Making It Work for You
Bruce Floyd
- Grew up in central Florida, ran track for the Florida Gators, earned two degrees from UF
- Managed Florida Gators social media for 10+ years — reached #1 in national fan engagement
- Founded Adroit Media to help organizations use AI to simplify workflows and grow — without the tech frustration
- Specializes in making complex ideas clear and actually useful
- Gainesville local · Live music fan · Cat dad
Hour 1 — Context
- Why most people use AI wrong
- The AI landscape right now
- Your organization's AI policy
- How to talk to AI effectively
- ↓ 10-minute break
Hour 2 — Hands On ✋
- Warm-up: multimodal AI exercise
- Live data demo
- Volunteer management exercise
- Build your org's AI brief
- Q&A + debrief
wrong.
Chatbots can answer questions — but they're not the same as search engines.
Search Engine
- Finds live web pages
- Updated in real time
- Shows you where information came from
- Helps you locate sources, websites, and documents
AI Chatbot
- Generates a written answer from its training plus tools it may use
- Has a knowledge cutoff unless it checks live sources
- May use web search for current questions — but not every time
- Can sound confident even when wrong, incomplete, or outdated
⚠️ For recent events, funding rules, statistics, or policy changes: check whether the AI used live sources — and always verify the answer.
There are two types of AI — most people only use one.
| Standard Chatbot | Reasoning Model | |
|---|---|---|
| How it works | Predicts the next word, fast | Thinks through the problem step-by-step |
| Best for | Drafting, summarizing, brainstorming | Analysis, strategy, complex decisions |
| Examples | ChatGPT (default), Claude (default) | ChatGPT o3, Claude Extended Thinking |
| Speed | Seconds | Slower — but more thorough |
| Use when | Most daily tasks | Multi-step problems that require judgment |
AI hallucinates — and it sounds completely confident when it does.
AI doesn't know when it doesn't know something. It fills gaps with plausible-sounding content that can be entirely false.
What can go wrong
- Invented statistics that sound real
- Fake citations to real journals
- Wrong names, dates, and facts
- Outdated information presented as current
The simple habit
- Read AI output critically before sharing
- Verify any statistic, citation, or fact
- Never paste AI content directly into a funder report
- Think of AI as a smart intern — check their work
Right Now
The tools you need to know
Before you buy anything — check what you already own.
Google for Nonprofits (free)
- Gemini in Gmail, Docs & Sheets
- NotebookLM — document-based AI
- Deep Research + image generation
- Up to 2,000 users, enterprise security
- nonprofits.google.com
Microsoft 365 for Nonprofits
- Copilot Chat may already be included
- Works in Outlook, Word, Excel, Teams
- Full M365 Copilot at nonprofit pricing
- Summarize meetings, draft documents
- microsoft.com/en-us/nonprofits
💡 Ask your IT contact: "Are we enrolled in Google for Nonprofits? What Microsoft licenses do we have?"
One new capability worth knowing: Deep Research
Available in ChatGPT, Gemini Advanced, and Perplexity. Browses the real web, reads multiple sources, and produces a cited research report — in minutes.
Use it for
- Grant landscape research — who's funding what
- Peer org comparisons and benchmarking
- Community needs data and demographics
- Policy background for program applications
How it differs from a regular prompt
- Actually browses live web sources
- Synthesizes many pages of information
- Cites its sources — you can verify
- Takes 2–5 minutes vs. instant response
Use Deep Research to profile a nonprofit.
Open ChatGPT, Gemini, or Perplexity → select the Deep Research option → paste this prompt, replacing [NONPROFIT NAME] with your own org or one you're curious about:
⏱ Deep Research takes 2–5 minutes to run. Start it — we'll debrief together when it finishes.
Needs an AI Policy
Why your organization needs an AI policy
Protect your clients
- Free AI tools may store conversation data
- Client names, case notes, and records should never go into a free chatbot
- HIPAA and confidentiality obligations still apply
Protect your organization
- AI errors in grant reports create liability
- Funders are beginning to ask about AI use
- Consistent staff guidance prevents reputational risk
🔒 A policy doesn't restrict AI use — it gives staff clear guidance so they can use it confidently and safely.
Some of your staff will have concerns about AI.
AI data centers consume significant electricity and water. The concern is real — even if an individual org's usage is small.
Major AI models were trained on scraped internet data, often without creator consent. Lawsuits are ongoing. If your org works with artists or the creative community, this tension is worth naming.
AI is changing what certain tasks look like. Some staff — especially in administrative roles — have legitimate reasons to feel uncertain. Don't pretend otherwise.
A policy forces the question: what does your organization actually believe about AI? That's not a liability — it's how you build trust.
What a good policy covers
1. Data Privacy
- Define what information can and can't go into AI tools
- Distinguish free tools vs. enterprise/paid accounts
2. Accuracy Standards
- Require human review before anything goes out
- No AI-generated stats without verification
3. Attribution
- When to disclose AI use to funders and stakeholders
- Be proactive — some funders now ask
4. Staff Guidance
- Which tools are approved for work use
- Who is responsible for AI-generated content
You don't have to start from scratch.
CNE has prepared a sample AI policy template you can adapt for your organization. It's today's handout.
What to do with it
- Customize for your org's size and programs
- Review with your ED and key staff
- Adopt it at a staff or board meeting
- Revisit annually — the landscape moves fast
This week's action
- Take the handout home today
- One afternoon of editing is all it takes
- Even a one-page policy is a strong start
- Your funders will appreciate your thoughtfulness
to AI
The Role / Task / Format formula
"You are a volunteer coordinator…"
If AI gives you a bad result, one of these three things is usually missing or too vague.
Specific prompts get real results.
❌ Vague
"Write a volunteer recruitment post."
✅ Specific
"You're a nonprofit volunteer coordinator. Write a Facebook post recruiting summer volunteers for a youth mentoring program serving 150 kids ages 8–14 in Gainesville. We need volunteers for weekday afternoons in June and July. Warm, conversational tone — not corporate. Under 120 words. End with a clear call to action."
AI is a conversation, not a vending machine.
❌ One-shot thinking
- Write one prompt, expect perfection
- Start over when the result isn't right
- Accept a mediocre answer because "that's what AI does"
✅ Iterative thinking
- Start with a solid prompt, review the output
- Push back: "Make it warmer." "Shorter." "Add a statistic."
- Build on what came before — don't start fresh
- Treat it like editing a draft with a smart collaborator
Come back with your phone or laptop open — Hour 2 is hands-on.
to Real Insight
Most people only think of three output formats.
💡 You don't need to know how to build an HTML file. You just need to know to ask for one.
Most nonprofits are sitting on program data, donor records, survey results, and evaluation reports they never fully analyze — because it used to require Excel expertise or a data consultant. That changes today.
Volunteer Management
Your organization runs a summer youth program. You need 10–15 new volunteers for June and July — weekday afternoons, no experience required. Using the Role/Task/Format formula, write a volunteer recruitment post for Facebook or your email newsletter.
💡 Tip: Swap in your own org's name, program, and audience if you'd like. Make it real.
🔁 If time allows: After your first result, push back on one thing — tone, length, or call to action. See what happens.
"You are a nonprofit volunteer coordinator. Write a Facebook post recruiting summer volunteers for a youth mentoring program in Gainesville, FL — weekday afternoons in June and July, no experience required. Warm, conversational tone. Under 120 words. End with a call to action linking to our signup form."
Yourself Every Time
AI is starting to remember you — but not the way you need it to.
Some tools now have memory features that pick up details over time. Useful — but passive memory isn't the same as a deliberate context document.
| Passive AI memory | Your org brief |
|---|---|
| Picks up fragments over time | You define exactly what it knows |
| Per-user — not shared | One document, your whole team |
| Can misremember or get things wrong | You wrote it — it's accurate |
| Locked to one tool | Works in any AI, anywhere |
You don't want AI to slowly figure out who you are. You want it to know — completely, correctly, from the first word — every time.
Your org brief: context you build once, use forever.
A one-page document that tells AI everything it needs to know about your organization. Load it as your starting context and every AI conversation begins with a fully informed collaborator.
Claude Projects
- Create a Project for your org in Claude.ai
- Paste your brief as "Project Instructions"
- Every conversation in the project uses it automatically
Custom GPTs (ChatGPT)
- Go to Explore GPTs → Create
- Paste your brief into the system prompt
- Save and use your org's own custom AI
💡 Simplest option today: paste your brief at the top of any new conversation — works in any tool, no setup required.
Open ChatGPT or Claude → paste this prompt → fill in your answers → AI builds your brief.
Help me build an org brief — a context document I'll paste at the start of any AI conversation so you always know who I am and how to communicate for my organization. I'll fill in my answers below. Compile everything into a clean, formatted org brief I can copy and save.
Mission (org name + what you do in one sentence):
Primary programs (2–3 bullets):
Who we serve (age, geography, circumstances):
Our voice and tone (3 adjectives + one example sentence):
What we'd never say (tone or values that would feel wrong):
Common AI tasks (3–4 things we'd ask AI to help with):
Fill in as much or as little as you know right now. The more detail you give, the more it sounds like you — but even a partial brief is worth building.
Now use it.
Paste your brief at the top of a new conversation and give it the same volunteer recruitment task from Exercise 1. Compare the results.
- Open a fresh conversation in ChatGPT or Claude
- Paste your org brief at the very top — "Here's context about my organization:"
- Then give your task: write a volunteer recruitment post for [your program]
- Notice the difference in tone, specificity, and how it "sounds like you"
📂 Next step at home: load your brief into a Claude Project or Custom GPT so it's automatic every time.
Here's what you built today.
🎯 Your Org Brief
- A context document that makes AI sound like your organization
- Reusable in any AI tool, any time
- Ready to share with your whole team
- Load it into a Claude Project or Custom GPT this week
🧠 Skills You Have Now
- You know the difference between a chatbot and a reasoning model
- You can write a Role/Task/Format prompt
- You know to ask for HTML output
- You have a policy framework for your org
Monday morning action plan
- Check what you already have. Ask your IT contact: "Are we enrolled in Google for Nonprofits? What Microsoft licenses do we have?" You may have powerful AI tools you haven't turned on yet.
- Load your org brief. Paste what you built today into a Claude Project or a Custom GPT. Takes 10 minutes. From now on, AI already knows your organization.
- Try one real task. Pick something on your actual to-do list this week — a message, a summary, a report. Use the Role/Task/Format formula. See what comes back.