This is the second article in our “AI for Nonprofits” series. If you’re new to AI, read Part 1: Getting Started, If you’re comfortable with basic prompting, you’ve found some helpful use cases, and you’re starting to wonder: “What else can these tools do?”, this article is for you. It explores advanced techniques that transform AI from a helpful assistant into a true productivity multiplier for nonprofit work. It references examples from both ChatGPT and Claude, often using the + sign or tools icon to access additional features.
Memory & Context Management
The most important skill in using AI tools is knowing how to structure your prompt and provide context.
Memory
AI tools maintain context throughout a conversation and use the instructions from your account profile and project. As tools improve, they are starting to develop “memory”.
ChatGPT handles this automatically through memory options in your account settings. Go into settings > personalization > memory > manage and see what it has remembered about your previous conversations. This is a fairly new feature, so expect these memories to improve over time.
Claude handles this by letting you explicitly reference past chats which it can read directly. Try asking Claude to search through previous chats and summarize what you have discussed on a given topic. Now that summary is available for the rest of your conversation.
Prompt Engineering
Moving beyond basic questions unlocks dramatically better results:
Chain-of-Thought Prompting: “Think step-by-step through this budget allocation decision. Show your reasoning for each consideration.”
Role-Based Prompting: “Act as a foundation program officer reviewing our proposal. What concerns would you have? Be critical.”
Structured Output Requests: “Provide three fundraising approaches in a table with columns for: Approach, Time Required, Cost, Pros, and Cons.”
Few-Shot Examples: Show AI 2-3 examples of your writing style or format, then ask it to match that style for new content.
Context Management
Take advantage of attaching files to your project and adding the web search tool to your conversation. Consider adding a link to your website, your strategic plan,, your organization chart, and any other information germane to your specific request. Most AI tools can handle 100-200 pages of context (about 200,000 words). For larger document sets, ask the AI to summarize key sections first, then use those summaries for your main task.
Collaborating on Artifacts
AI tools specialize in generating text, but they are getting better at generating other artifacts like spreadsheets, presentations, or even interactive apps. Regardless of the type of artifact, the first draft will likely need revision. You will want to make changes, and you may even want to invite others on your team to suggest updates as well.
Artifacts (Claude)
Artifacts create interactive documents in a side panel—standalone content you can edit, share, and publish. They’re perfect for content that will be used outside the conversation. Claude does not currently allow you to update these artifacts directly, but you can describe to Claude what you want changed, or you can copy and paste the content into a shared Google Drive document, work with your team to update it, and then tell Claude to read your updates and update the artifact.
Canvas (ChatGPT Plus)
Canvas is a side-by-side workspace for real-time document and code editing. Ask ChatGPT to create a proposal, click “Use Canvas,” then edit directly while requesting changes.
Custom GPTs (ChatGPT Plus)
Create specialized AI assistants for recurring tasks. Build a “Grant Writer GPT” loaded with foundation guidelines, your case for support, and successful proposals. Anyone with the link can use your Custom GPT.
Shared Conversations
Both platforms allow sharing conversations via link. This lets you collaborate with colleagues or show stakeholders your AI-assisted work process.
Multimodal Capabilities
In addition to text, AI tools have become multimodal, supporting audio, images, and video.
Uploading Images
Most AI platforms now analyze images. Take photos and let AI interpret them.
Nonprofit applications:
- Whiteboard brainstorming sessions → organized summaries
- Event photos → social media captions
- Handwritten forms → digital transcription
- Data charts from reports → analysis and insights
- Competitor materials → competitive analysis
Example: “Transcribe this whiteboard, organize ideas by theme, and identify the top 3 priorities we should focus on.”
Generating Images
Create custom graphics for social media, presentations, and marketing materials.
ChatGPT supports image generation directly. Try “Create an image showing diverse volunteers at a food bank with warm, welcoming lighting.”
Claude does not generate images, but it does have integration with Canva. Canva is an excellent design tool that provides access to a large library of images, and includes the ability to use AI to generate your own images. Canva provides a free license for nonprofits.
Voice Mode
Both Claude and ChatGPT support voice mode for hands-free interaction in their mobile apps. Dictate instead of typing—perfect after site visits, during meetings, or when multitasking.
Basic Voice Mode (Free): Transcribes your speech and responds in text. Works well for dictation and straightforward queries.
Advanced Voice Mode (ChatGPT Plus): Conversational AI with natural interruptions, emotional tone recognition, and fluid dialogue. It’s like talking to a colleague.
Nonprofit use cases:
- Dictating meeting notes while facilitating
- Capturing ideas during site visits
- Hands-free brainstorming while commuting
- Accessibility for team members who prefer speaking
Deep Analysis & Research
Extended Thinking & Deep Research
Claude’s extended thinking mode takes extra time to reason through complex problems, showing its thinking process. This is valuable for nuanced analysis, strategic decisions, and multi-faceted problems. Chat GPT has two tools, Think Longer and Deep Research. Think longer spends more time analyzing and collecting ideas from its existing knowledge base before generating the response, while deep research adds web search to do extended thinking using the latest information available on the internet.
Research process:
- You pose a research question
- AI creates a research plan
- Conducts multiple searches
- Synthesizes findings
- Provides comprehensive report with citations
Nonprofit applications:
- Best practices in volunteer retention (with current data)
- Funding landscape for environmental education in Ontario
- Sector trends in youth mental health programs
- Comparative analysis of similar organizations
Data Analysis
Upload CSV or Excel files to segment donors, analyze attendance trends, identify retention patterns, or spot program outcomes.
Analytical capabilities:
- Statistical analysis and visualizations
- Trend identification
- Segmentation and clustering
- Predictive insights
- Data cleaning and preparation
Example: “I’ve uploaded 3 years of donor data. Analyze retention patterns. What percentage of first-time donors give again? Which donor segments have the highest lifetime value? Create visualizations showing trends.”
Connectors
AI platforms are extending beyond their chat interface to connect with your existing tools and information sources.
Web Search
Most AI tools now search the web for current information beyond their training data, which is typically about a year old. This may happen automatically, but to ensure web search is used click the + sign in the conversation and add the web search tool.
Google Drive Integration
Claude and ChatGPT can connect to your Google Drive (with your permission) to read and analyze documents stored there.
Capabilities:
- Search across your Drive for relevant information
- Analyze multiple documents without downloading
- Reference organizational knowledge stored in Drive
- Create summaries of folder contents
Privacy: You control access. AI only sees files you explicitly grant permission to access.
Desktop Integration
Some AI platforms offer desktop applications that can:
- Read content from your screen (with permission)
- Analyze local files without uploading
- Integrate with local applications
- Work offline for some features
Claude Desktop can reference files on your computer and see your screen content when you grant permission. ChatGPT requires uploading files or screenshots at this time.
Additional Tools & MCP Servers
Model Context Protocol (MCP) allows AI to connect with your existing business tools—CRMs, calendars, databases, and custom applications. Check if your AI platform offers connectors for tools you already use.
Read vs. Update Access
When granting connectors access to your tools, understand the difference:
Read Access: AI can view and analyze information but cannot modify it. This is safer for most use cases.
Update Access: AI can make changes to documents, calendars, databases. Only grant this when necessary and appropriate.
Best practice: Start with read-only access. Only enable update access for specific, controlled use cases.
Privacy & Security
Incognito Mode (ChatGPT Plus / Claude Pro)
Paid subscriptions offer private conversation modes where:
- Nothing is saved to your conversation history
- Data isn’t used for training AI models
- Conversations are deleted when you close them
When to use incognito:
- Sensitive strategic planning
- Confidential financial analysis
- Personnel discussions
- Competitive intelligence
Connector Security
When connecting AI to other tools:
Verify permissions: Understand exactly what access you’re granting. Read permission dialogs carefully.
Use service accounts: For organizational use, consider dedicated service accounts rather than personal credentials.
Regular audits: Periodically review connected applications and revoke access you no longer need.
Data handling: Understand where your data is stored, how long it’s retained, and who can access it.
What Never to Share
Even in incognito mode, avoid sharing:
- Actual donor names or personal information
- Banking details or financial credentials
- Social Insurance Numbers or health information
- Proprietary strategy that would harm you if disclosed
- Information subject to confidentiality agreements
Use placeholders: Replace sensitive data with “[Donor Name]”, “[Amount]”, or “[Client ID]” when seeking AI assistance.
Establishing Organizational Policies
Create guidelines for your nonprofit:
Approved Tools: Which AI platforms are acceptable for organizational use?
Prohibited Information: What data should never be shared with AI?
Privacy Settings: Required privacy configurations before use.
Approval Requirements: Which activities require supervisor approval?
Review Processes: How AI-generated content should be reviewed before publication.
Action Step: Draft a one-page AI usage policy for your organization covering these five areas.
Conclusion: From User to Power User
The difference between AI users and AI power users isn’t just access to premium features—it’s understanding when and how to apply advanced techniques strategically. Start experimenting with the advanced techniques identified above, and subscribe to updates from your AI tool provider so you get informed as new features and capabilities get rolled out.
Questions? Want to go deeper?
🎓 Register for our “Practical Use of AI for Nonprofits” workshop
🤝 Request personalized AI consulting from our expert team
Our next blog will cover Workflow Automation and AI Agents.
