How to Automate Customer Payment Reminders With AI

· DoubleX

Quick answer

To automate payment reminders with AI, connect your receivables spreadsheet and email to an agent, describe your collection schedule in plain words, and run it daily. The agent flags overdue invoices and drafts emails, and you approve each one before it goes out. DoubleX Agentes supports this workflow.

Chasing late payments is repetitive work: open the spreadsheet, find who is overdue, check the history, write a polite email, repeat tomorrow. An AI agent can do the reading, sorting, and drafting while you keep the final say. This guide shows how to build a collection schedule (a series of reminders based on days overdue) that runs every day and never sends anything without your approval.

What information does the agent need to collect payments?#

The agent needs one row per invoice with enough data to decide who to contact and what to say. A collection agent is an AI assistant that reads your receivables data and prepares reminders based on rules you define. If your data is incomplete, it will ask or skip the row instead of guessing.

A minimal receivables spreadsheet should include these columns:

ColumnWhy the agent needs it
Customer namePersonalizes the message
Contact emailWhere the reminder is addressed
Invoice numberReferenced in the email
AmountShown in the message
Due dateUsed to calculate days overdue
Status (open, paid, disputed)Prevents reminders on paid or disputed invoices
Last reminder sentAvoids repeating the same message
Payment link or bank detailsMakes paying easy

How do you connect the receivables spreadsheet and email?#

You connect each system once through Connectors, then the agent can use them. In DoubleX Agentes, the steps look like this:

  1. Create your account (free, no card required) and take the diagnostic, a 5 to 10 minute conversation about your routine, or create a new agent in Agents.
  2. Open Connectors and connect Google Sheets (or Airtable) for the receivables data.
  3. Connect Gmail or Outlook. A personal connection, such as your own mailbox, is valid only for you.
  4. In the agent's settings, choose which systems it can use and which actions are allowed. Allow the spreadsheet to be read, and allow Gmail to create drafts only, not send.
  5. Ask the agent to run the collection routine once, manually, to check the result before scheduling it.

For more on how connections work, see the Connectors documentation.

How do you describe your collection schedule in your own words?#

You write the schedule as you would explain it to a new colleague. There is no need to write code or a formal prompt. A clear description covers the stages, the days, the tone, and the exceptions.

For example:

  • Before the due date: a friendly heads-up 3 days ahead.
  • 1 to 7 days overdue: a polite reminder with the invoice and payment link.
  • 8 to 20 days overdue: a firmer message asking for a payment date.
  • More than 20 days overdue: do not email; list the account for me to call.
  • Never contact customers marked as disputed.

If you want the schedule to be reusable, save it as a Skill, a reusable set of instructions that can be attached to any agent. Details are in the Skills documentation. You can also store your standard wording and payment terms in the Brain, the company knowledge base the agent consults.

How do you schedule the daily routine?#

You schedule it with a trigger, which makes the agent work on its own at a set time. A common choice is every weekday at 8 a.m. The routine runs with the permissions of whoever created it, so the same limits you set for the agent still apply.

Each morning, the agent reads the sheet, calculates days overdue, picks the right stage for each invoice, and prepares drafts. You get a list of what it found and what it prepared. See the Triggers documentation for the available options.

Examples of requests you can copy#

  • "Read the Receivables sheet and list every invoice overdue by more than 7 days, grouped by customer, with the total owed."
  • "For each invoice 1 to 7 days overdue, create a Gmail draft with a polite reminder that includes the invoice number, amount, and due date."
  • "Skip any row marked Disputed or Paid, and tell me how many rows you skipped and why."
  • "Every weekday at 8 a.m., run the collection schedule and send me a summary of the drafts waiting for my approval."
  • "Rewrite this reminder in a warmer tone for a customer who has always paid on time."

How do you keep the tone and approval before sending?#

You keep control by limiting what the agent is allowed to do and by reviewing each message. Before sending, deleting, or changing anything in your systems, the agent asks for confirmation and shows the steps it took. Every request is saved as a task with the full history: what was asked, which tools were used, and the result.

To keep the tone consistent:

  • Put your preferred wording and sign-off in the instructions or in a Skill.
  • Set Gmail to create drafts only, so nothing leaves without you.
  • Keep spreadsheet access read-only if the agent does not need to update the "last reminder" column.
  • Ask the agent to flag unusual cases, such as large amounts or long-standing customers, for a personal message.

For related finance workflows, see AI Agents for Finance Teams: What to Automate First and How to Automate a Weekly Cash Flow Report With AI.

Frequently asked questions#

Can the agent send reminders without my approval?#

It does not have to. You choose which actions the agent can perform, and it asks for confirmation before sending. Limiting Gmail to drafts keeps every email in your hands until you approve it.

Do I need to write code or a prompt?#

No. You describe the collection schedule in your own words and the system builds the agent. For developers, there are also a REST API, a CLI, and an MCP server.

What if my receivables live in Airtable or a database instead of Google Sheets?#

Airtable, PostgreSQL, and MySQL are available as connectors. Databases connect in read-only mode by default, which suits a reminder routine well.

Is my customer data used to train AI models?#

No. Company content is not used to train models, and requests only go to AI providers that do not use what they receive for training. Each company also has its own separate database.

Can I see what the agent did on each run?#

Yes. Every run is recorded as a task with the request, each tool the agent used, and the result, so you can check who asked for what and when.

Related guides

More Finance guides