Plumbing Invoice Follow-Up: A 14-Day Collection Workflow

Plumbing Invoice Follow-Up: A 14-Day Collection Workflow

Plumbing invoice follow-up should start before a balance becomes an awkward phone call. The best workflow verifies that the invoice is correct and delivered, gives the customer a simple way to pay or ask a question, records every response, and stops the moment the balance is resolved.

An AI employee can own that routine from the invoice event to the final status. It can watch the billing system, check the job record, send approved reminders, answer factual questions, resend documents, update the customer record, recognize a payment or dispute, and keep the owner out of routine chasing.

The owner gets involved when the company needs to make a real business decision: approve a credit, change terms, handle a disputed scope, write off a balance, or move an account into formal collection. That exception should arrive as a short brief with the invoice, job history, contact attempts, customer response, and exact decision needed.

This guide gives you a 14-day workflow, five customer messages, an exception format, a daily scorecard, and a test plan.

Why unpaid invoices turn into owner work

The usual problem is not that the company forgot to create an invoice. The invoice exists. The job may even show as completed and profitable in a report. Cash still has not arrived.

The U.S. Small Business Administration's guide to managing business finances explains that accrual accounting can record a sale before payment lands. It also lists accounts receivable and available cash among the financial functions a business needs someone to manage. Booked revenue and usable cash are different facts.

That gap becomes painful in a plumbing company because the next payroll, supply-house bill, fuel charge, and equipment payment do not wait for a homeowner or property manager to find an old email.

Cash discipline matters while many small firms already face financing pressure. The Federal Reserve Banks' 2025 Small Business Credit Survey reached more than 6,500 small employer firms. About half had their funding needs met, about one-third faced a gap despite applying for financing, and the rest needed financing but did not apply. The survey does not say unpaid plumbing invoices caused those gaps. It does show why an owner cannot treat accounts receivable as harmless paperwork.

The manual process usually breaks in predictable places:

• the invoice went to the wrong email address;
• the customer never received the payment link;
• a property manager needs a purchase order or separate recipient;
• the invoice does not match the approved estimate;
• a technician note or signature is missing;
• a promised discount was never applied;
• the payment posted in one system but not another;
• a customer replied to an inbox nobody watches;
• staff do not know whether to resend, call, pause, or escalate;
• the owner sees the balance only after it has aged for weeks.

A reminder alone does not fix those failures. It can make them worse by asking a customer to pay a bill that is wrong, already paid, or missing the document their office requires.

Define what "recovered" means

An invoice is not recovered because an email was sent.

Call the workflow complete only when the AI employee records one of these outcomes:

1. paid in full and reconciled to the correct invoice;
2. approved payment arrangement recorded with the next due date;
3. valid administrative issue fixed and the corrected invoice delivered;
4. customer dispute packaged for an authorized decision;
5. approved credit, write-off, or cancellation completed;
6. formal collection or legal review accepted by the authorized owner;
7. documented stop under company policy.

Every open invoice should also have a next action, owner, and due time. "Waiting" is not a useful status unless the record says who is expected to do what and when.

Give the AI employee the whole routine job

The AI employee's trigger is a finalized invoice or an invoice that enters a defined aging stage. Its job ends when the balance has a recorded outcome and every connected system agrees on that outcome.

A capable invoice-recovery employee can:

1. Read the invoice, customer, job, estimate, payment, and communication records it is authorized to access.
2. Confirm the invoice has a number, issue date, due date, customer name, service address, line items, total, payment instructions, and delivery destination.
3. Check for an existing payment, credit, dispute, returned message, duplicate invoice, or open correction before contacting the customer.
4. Deliver or resend the invoice through the approved channel.
5. Ask the customer whether they need the invoice resent, a receipt, a job reference, a purchase order field, or another billing contact.
6. Answer approved factual questions from the job and invoice records without inventing what happened on site.
7. Create and monitor the next follow-up based on the due date, customer type, balance band, and company policy.
8. Stop reminders when payment, a valid dispute, an opt-out rule, or another stop condition appears.
9. Update the field-service platform, accounting system, CRM, and recovery board within its assigned permissions.
10. Send a decision-ready exception only when routine policy cannot resolve the account.

That is the difference between an AI employee and an automatic email sequence. The sequence sends on a timer. The employee checks the situation, communicates, fixes routine administrative problems, and owns the result.

For the broader model, read what an AI employee is.

Build one invoice-recovery record

Do not make the employee reconstruct the account from five browser tabs every morning. Create one recovery record for each invoice.

Include:

• invoice ID and direct link;
• customer and billing contact;
• job ID, service address, and completion date;
• original estimate or authorization link when applicable;
• invoice date, due date, original amount, credits, payments, and open balance;
• invoice delivery channel, recipient, and delivery status;
• payment link or approved payment instructions;
• customer type, such as homeowner, property manager, commercial account, or warranty administrator;
• required purchase-order, work-order, signature, tax, or supporting-document fields;
• last customer message and timestamp;
• current status;
• next action, owner, and due time;
• dispute, complaint, hardship, legal, or safety flags;
• every reminder and system change with timestamps;
• final outcome and close date.

Preserve the original invoice and customer messages. A summary helps the owner make a decision, but it should not replace the source record.

The 14-day plumbing invoice follow-up workflow

The sample below uses the invoice due date as Day 0. Change the timing to match your contracts, customer types, local requirements, and normal payment cycle. A commercial property manager on net-30 terms should not receive the same sequence as a residential customer whose payment was due at job completion.

Before Day 0: invoice delivery check

The AI employee verifies the invoice before the first payment request.

It should confirm:

• the job is actually ready to invoice;
• the invoice total matches the approved record available to the workflow;
• required credits, deposits, and prior payments appear;
• the billing contact and delivery channel are current;
• supporting documents required by that account are attached or linked;
• the payment method works;
• no open complaint, disputed scope, or correction already exists.

If a routine field is missing, the employee fixes it when policy and system permissions allow. If the invoice conflicts with the approved estimate or requires a commercial decision, it pauses that invoice and sends an exception brief. Other invoices continue normally.

Delivery message:

"Hi [name], this is [company]. Invoice [number] for [service address or job reference] is ready. The balance is [amount], due [date]. You can review and pay here: [secure link]. If your billing office needs a different contact or document, reply here and I will update the record."

The employee logs delivery evidence. A message marked "sent" is not the same as a message delivered.

Day 0: due-date check

On the due date, the AI employee checks payment status before sending anything.

If paid, it closes the recovery record and sends an approved receipt when required. If the payment is pending, it waits under the company's payment-processing rule. If the balance remains open, it sends a short factual reminder.

"Hi [name], a quick billing note from [company]: invoice [number] for [amount] is due today. Here is the secure payment link: [link]. If you already paid or something on the invoice needs attention, reply with the details and I will check the record."

A reply creates work for the AI employee, not a new inbox task for the office. It checks the record, answers approved factual questions, resends documents, corrects authorized contact fields, or routes a genuine dispute.

Day 3 overdue: identify the blocker

The third-day touch should find out why payment has not moved.

"Hi [name], invoice [number] still shows an open balance of [amount]. Is the blocker the invoice, the payment link, a missing document, or something about the completed work? Reply with what you need and I will route it correctly. Payment link: [link]."

The response should move into one of these paths:

• ready_to_pay: provide the secure link and monitor the payment event;
• needs_invoice_copy: resend the approved document and confirm delivery;
• wrong_billing_contact: update the authorized billing recipient and redeliver;
• missing_account_document: collect or attach the approved job, purchase-order, signature, or tax record;
• payment_already_sent: collect the date and reference, then check the payment system;
• factual_question: answer from approved records and log the source;
• scope_or_quality_dispute: pause collection messaging for that invoice and prepare an exception;
• payment_arrangement_request: gather the requested timing and route it under company authority limits;
• cannot_identify_customer_or_job: block the send and fix the data problem first.

This is where the workflow earns its keep. It resolves the small blockers that make a generic sequence keep repeating itself.

Day 7 overdue: state the status and next step

By Day 7, the message should be direct without becoming hostile.

"Hi [name], invoice [number] has an open balance of [amount] and is now 7 days past due. I have not recorded a payment or an open invoice issue. Please pay here: [link], or reply today if the balance or job record needs review."

If company policy allows a specific next step, state it accurately. Do not threaten a late fee, service hold, collection action, lien, report, or lawsuit unless the contract, law, and authorized company decision support that exact action.

The AI employee can apply a preapproved account hold within its assigned authority, but it should not improvise a consequence to make the message sound stronger.

Day 14 overdue: prepare the owner decision

At Day 14, routine messaging may continue only if company policy says it should. The important output is a decision-ready account brief.

Use this format:

• Account: customer, invoice, job, service address, and direct links.
• Balance: original amount, credits, payments, fees authorized by the record, and current open balance.
• Delivery evidence: recipients, channels, delivery status, and dates.
• Work completed: reminders, documents resent, contact corrections, and factual questions answered.
• Customer position: exact response or "no response" with the last contact time.
• Dispute status: issue raised, source evidence, and affected amount.
• Company policy: the allowed next actions for this customer and balance band.
• Options: continue routine follow-up, approve terms, approve a credit, place an account hold, write off, or send for qualified collection or legal review.
• Decision needed: one specific choice and its deadline.

The owner should be able to decide without opening the field-service platform, accounting system, email, and text thread.

After the decision, the AI employee carries it out within its permissions, tells the customer what the company actually decided, updates every record, and monitors the next event.

Handle disputes without losing the account history

A customer who says "I don't owe this" should not keep receiving the same reminder while the office debates what happened.

The AI employee should:

1. Pause collection messages on the disputed amount.
2. Preserve the customer's exact words and timestamp.
3. Gather the approved estimate, change authorization, technician notes, photos, signatures, prior credits, invoice, and communication record available in connected systems.
4. Separate factual record questions from decisions about workmanship, scope, discounts, liability, or legal position.
5. Send the authorized person a short brief with the disputed amount and exact decision needed.
6. Continue unrelated work, such as confirming the correct billing contact or reconciling an undisputed payment.
7. Record the decision and communicate it using approved language.

The AI employee should not invent a technician explanation, admit fault, deny responsibility, negotiate beyond its limit, or calculate a legal deadline. Those are business, technical, or legal decisions for the authorized person.

Set collection rules before the employee goes live

Invoice follow-up touches contracts, customer communications, fees, credit decisions, and sometimes formal debt collection. The legal boundary depends on the debt, the company, the collector, and the jurisdiction.

The Consumer Financial Protection Bureau's explanation of federal debt-collection limits says the Fair Debt Collection Practices Act covers debts primarily for personal, family, or household purposes. It generally does not cover business debts or collection by the original creditor. The CFPB also notes that state laws can provide broader protections and may cover original creditors. The Federal Trade Commission's Debt Collection FAQs likewise state that business debts are outside the FDCPA.

Do not turn that distinction into "no rules apply." Have qualified counsel review your customer types, contracts, communication channels, timing, fees, dispute handling, account holds, third-party collection process, liens, and state or local requirements.

Give the AI employee written policy for:

• which invoices enter the workflow;
• timing by customer and contract type;
• approved channels and contact hours;
• verified payment links and data-handling rules;
• late fees and when they may appear;
• discount, credit, and payment-plan authority;
• dispute stop conditions;
• account-hold authority;
• formal collection and legal-review thresholds;
• opt-outs and channel preferences;
• required records and retention;
• who can approve each exception.

These policies should sit behind the workflow. Customers should experience a capable billing employee, not a lecture about guardrails.

Connect invoice events to the workflow

The workflow needs reliable events from the systems that hold the invoice and payment status.

At minimum, identify how the AI employee learns that an invoice was:

• created or finalized;
• delivered or returned;
• viewed when that event is available;
• partially paid or paid in full;
• credited, voided, or marked uncollectible;
• changed after delivery;
• disputed by the customer.

Stripe's documentation for invoice emails and reminders is one example. It supports reminders for unpaid invoices and webhooks for invoice events such as finalized, paid, or marked uncollectible. Your plumbing company may use a field-service platform, accounting package, payment processor, or a combination. Verify the real events and permissions in your own accounts before designing around them.

Do not let two systems send overlapping reminders. Choose one owner for customer communication and make every other system feed status into that workflow.

Copy this daily owner scorecard

The owner should see exceptions and movement, not a dump of every message.

Use these fields:

• total open invoice balance;
• invoices newly due today;
• invoices 1 to 6 days overdue;
• invoices 7 to 13 days overdue;
• invoices 14 or more days overdue;
• payments posted since the prior report;
• balances resolved through corrected billing records;
• active disputes and disputed amount;
• payment arrangements due or missed;
• undelivered invoices;
• accounts with no valid billing contact;
• decisions needed today;
• records with no next action or owner.

Track the workflow itself with:

• percentage of invoices with verified delivery;
• percentage of open invoices with a next action and due time;
• median time to classify a customer reply;
• number of duplicate or incorrect reminders prevented;
• dollars paid, corrected, credited, disputed, written off, or moved to an approved next stage;
• exception rate by reason;
• unreconciled payment events.

Do not promise a recovery percentage before you have your own baseline. Measure thirty to sixty days of real outcomes, then improve the rule or message causing the most avoidable exceptions.

Test the workflow with controlled invoices

The National Institute of Standards and Technology describes its AI Risk Management Framework as a voluntary tool for adding trustworthiness considerations to AI design, use, and evaluation. For invoice recovery, that means permissions, test cases, logs, exception rules, and an authorized override.

Test at least these paths before live customer contact:

1. correct invoice delivered to the right customer;
2. invalid email address;
3. duplicate customer and duplicate invoice;
4. payment posted just before a scheduled reminder;
5. partial payment;
6. credit applied after invoice delivery;
7. wrong billing contact;
8. property manager who requires a purchase order;
9. customer who says payment was already sent;
10. customer who disputes the total;
11. open workmanship complaint;
12. payment-plan request outside the employee's authority;
13. customer who asks for a different channel;
14. invoice changed after a reminder was prepared;
15. disconnected accounting or payment system;
16. two systems attempting to send the same reminder.

For every case, verify the message, recipient, amount, link, timing, system updates, stop conditions, and audit log. A good draft message is not enough. The full account must end in the right state.

Common mistakes

Starting after the invoice is already old

The workflow should verify delivery and billing details when the invoice is created. Waiting until Day 30 preserves every upstream error.

Sending before checking payment

A payment may land minutes before the scheduled reminder. Read the current payment state immediately before sending.

Treating every customer the same

Residential, commercial, warranty, property-management, and construction accounts can have different terms and document requirements. Route by the actual account policy.

Escalating every reply to the office

The employee should handle document resends, contact corrections, payment-link problems, and approved factual questions. Escalate only the decision it cannot make.

Letting two systems contact the customer

Accounting software and a field-service platform may both have reminder features. Duplicate messages make the company look disorganized and can create compliance problems.

Using pressure instead of facts

State the invoice, amount, due date, secure payment route, and next authorized step. Never invent urgency or threaten action the company has not approved.

Measuring sends instead of outcomes

The goal is not a high reminder count. The goal is an accurate balance with a recorded outcome and no unnecessary owner work.

Put one week of invoices through the map

Pull every invoice finalized last week. For each one, mark the delivery evidence, balance, customer type, required documents, last response, current status, next action, and decision owner.

You will quickly see whether the problem is customer follow-up, invoice accuracy, billing contacts, payment reconciliation, or missing account rules.

ComfortGrowth AI builds AI employee workflows for plumbing companies, including invoicing and collections operations that connect to the systems a company already uses. Our after-hours plumbing missed-call workflow shows the same ownership model on the front end of the customer journey.

Request a free business audit and we will map one unpaid-invoice path from finalized invoice to recorded outcome, including the routine work the AI employee can own and the few decisions that should stay with you.