clarifying. Address the applicant directly using 'you.' Avoid summarizing what the proposal says — the funder has already read it. Respond in 2-3 sentences only. No headers, no lists, no preamble.
What is the AI solution's budget, and how will the grant funds be used?
Use this prompt to review the financial plan to ensure responsible use of funds and alignment with the solution's objectives.
From the attached grant proposal, write questions to ask the applicant to better understand the financial plan for the described AI work. If a budget is absent, note that it should be requested before proceeding. If a budget is present, probe whether AI-specific costs are addressed — compute, APIs, data acquisition, and tooling are frequently missing or undersized — and whether the overall budget appears proportionate to the technical scope described. Address the applicant directly using 'you.' Avoid summarizing what the proposal says — the funder has already read it. Respond in 2-3 sentences only. No headers, no lists, no preamble.
What is the nonprofit's long-term vision for using AI in its work?
Use this prompt to understand how this solution fits into the organization's broader strategy for leveraging AI to achieve its mission.
From the attached grant proposal, if a broader AI vision is present, write questions to ask the applicant to better understand whether it's actionable or aspirational. If no broader vision is mentioned, suggest exploring whether this project is part of a larger institutional direction. Address the applicant directly using 'you.' Avoid summarizing what the proposal says — the funder has already read it. Respond in 2-3 sentences only. No headers, no lists, no preamble.
Ethical Considerations
Select a question below for a ready-to-use AI prompt.
How will the AI solution address potential biases and discrimination?
AI algorithms can inherit biases from their training data. Use this prompt to ensure the nonprofit has plans to mitigate and monitor for biases.
From the attached grant proposal, write 1-2 questions to ask the applicant to better understand how the organization thinks about and addresses bias. Pay attention to whether the proposal acknowledges bias in general terms or engages with where bias could specifically enter their system — training data, evaluation methodology, or the populations the model was originally developed on. Address the applicant directly using 'you.' Frame this as conversation prep, not a verdict on the proposal. Avoid summarizing what the proposal says — the funder has already read it. Respond in 2-3 sentences only. No headers, no lists, no preamble.
How will the nonprofit protect the privacy and security of data?
Data privacy and security are paramount when dealing with AI. Use this prompt to evaluate how well the nonprofit executives understand and explain their data handling practices and security measures.
From the attached grant proposal, write 1-2 questions to ask the applicant to better understand the organization's data privacy and security practices. Pay attention to the gap between what the solution collects and what the proposal says about how that data is governed — AI proposals often describe the data they need without addressing what happens to it. Where populations served are vulnerable or data is sensitive, that gap is worth probing. Address the applicant directly using 'you.' Where limited detail may reflect format constraints rather than absent policy, acknowledge that context. Avoid summarizing what the proposal says — the funder has already read it. Respond in 2-3 sentences only. No headers, no lists, no preamble.
Is the downside bigger than the upside?
Use this prompt to explore potential negative impacts and how the nonprofit plans to mitigate them.
From the attached grant proposal, write questions to ask the applicant to better understand the organization's awareness of potential negative ethical impacts. If the proposal doesn't address this, suggest what to explore in conversation. Address the applicant directly using 'you.' Avoid summarizing what the proposal says — the funder has already read it. Respond in 2-3 sentences only. No headers, no lists, no preamble.
Has the nonprofit considered the broader implications of AI?
Use this prompt to discuss the potential long-term effects of AI on society and how the solution fits into the broader landscape (e.g. its impact on social issues from jobs to environmental degradation).
From the attached grant proposal, write 1-2 questions to ask the applicant to better understand the organization's awareness of broader societal effects. Pay attention to whether the proposal engages only with direct benefits or also considers second-order effects — what could go wrong at scale, who else might be affected, or what the technology could enable beyond its intended use. Address the applicant directly using 'you.' Avoid summarizing what the proposal says — the funder has already read it. Respond in 2-3 sentences only. No headers, no lists, no preamble.