The free companion guide · Claude + Neon

Give your AI
a filing cabinet.

Better follow-ups start with better business records. Learn what to organize, then copy the prompt to build your first sample database.

Get the starter prompt ↓Download the PDF ↗

With Jamal Henry · For service-business owners · No signup needed

01 / The foundation

AI needs the details to do the job.

Before an assistant can help with an invoice or a customer follow-up, it needs the relevant records. A database gives that information a consistent structure your software can read and update.

Think of a filing cabinet: each drawer has a job, each record has a place, and related records connect. One customer can have several jobs. A job can have an invoice.

02 / Two kinds of information

The binder and the cabinet.

The binder

Your knowledge base

How your business operates: services, policies, tone of voice, and instructions. It explains how you want work done.

The cabinet

Your business database

What is happening: customers, jobs, invoices, and recorded AI actions. It supplies the details for a particular task.

Both matter. This is a helpful teaching distinction; they do not have to be separate products.

03 / Keep it simple

Start with four tables.

Customers

Names, contact details, and customer IDs.

Who is this customer, and how do we reach them?

Jobs

Work, dates, technician, and job status.

What have we done for this customer?

Invoices

Amounts, due dates, and payment status.

Which invoices still need attention?

AI action log

Actions your workflow records, when, and their results.

What did the assistant actually do?

These are a starting point. The action log only works when your workflow writes events to it.

04 / Put it into practice

Your business. Your starting prompt.

Copy this into Claude, then replace the example company, services, and job process with your own. Start in a separate demo project with fictional data and review the generated SQL before running it.

Claude + Neon starter prompt

Edit the example below before copying, if you like.

I own a residential plumbing company in Houston, TX. We do water heater installs and repairs, drain cleaning, repipes, faucet and fixture repair, and sewer line inspections. Customers reach us by phone, text, and a website form. A job goes: first contact, quoted, scheduled, done, invoiced, paid.

Build me a simple database that an AI assistant can read and write to. Give me exactly four tables:
• customers (who they are and how to reach them)
• jobs (what we did, when, which tech, for how much)
• invoices (amount, due date, paid or not)
• ai_action_log (every text, email, or call the AI makes, with the result)

For each table:
(1) Explain the columns in plain English, one line each.
(2) Give the exact SQL to create the table in Neon (Postgres) with sensible rules, like a date column that only accepts dates and a status column limited to scheduled, in_progress, done, or cancelled.
(3) Give me 8 rows of realistic fake Houston customers and matching jobs, invoices, and log entries as INSERT statements, with at least two unpaid invoices.

Then give me three questions I can ask the database this week as SQL:
• Who owes me money and how late are they?
• Who has not been serviced in over a year?
• What did the AI send out this week?

Keep every explanation at a 5th-grade reading level.

Download the prompt as a text file ↗

05 / Your first small win

Ask one useful business question.

Choose a task. Find unpaid invoices, customers due for a follow-up, or recent AI activity.
Check the structure. Ask how customers, jobs, invoices, and actions are linked. Review which validation rules were implemented.
Test with fictional records. Check whether the answers match the sample data. In the demo, AI activity is also sample data.
Plan the workflow. Decide which system owns each record and how updates should move between your database and CRM.

Rules help. They need design.

A database can enforce valid dates, allowed statuses, and relationships. These rules reduce specific errors; they do not guarantee every record or AI answer is correct. Spreadsheets can have validation too.

Keep the CRM you use.

A database can work alongside your CRM. Creating it does not automatically synchronize records or send messages. Follow-ups need their own triggers, permissions, and workflow.

Keep building with ComfortGrowth AI

The foundation comes first.

Follow the series for practical ways to turn organized business information into useful AI workflows.

Visit the YouTube channel ↗

Give your AIa filing cabinet.