The unpaid invoice problem is rarely one big mistake

Most agency owners don’t have a billing problem. They have a follow-up problem.

The work gets completed. The invoice goes out. The client relationship is generally good. Then the invoice reaches 30 days, 45 days, or 60 days because nobody has made the next call, sent the right email, or checked whether there was an issue buried in a client reply.

That delay puts pressure on the whole agency.

Payroll is still due every two weeks. Freelancers need to be paid. Software subscriptions keep hitting the bank account. Media spend may need to be funded before a client reimburses it. Meanwhile, account managers are focused on client delivery, finance teams are chasing numbers, and partners are stepping in when an overdue balance gets large enough to become uncomfortable.

For a marketing or creative agency doing $1 million to $25 million in annual revenue, we usually see cash leakage from slow invoicing, missed follow-up, disputed line items, and avoidable write-offs land somewhere in the $60,000 to $180,000 annual range.

That doesn’t mean all of it is permanently lost. Some is collected eventually. The real issue is that money arrives late, takes too much labour to recover, or gets discounted because the agency didn’t identify a problem early enough.

An AI invoice follow-up workflow doesn’t replace the client relationship. It makes sure your team protects it without relying on somebody’s memory, inbox discipline, or willingness to send a second reminder.

Why agencies let receivables drift

Agency invoicing has a few quirks that make normal collections processes fail.

First, invoices often relate to work that isn’t straightforward. A client may see a retainer fee, production costs, platform fees, project milestones, media management, or pass-through expenses on the same invoice. If they question one item, they may hold the whole payment. That question might sit in an account manager’s inbox while they prepare a campaign launch or respond to a Slack thread.

Second, the person who owns the client relationship usually doesn’t own the accounts receivable process. Finance can see the aged debt. The account manager knows the client context. The owner knows which clients deserve a firm call and which ones need a careful conversation. The information sits in three places, and nobody has a complete view.

Third, agencies often have inconsistent payment behaviour across their book. A long-standing client might pay in 14 days until a new finance contact joins. Another client may always pay at 45 days despite 30-day terms. A third is late only when an invoice contains production costs. These patterns matter, but they aren’t visible in a generic overdue report.

The common manual process looks like this:

• Finance exports an aged receivables report from Xero, QuickBooks, NetSuite, or the agency’s financial system.
• Someone sorts invoices by days overdue and balance.
• They check the CRM, Slack, email, project management system, and account notes for context.
• A reminder email gets drafted, often from a generic finance inbox.
• The account manager is copied if the client is sensitive or strategically important.
• Replies arrive across several channels.
• A promise to pay is mentioned in a call or email but isn’t logged.
• The invoice remains on the report next month because no one knows who owns the next action.

This doesn’t break because your people don’t care. It breaks because it is repetitive, context-heavy work with no natural owner.

It also steals time from higher-value work. The same account manager who should be protecting renewal opportunities and finding new work inside an account is instead checking whether a finance contact has processed an invoice.

What an AI invoice follow-up agent actually does

An AI Invoice Follow-Up Agent is not a chatbot that blindly sends payment reminders. That would be a fast way to damage client trust.

A useful agent works from a clear set of operating rules. It knows the status of the invoice, the payment terms, the client relationship, prior communication, and the escalation path. It prepares actions for your team where judgment matters, and it can automate low-risk actions where the process is already clear.

Here is what that looks like end to end.

It starts with the financial record

The agent connects to your accounting platform and reads the relevant invoice data:

• Client name and billing contact
• Invoice number and issue date
• Due date and payment terms
• Amount owing
• Invoice line items
• Previous payments and credit notes
• Days overdue
• Payment history for that client
• Current account balance
• Any invoice disputes or notes

It then applies rules you define. A $1,200 invoice that is three days overdue doesn’t deserve the same treatment as a $48,000 production invoice at 42 days overdue.

You can set thresholds by invoice value, account tier, client type, contract terms, or payment history. A client that typically pays seven days late may receive a friendly reminder. A client whose debt has moved from 15 to 45 days over the last quarter may need a different response.

The objective is not to treat every late invoice as a collections issue. It is to make the right next action visible early.

It checks for client context before drafting anything

This is where most automated dunning tools fall short.

Before preparing a follow-up, the agent checks connected systems for account context. That could include your CRM, project management platform, email history, shared account notes, and Slack or Teams conversations where client matters are discussed.

It looks for signals such as:

• A client has raised a question about the invoice
• A project scope change is still unresolved
• A purchase order is missing
• A new finance contact has taken over
• The client has asked for invoices to be resent to a different address
• The account manager has recorded a promised payment date
• A campaign is paused because the client is under budget pressure
• The client is in an active renewal or expansion discussion

That context changes the follow-up.

If the invoice was sent to the wrong billing contact, the agent drafts a correction. If the client requested backup documentation, it prepares a response with the right attachments. If an account manager has already committed to a call, it flags the invoice rather than sending another automated message.

This is the difference between automation and operational intelligence. The goal is to remove the manual searching, not remove judgment.

The Omni apps layer is designed for this kind of connected work, where the valuable answer sits across finance, CRM, communications, and delivery systems.

It drafts the message in the right voice

For low-risk invoices, the agent can prepare a simple reminder that is clear, polite, and specific. It includes the invoice number, amount, due date, payment link or remittance details, and a direct request for an expected payment date if there is an issue.

For more sensitive accounts, it drafts the follow-up for review by the account manager or finance lead.

The difference is practical. A generic message says, “Your account is overdue.” A context-aware message says, “We wanted to check that invoice 10482 for the August retainer reached the right team. It was due on 15 September. If there is a PO or supporting document you need from us before it can be processed, please let us know and we’ll send it through today.”

That message gives the client an easy path to resolve the problem. It doesn’t make them defensive.

The agent can also draft internal notes. For example, it might tell the account manager that a client now has $27,500 overdue across two invoices, has paid 20 days late for the last three months, and has not responded to two finance reminders. The recommended next step may be a direct call from the account lead, not another email.

It routes exceptions to the person who can solve them

The agent should not turn every late payment into an owner escalation. That creates noise and your team will stop trusting it.

Instead, it routes work based on cause and risk.

An invoice with a missing PO goes to the account manager or project lead who can obtain it. A billing-address issue goes to finance. A client who disputes a line item goes to the delivery owner with the relevant statement of work and change request. A strategic client with a large overdue balance goes to the partner responsible for the relationship.

The agent creates a task, includes the relevant evidence, suggests the next action, and sets a follow-up date. It can notify the right person in Slack, Teams, email, or your task system.

The task is not “chase payment.” It is “Client says the $8,400 video production charge needs approval against PO 7281. Contact Sarah Jones in procurement, attach the approved scope variation, and confirm payment timing by Thursday.”

That specificity is where time gets recovered.

The connection to account management capacity

Invoice follow-up may sit with finance, but it affects account capacity.

Your account managers are already carrying a broad load. In many agencies, AMs spend 30% to 50% of their time assembling reports, building decks, responding to client updates, and coordinating delivery. Each person may cap out at six to 10 meaningful client accounts before quality starts to slip.

When cash collection becomes another manual responsibility, that ceiling comes down further.

This is also why invoice follow-up should not be treated as an isolated bot project. It belongs inside a wider operating model.

The Reporting Agent can pull performance data from connected platforms, draft the monthly report, and prepare the account manager’s email summary. That reduces the report-building burden that crowds out commercial work.

The Account Health Agent watches client accounts daily, flags risk and opportunity, and drafts the next-step message before the account manager has to ask. A worsening payment pattern is one of those health signals. A late payer may be under financial pressure, unhappy with delivery, or simply dealing with an internal process change. Each possibility requires a different response.

When the Invoice Follow-Up Agent and Account Health Agent share the same account context, your team can see the full picture. A delayed payment is no longer just a finance problem. It becomes a prompt to check relationship health, scope, renewal risk, and delivery concerns.

You can see Omni for marketing and creative agencies to understand how these connected agent workflows are mapped against the actual constraints in an agency operating model.

Where the dollars come from

A $60,000 to $180,000 leakage band can sound abstract until you break it into ordinary agency events.

Consider an agency with $5 million in annual revenue and a mix of retainers and project work. It may carry $400,000 to $700,000 in receivables during busier periods. If a portion of that book moves from 30 days to 60 or 75 days without active follow-up, the agency may need to use its overdraft, delay contractor payments, or draw cash from elsewhere in the business.

Then there is the labour cost. Finance staff spend hours every week preparing reports, checking email threads, writing reminders, and updating notes. Account managers become involved when messages go unanswered. Partners get pulled in for the largest accounts.

The final cost is often discounting. A client raises a question late. The agency can’t quickly find the approved scope change, delivery evidence, or email confirmation. To get paid, the team agrees to reduce the invoice. Some discounts are commercially sensible. Others happen because the information was scattered and nobody had time to fight the point.

The agent does not create profit out of thin air. It gives you earlier visibility, tighter follow-up, and a more consistent evidence trail. That protects cash and reduces the amount of senior time consumed by routine debt management.

If you want to find the exact process gaps in your agency, Book a 60-min Omni Audit. The session is built around your operating reality, not a software demo.

The controls that make this safe

Agency owners are right to be cautious about automated client communication. You have spent years building trust. No one wants an agent sending a blunt message to a client who is in the middle of a legitimate dispute.

The answer is not avoiding automation. It is setting proper controls.

Start with draft-only mode. For the first few weeks, the agent prepares messages and next-step tasks, but your team approves them. This lets you review tone, accuracy, routing, and escalation logic.

Set clear approval rules. For example, messages for invoices below a certain value and less than 14 days overdue may send automatically after an initial review period. Anything above a set value, any disputed invoice, or any strategic account requires approval.

Define exclusions. You might exclude clients in renewal negotiations, government clients with fixed payment processes, media invoices, or accounts with active legal issues.

Keep an audit trail. Every action should show what data was used, what message was drafted or sent, who approved it, and what happened next. This is especially important when a client says they never received an invoice or challenges the timing of a reminder.

Use human escalation as a feature, not a failure. The agent is there to identify where human attention creates the most value. It should make those moments easier to act on.

For a broader view of how these workflows are governed, the Omni advisory approach focuses on process design, ownership, controls, and measurable commercial outcomes.

A sensible first 30 days

You don’t need to rebuild your finance system to begin.

In the first week, map the current receivables process. Identify where invoice data lives, who sends reminders, who owns client context, and how exceptions are tracked. Look at the last 90 days of overdue invoices. Find repeated causes, not just overdue totals.

In week two, define your follow-up sequences and escalation rules. Start with the common cases. Invoice due in seven days. Invoice overdue by seven days. No response after 14 days. Client asks for backup documents. Client disputes a line item. Promise to pay missed.

In week three, connect the relevant systems and test the agent against historical invoices. This is where you find edge cases, such as invoices that should not be chased because of an agreed payment plan or a project hold.

In week four, run in draft-only mode with a defined group of clients. Measure time saved, response rates, days sales outstanding, invoices resolved before escalation, and manual touches per invoice.

You can then expand into adjacent work. The Content Production Agent can produce first-pass content from approved briefs so your team edits instead of starting from an empty page. Combined with reporting and account health workflows, that gives your agency more capacity without assuming every growth target requires another account management hire.

You can also browse the EDNA insights library for practical examples of where AI agents fit inside business operations.

The right next step is to audit the workflow

The useful question isn’t, “Can AI send overdue invoice reminders?”

Of course it can.

The better question is, “Where does our cash collection process break, what context does the team need to resolve it, and which actions can be handled without adding risk to client relationships?”

That is what we work through in an Omni Audit.

It takes 60 minutes. You leave with three outputs: a view of the workflows creating the most leakage, a prioritised agent opportunity map, and a practical first implementation path. There is no deck full of generic AI ideas.

For agencies, that conversation often starts with receivables because the cash impact is immediate. It then moves into reporting workload, content production costs, and the account management ceiling that keeps forcing headcount as the main growth lever.

Read more about the AI audit for marketing and creative agencies, then Book my Omni Audit when you’re ready to put real numbers around the opportunity.