Building Your First AI Strategy: A Step-by-Step Guide

You know you need an AI strategy. But where do you start? This lecture walks you through building one from scratch.

Step 1: Identify Problems (Not Solutions)

Don't start with "let's use ChatGPT." Start with "what's hard right now?"

Problems nonprofits face:

• Fundraising emails take hours to write. Writers block. Copy often weak.
• Donor data is scattered. Hard to segment for targeted asks.
• Grant research takes forever. Someone spends weeks finding relevant grants.
• Program photos aren't organized. Hard to pull impact stories quickly.
• Manual data entry. Volunteer hours, donation amounts, program participation.
• Reporting is tedious. Compiling impact numbers from multiple systems.

Pick your top three problems. Write them down.

Step 2: Evaluate Which Problems AI Can Solve

Not all problems need AI. Some need better processes, better tools, better data.

AI is good at: writing, summarization, pattern recognition, categorization, prediction, image analysis.

AI is bad at: solving structural problems (if your CRM is broken, AI won't fix it), subjective judgment calls, decisions requiring context.

Example: "Donor data is scattered" isn't an AI problem. It's a data problem. Fix your CRM first. Then AI can help analyze it.

Example: "Writing fundraising emails is slow" IS an AI problem. ChatGPT can draft emails. Your team refines them. Saves time.

Step 3: Size the Opportunity

For problems AI can solve, estimate impact:

Fundraising emails: currently take 4 hours per appeal. AI drafts in 30 minutes (your team edits). Saves 3.5 hours per appeal. You send 12 appeals per year. Saves 42 hours per year. That's one month of work.

Grant research: currently takes 20 hours per grant cycle. AI accelerates to 10 hours (AI finds candidates, you review). Saves 10 hours per cycle. Do this 4 times per year. Saves 40 hours per year.

Quantify impact: money saved (staff hours × hourly rate), quality improvement (better grant targeting = higher award rate), speed (faster decisions).

Step 4: Build Your Use Cases

For each problem AI solves, write a use case:

Use Case 1: Email Writing

• Problem: Fundraising emails take hours
• AI solution: ChatGPT drafts emails from brief outline
• Tools: ChatGPT (free or $20/month), or integrate Claude API
• Workflow: Development director writes brief (donor segment, ask amount, key message). Pastes into ChatGPT. Gets draft. Edits 15%. Sends.
• Impact: 3.5 hours saved per appeal
• Risk: Quality varies. Output needs editing. Requires brand voice training.
• Success metrics: emails sent faster, higher open/click rates, team satisfaction

Write similar one-pagers for each use case. They become your strategy.

Step 5: Prioritize

Start with highest impact, lowest risk use cases. Not the most exciting ones.

Matrix:

Step 6: Pilot the First Use Case

Don't implement all at once. Pilot one use case for 3 months. (See 5.4.5 Pilot Project Framework.)

Learn: what works, what doesn't, what needs refinement. Then scale.

Step 7: Document Your Strategy

Write a one-page AI strategy:

• Vision: "We want AI to amplify our team's capacity, not replace them."
• Use cases (prioritized)
• Timeline: "Q2: Email writing pilot. Q3: Grant research pilot. Q4: Evaluate and scale."
• Budget: "Tools ($100-500/month). Staff time (10-15 hours/week for pilot)"
• Governance: "Executive Director approves all AI tools. Privacy/Security reviews new tools."
• Ethics: "We won't use AI to replace jobs. We'll use it to amplify capacity. All staff impacted gets retrained."

Share with board. Make sure they understand: this is about efficiency, not cutting staff.

Key Takeaway

A good AI strategy starts with honest assessment of problems, evaluates which ones AI can actually solve, sizes the opportunity, builds specific use cases, and pilots one before scaling. It's not about having the latest AI. It's about using AI to solve real problems.