Pennsylvania NAACP workshop · October 24, 2026

Stay Woke.
Stay Connected.

Using AI with Wisdom, Purpose, and Power

Facilitated by Pastor Darrell K. Cook · Participant resources

Understand enough to challenge the harms. Develop enough skill to use the opportunities. Our communities should help shape technology, not simply have it happen to us.

Download participant handout · PDF ↓

Save the handout before the session for use without Wi-Fi. You can also use your browser’s Print command to print or save this page as a PDF.

Return anytime: SBCImpactErie.com/ai

Our hour together

Learn. Question. Practice.

0–5 minWhy it matters. Connect technology, community power, and everyday work.
5–13 minFrom answers to assignments. Understand chatbots, assistants, and agents.
13–25 minRights and risks. Examine fairness, misinformation, privacy, and physical impacts.
25–37 minPractical opportunities and demo. Give a bounded assignment, inspect the draft, and revise it.
37–45 minPaired exercise. Design an assignment for one recurring branch task.
45–50 minFive supervision checks. Facts, Fairness, Privacy, Permission, Ownership.
50–58 minQuestions and answers. Bring your questions, concerns, and possible uses.
58–60 minOne responsible next step. Choose a task, a verification step, and a boundary.

From answers to assignments

What kind of help are you asking for?

Chatbot

Responds to questions and conversation. Try asking it to explain a public document in plain language, then check the explanation against the original.

Assistant

Helps with context, documents, and available tools. You might give it approved facts and ask for an event-plan draft.

Agent

Pursues a goal across multiple steps and tools, within the permissions it is given. Its work still needs boundaries, review, and a responsible human.

You remain accountable

These capabilities overlap; they are not a rigid replacement history. “Digital teammate” is a metaphor. A tool does not take ownership of your judgment or decisions.

A plan described in chat is not proof that a tool acted. If the demo only produces text, treat it as a workflow simulation, not actual agent execution. Verify any claimed action in the destination tool.

Rights and risks

Ask who benefits—and who bears the cost.

People and decisions

Bias and discrimination: challenge unfair outcomes in consequential decisions. Ask what evidence, human review, and ways to contest a decision are available.
Misinformation and deepfake fraud: confident text, an image, or a familiar-sounding voice is not verification. Check important claims with the original source or a known contact channel.
Privacy: use fictional or public data for practice. Keep confidential complaints, member lists, and children’s records out of prompts.

Data centers and our communities

Who pays for new infrastructure, and who pays if a large customer leaves?
What are the local water, power, noise, and pollution impacts?
Which jobs are temporary, and which are permanent?
How can residents influence decisions before approval?
Which promised community benefits are enforceable, and who checks delivery?

Use local evidence and project-specific questions. There is no single water-use estimate that describes every AI prompt or data center.

Read the NAACP Frontline Framework and NIST’s generative AI risk profile. Pennsylvania’s large-load tariff order and model guidance address cost allocation; they are not a promise of universal protection from higher bills.

Practical opportunities

Start with useful, reviewable work.

Turn rough, fictional meeting notes into an action list with owners and unresolved decisions.
Turn approved facts into event plans, invitation drafts, and checklists.
Turn public documents into sourced questions for a hearing. Ask for page or section references, then verify each one.

AI can widen access to capabilities that once required specialist staff. Judgment, community knowledge, and verification still matter.

Keep the first task low risk. Ask for a draft. Use fictional or public information. Review before anything is sent, published, purchased, or used to decide about a person.

Try it together

Copy. Review. Improve.

Use these prompts in the AI tool you choose. The workshop, budget, and people below are fictional. No real participant information is needed. If copying is unavailable, select the text manually.

1. Give a bounded assignment

I’m preparing a fictional, nonpartisan community digital-literacy workshop for 30 adults with mixed technology experience. We have 60 minutes, four volunteers, a $150 budget, and unreliable Wi-Fi. The venue and date are not confirmed. Create a timed agenda, volunteer assignments, a draft invitation, and a checklist of decisions still needed. Do not invent speakers, partners, confirmed logistics, or expenses. Mark assumptions clearly. Keep everything in draft form for my review.

2. Supervise and revise

Review those drafts for contradictions and missing information. Revise the plan so it works without a projector or internet. Identify what can be prepared now and what needs my decision. Do not send messages, publish anything, spend money, or collect participant information.

What to look for: Does the agenda fit 60 minutes? Are responsibilities clear? Did the tool invent a confirmed venue, a partner, or a price? Does the revision actually work offline? Keep what is useful and correct what is unsupported.

Paired exercise · 8 minutes

Turn one recurring task into an assignment.

Work with a partner. Choose a routine branch task you can describe using fictional or public information. Write an assignment using these five parts:

Job and purpose: What should the tool help accomplish, and why?
Audience: Who needs to understand or use the result?
Source facts: What verified information may it use? What is still unknown?
Constraints: What must it avoid, preserve, or ask you before doing?
Output: Do you need a table, checklist, plain-language draft, or set of sourced questions?

Swap assignments. Your partner identifies one missing fact and one boundary to make clearer. No internet? Write the assignment on paper and review it together.

Before you use the result

Five supervision checks

Facts: What claims, names, dates, numbers, and sources must I verify?
Fairness: Who might be excluded, misrepresented, or treated unfairly?
Privacy: Did I share information that should not be in this tool?
Permission: What may the tool do, and what needs my explicit approval?
Ownership: Who reviews, corrects, and takes responsibility for the final result?

Keep learning

Sources to return to

This is a growing resource page. Pastor Cook will continue adding practical tools, helpful sources, and additional learning content over time.

Read the original material, bring questions, and check how it applies to your community.

NAACP · Frontline Framework Community Guiding Principles ↗Community-centered questions and principles for data-center development.
NIST · Generative Artificial Intelligence Profile (PDF) ↗A detailed risk-management reference covering issues such as unreliable output, harmful bias, privacy, and information integrity.
Pennsylvania PUC · Large Load Customers ↗Final order and model tariff guidance on interconnection and large-load costs, including cost shifting and exit fees. Not universal bill protection.
Anthropic · Building effective agents ↗A technical explanation of workflows, agents, tools, and when simpler approaches are sufficient. Published in 2024; specific tooling has evolved.

Links checked October 24, 2026. These are learning resources, not an endorsement of a product or a claim that any organization endorses this page.

Your responsible next step

One task. One check. One boundary.

I will try: one low-risk, recurring task.

I will verify: one thing before I use the result.

I will protect: one boundary the tool may not cross.

Stay curious. Stay critical. Stay in charge.

Save the participant handout ↓