Workshop companion

Three practical ways to put AI to work.

Choose a demonstration, copy the sample information and prompt, and try it in the AI tool you prefer.

Start with the work—not the technology.

These examples use fictional business information, so they are safe to explore. When using your own material, remove sensitive information and follow your organization’s policies.

← Return to all workshop resources

Demonstration 1

1. Find one useful AI opportunity

A small business owner knows that work is slipping through the cracks but does not know where AI would genuinely help.

What this demonstrates

AI can examine a business process, identify recurring friction, and suggest small experiments while keeping people responsible for decisions.

Fictional sample input

Gulf Coast Home Services is a fictional 12-person company serving Baldwin County. The team schedules estimates by phone and email, receives website inquiries, prepares estimates from technicians’ notes, sends appointment reminders, follows up on unanswered quotes, and answers repeated questions about service areas, pricing ranges, and appointment windows. The owner reviews all quotes and customer-facing messages. The company uses Google Workspace and a basic scheduling system.

Prompt

Act as a practical AI workflow advisor for a small business.

Review the business information below and identify five recurring tasks where AI could reduce time or improve consistency without removing human judgment.

Rank the opportunities by ease of implementation and potential value.

For each opportunity, provide:

1. The recurring task
2. The current friction
3. How AI might assist
4. What information AI would need
5. Where a person must review or approve the work
6. One small experiment we could run this week

Do not recommend replacing employees. Do not automate decisions involving money, safety, legal commitments, hiring, or customer promises.

Before providing the final opportunity map, ask me up to three useful clarifying questions.

BUSINESS INFORMATION:

[Paste a business description or public website address here]

Try in ChatGPT Try in Claude

What to notice

Does the AI begin with useful questions?
Are the recommendations specific to the business?
Does each recommendation contain a human review point?
Which suggestion would be easiest to test this week?

Make it your own

Replace the fictional profile with a description of your organization or a public website address. Do not include confidential information.

Demonstration 2

2. Turn rough information into a useful reply

A customer’s request has changed several times, and the employee needs to respond quickly without promising something the business has not confirmed.

What this demonstrates

AI can organize incomplete information, identify missing decisions, and prepare a first draft for a person to verify.

Fictional sample input

CUSTOMER MESSAGE

Hi, this is Linda. We have 60 chairs scheduled for delivery Saturday afternoon, but the weather looks bad. Can you bring them Friday instead? We may move from the beach pavilion to the Orange Beach Community Center. We might also need 80 chairs. Please let me know.

INTERNAL NOTES

- Friday delivery may be possible after 3:00 p.m.
- The truck can carry 80 chairs.
- The price changes if the chair count changes.
- The venue must be confirmed before the route is updated.
- Do not promise Friday delivery until Marcus checks the schedule.

Prompt

Help me prepare a customer reply using only the customer message and internal notes below.

First, create a section called “Questions before sending” that lists every fact, decision, or approval that is still missing.

Then draft a warm, concise customer email that:

- Confirms only what is already known
- Clearly identifies what still needs confirmation
- Asks only the necessary questions
- Does not invent a price, delivery time, availability, policy, or promise
- Uses plain, professional language

Finish with a section called “Human check” listing what an employee must verify before sending the email.

CUSTOMER MESSAGE AND INTERNAL NOTES:

[Paste the information here]

Try in ChatGPT Try in Claude

What to notice

Did the AI distinguish facts from possibilities?
Did it avoid making an unauthorized promise?
Did it identify the missing decisions?
Does the draft sound human and useful?

Make it your own

Try this with a nonsensitive customer message, rough notes, or a reply you have been postponing.

Demonstration 3

3. Turn meeting notes into action

A team has a useful conversation, but the decisions and next steps remain buried inside rough notes.

What this demonstrates

AI can transform unstructured notes into a summary, action list, and repeatable process without inventing owners or deadlines.

Fictional sample input

WEEKLY TEAM HUDDLE

- Website inquiries sometimes wait two business days for a reply.
- Amanda checks the form each morning and afternoon, but no one covers it when she is away.
- Luis prepares estimates but needs photographs and the service address before he can price the work.
- The goal is to acknowledge every inquiry within one business day.
- The team may test an automatic acknowledgment, but a person must approve all estimates and customer commitments.
- Ryan will draft standard answers to five common questions by Friday.
- Amanda will map the current inquiry process by Thursday.
- Next week, the team will decide whether to test the process with new inquiries.

Prompt

Organize the meeting notes below into a practical follow-up document.

Create:

1. A concise meeting summary
2. A list of confirmed decisions
3. An action table with task, owner, deadline, and status
4. A list of unanswered questions
5. A simple repeatable checklist for handling this process in the future
6. The three most important next actions

Do not invent an owner, deadline, policy, decision, or fact. Use “Unassigned,” “TBD,” or “Not decided” when the notes do not provide the answer.

Finish with a section called “Human check” identifying anything the team should verify before acting.

MEETING NOTES:

[Paste the notes here]

Try in ChatGPT Try in Claude Explore your sources in NotebookLM

What to notice

Are explicit decisions separated from assumptions?
Are owners and deadlines preserved accurately?
Does the checklist make the work easier to repeat?
What remains unanswered?

Make it your own

For a larger collection of meeting notes, policies, procedures, or reference documents, add the materials as sources in NotebookLM and ask it to answer using only those sources.

Optional NotebookLM prompt

Using only the selected sources, create:

1. A concise summary
2. Confirmed decisions
3. An action table with owners and deadlines
4. Unanswered questions
5. A repeatable checklist
6. Statements that may require verification

Cite the supporting source for each major conclusion. If the sources do not provide an answer, say so plainly rather than inferring one.

Recommended tools

Choose the tool that fits the work

ChatGPT A general-purpose conversational workspace for drafting, analysis, files, and iterative problem-solving.Claude A conversational workspace well suited to careful drafting, analysis, and working through substantial context.NotebookLM A source-grounded workspace for exploring documents, notes, websites, and other materials you provide.

Before sharing sensitive business information, review your organization’s policies and the privacy and data settings for the account you are using.

Complimentary AI Readiness Discovery

Bring the task that keeps following you home.

In 30 minutes, we’ll look at the work, identify where AI may help, and choose one practical first experiment—with human judgment kept in control.

Find a time