As of 13 August 2026, AI can only partly create a custom CRM for your small business.

What to actually do

1. Hand it to a person

   The route this page recommends

   Someone with a licence or accountable authority has to sign this before it counts.

2. Use a tool built for this

   Second choice

3. Do it yourself

   The distant third

   A chat interface, power-user skill, and roughly 2 hours until you can act on the result.

   How to actually do it

   1. Open a spreadsheet and list your current customer, lead, opportunity, task and note fields, marking which fields are required and which contain personal data.
   2. Write down the real workflow from first enquiry to completed sale or follow-up, including every user role, approval, reminder and exception.
   3. Collect the details of any email, calendar, website form, accounting or other system that the CRM must connect to, including what data should move in each direction.
   4. Paste the requirements into Lovable or another AI development tool using the prompt, and answer its clarification questions without entering real customer records.
   5. Create sample records covering a new lead, an existing customer, a duplicate contact, a closed opportunity, a failed reminder and each user role, then run the generated tests and manual workflows.
   6. Compare the fields, permissions, reports, exports and automations against your spreadsheet and written process, correcting each mismatch before considering a pilot.
   7. Ask a developer or security specialist to check authentication, permissions, exposed endpoints, backups, deletion, audit logs, data retention and integrations before importing real data.
   8. Run a limited pilot with copied or anonymised data, compare the results with your existing records, and only then plan migration and staff training.

   Prompt

   Create a small-business CRM for [business type] using the requirements below. Build a working full-stack application with a clear, simple interface and a relational data model.
   
   Business process:
   [describe how a lead becomes a customer and what happens afterwards]
   
   Users and permissions:
   [list each user role and exactly what that role may view, create, edit, export or delete]
   
   Records and fields:
   [list the records you need, such as contacts, companies, leads, opportunities, tasks and notes, with the fields and required or optional status for each]
   
   Workflows and reminders:
   [list status changes, notifications, follow-up tasks and any rules that should run automatically]
   
   Reports and searches:
   [list the reports, filters, saved views and exports required]
   
   Integrations:
   [list any email, calendar, forms, accounting or other systems, or write none]
   
   Build rules:
   - Use sample data only while building. Do not request or include real customer data.
   - Keep the interface usable on a desktop and mobile screen.
   - Validate required fields and prevent duplicate contacts where practical.
   - Record who changed important records and when.
   - Include authentication, role-based access, input validation and safe error handling.
   - Do not claim that security, backups, privacy compliance or integrations are complete unless they are implemented and tested.
   - Add a clear way to export the data and document how backups would be configured.
   - Show the database schema, assumptions, known limitations and any manual setup required.
   - Create automated tests for the main create, edit, search, permission and export workflows.
   
   Before building, list ambiguities and ask only the questions that block a safe first version. After building, give me a test plan using sample records and identify which checks require a developer or security specialist.

   Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.