Automate Accounts Receivable Follow-ups: Build an AI Agent That Sends…

Every finance team has that one person — maybe it's you — who spends Monday mornings staring at an aging report, copying invoice numbers into emails, and politely asking customers to please, for the love of god, pay their bills. It's tedious. It's repetitive. And it's costing your business way more than you think.

Here's the thing: accounts receivable follow-up is one of those workflows that feels like it needs a human touch, but 70-80% of it genuinely doesn't. The first reminder? Template. The second reminder? Slightly firmer template. The third? Even firmer template with the invoice attached again because they "didn't receive it."

This is exactly the kind of work an AI agent can handle — not in some theoretical future, but right now, today, using OpenClaw. Let me walk you through how it works, what you can realistically automate, and where you still need a human in the loop.

The Manual Workflow (And Why It's Killing Your Team)

Let's be honest about what AR follow-up actually looks like at most companies. Not the aspirational version. The real one.

Step 1: Pull the aging report. Someone opens QuickBooks, NetSuite, Xero, or whatever you're running. They export the aging report — usually filtered by 30+, 60+, and 90+ days overdue. This takes 10-15 minutes if the system cooperates, longer if it doesn't.

Step 2: Research each account. Before you can send a reminder, you need context. Has this customer been contacted already? Is there a dispute? Did they short-pay? Are they a $500/month account or a $50,000/month account? This means checking the ERP, the CRM, email threads, maybe a shared spreadsheet where Karen logs her notes. Average time per account: 5-10 minutes.

Step 3: Draft and send the reminder. Copy-paste a template, swap in the invoice number and amount, maybe personalize the opening line if it's a bigger customer. Attach the invoice PDF because half the time they'll say they never got it. Send. Repeat 20-50 times.

Step 4: Log the activity. Go back to your CRM or spreadsheet. Record that you sent the email, what you said, and when. Set a follow-up reminder for next week if they don't respond.

Step 5: Escalate. If there's no response after the second or third touch, the tone shifts. Phone calls start. Maybe the sales team gets pulled in. At 90+ days, someone has to make the call on whether to involve a collections agency or write it off.

Step 6: Repeat forever.

A typical AR specialist spends 60-80% of their time on these low-value follow-up tasks. The average cost to manually collect a single commercial invoice is $15-$40. And companies running this process manually have DSO (Days Sales Outstanding) that's 15-25 days higher than their automated competitors.

That's real money sitting in other people's bank accounts instead of yours.

What Makes This So Painful

The time cost is obvious. But the second-order problems are worse.

Inconsistency. When three different people send collection emails, you get three different tones, three different levels of firmness, and three different compliance risk profiles. One person is too aggressive with a strategic account. Another is too soft with a serial late-payer. There's no system — just vibes.

Scale ceiling. One collector can effectively manage 150-300 accounts. That's it. Beyond that, things start slipping through cracks. Invoices age out silently. Small balances get ignored. And those small balances add up — the average U.S. company writes off 1.5-4% of receivables annually.

Data fragmentation. Notes live in email threads, ERP comments, CRM records, Slack messages, and Post-it notes on someone's monitor. When that person goes on vacation or quits, their institutional knowledge walks out the door with them.

Customer relationship damage. Ironically, manual processes often hurt relationships more than automated ones. Reminders go out at the wrong time (the customer already paid yesterday). The tone doesn't match the situation. Or worse — a good customer who's 5 days late gets the same treatment as a deadbeat at 90 days.

According to a 2026 Robert Half survey, 43% of finance leaders say collections is their single most time-consuming process. And yet only 23% of companies have achieved significant automation in this area.

That gap is your opportunity.

What AI Can Actually Handle Right Now

Let's be clear-eyed about this. I'm not going to tell you AI can replace your entire AR department. It can't. But it can handle the repetitive, rule-based, high-volume work that's eating your team alive.

Here's what an AI agent built on OpenClaw can realistically do today:

Automated dunning sequences. The agent monitors your invoices and triggers reminders based on rules you define. Three days past due? Friendly nudge. Fourteen days? Firmer reminder with the invoice attached. Thirty days? Escalation notice. The agent handles the timing, the tone, and the delivery — across email, SMS, or even WhatsApp.

Predictive prioritization. Not all overdue invoices are equal. An OpenClaw agent can analyze payment history, invoice size, customer segment, and past behavior to rank accounts by risk. Your team sees a prioritized list every morning instead of a raw aging report.

Personalized messaging at scale. This isn't "Dear {First_Name}" mail merge. An OpenClaw agent can reference the specific invoice, the customer's payment history ("We notice your last three invoices were paid within 15 days — this one seems unusual"), and adjust tone based on the relationship tier. It reads the context and writes accordingly.

Cash application support. When payments come in, matching them to the right invoice is still surprisingly manual at most companies. An AI agent can match 80-90% of payments automatically by cross-referencing amounts, reference numbers, and customer data — flagging only ambiguous cases for human review.

Dispute detection and routing. When a customer replies to a reminder saying "We're not paying because the shipment was damaged," the agent can classify the dispute type and route it to the right person — shipping, sales, or customer success — instead of it sitting in the AR inbox for three days.

You can find pre-built agent templates for workflows like this on Claw Mart, which saves you from starting from scratch. Browse what's available, customize it for your stack, and deploy.

Step-by-Step: Building the AR Follow-Up Agent on OpenClaw

Here's how to actually build this. I'll keep it practical.

Step 1: Define Your Data Sources

Your agent needs access to your invoice and payment data. At minimum, you need:

Aging report data (invoice number, customer, amount, due date, days overdue, status)
Customer contact info (email, phone, account manager)
Payment history (average days to pay, dispute history, credit terms)
Communication log (what's been sent already)

OpenClaw connects to most accounting systems via API — QuickBooks, Xero, NetSuite, Sage Intacct. If yours has an API, you can pull data in. If it doesn't, a CSV export on a scheduled basis works as a fallback.

Step 2: Set Up Your Dunning Rules

Define the logic tree for your follow-up sequence. Here's a starting framework:

TRIGGER: Invoice status = unpaid AND days_past_due >= 3

TIER 1 (3-7 days overdue):
  → Send friendly reminder email
  → Tone: casual, helpful
  → Include: invoice PDF, payment link
  → Channel: email

TIER 2 (14-21 days overdue):
  → Send firmer reminder
  → Tone: professional, direct
  → Include: invoice PDF, payment link, account summary
  → Channel: email + SMS (if enabled)
  → Flag for AR dashboard

TIER 3 (30-45 days overdue):
  → Send escalation notice
  → Tone: formal
  → CC: account manager
  → Include: full statement, payment options
  → Channel: email
  → Alert: Slack notification to AR team

TIER 4 (45+ days overdue):
  → STOP automation
  → Route to human collector
  → Include: full communication history, risk score, recommended action

In OpenClaw, you build this as a workflow with conditional branches. Each branch triggers the appropriate agent action — draft the email, select the template, pull in the invoice data, and send.

Step 3: Build Your Message Templates

The agent needs message templates for each tier, but they shouldn't read like templates. Here's where OpenClaw's language capabilities shine. Instead of rigid mail merge, you give the agent instructions like:

Write a payment reminder for {customer_name} regarding invoice 
{invoice_number} for {amount}, which is {days_overdue} days past due. 

Context: This customer's average payment time is {avg_days_to_pay} days. 
Their payment history is {good/mixed/poor}. This is reminder #{sequence_number}.

Tone: {friendly / professional / formal} based on tier.

Include: Direct payment link. Offer to resend invoice if needed. 
If this is a repeat late payer, mention the pattern diplomatically.

Do NOT: Threaten legal action. Use aggressive language. 
Make promises about credit terms.

The agent generates a contextually appropriate message every time. Not a static template — an actual personalized email that accounts for who this customer is and where they are in the process.

Step 4: Configure Escalation and Human Handoff Triggers

This is critical. You do not want an AI agent handling every situation autonomously. Set clear handoff rules:

Invoice amount > $25,000 → Human review before any communication
Customer tagged as "strategic" or "key account" → Human drafts or approves messages
Dispute detected in customer reply → Route to appropriate department
Customer requests phone call → Alert human collector immediately
Third reminder with no response → Queue for human outreach
Any legal or compliance-sensitive situation → Full stop, human only

In OpenClaw, these are guard rails you configure at the workflow level. The agent respects them absolutely.

Step 5: Connect Your Channels

Wire up the actual sending mechanisms:

Email: Connect via your email provider's API (Gmail, Outlook, SendGrid). The agent sends from your AR team's email address, not some generic bot address.
SMS: Twilio or similar integration for text-based reminders (increasingly effective — open rates are 95%+ vs. 20-30% for email).
Slack/Teams: Internal notifications when the agent escalates or needs human input.
Your accounting system: Write-back capability so the agent logs every action directly in your ERP. No more manual activity logging.

Step 6: Test on a Small Segment

Don't flip the switch on your entire AR book on day one. Start with:

Invoices under $5,000
Customers with good payment history (low risk of needing human nuance)
Only Tier 1 and Tier 2 reminders (keep humans on Tier 3+ initially)

Run it for 2-4 weeks. Monitor response rates, payment timing, customer feedback, and any edge cases the agent handles awkwardly. Adjust your templates, rules, and escalation triggers based on what you see.

Step 7: Expand and Optimize

Once you're confident in the system, expand the scope:

Add higher-value invoices with human approval gates
Enable Tier 3 automation for non-strategic accounts
Turn on SMS for customers who don't respond to email
Add predictive scoring to prioritize the human collector's daily queue

The agent gets better over time as it accumulates data on what works — which subject lines get opened, which send times get faster responses, which customers need a phone call versus an email.

What Still Needs a Human

Let me be direct about the boundaries. AI is great at pattern execution. It's not great at judgment calls that require business context, empathy, or strategic thinking.

Keep humans on these:

Complex dispute resolution. If a customer says the product was defective or the pricing was wrong, that requires investigation, coordination with other departments, and negotiation. An AI agent can detect the dispute and route it. It shouldn't try to resolve it.
Strategic account management. Your top 10 customers by revenue should never receive a fully automated collection email without someone reviewing it. The relationship dynamics are too nuanced, and the downside of getting it wrong is too high.
Hardship and payment plan decisions. When a customer asks for extended terms or says they're having cash flow problems, a human needs to evaluate the risk and make a call. The agent can surface the customer's history and recommend options, but the decision is human.
Final escalation. Sending an account to a collections agency or initiating legal action is a business decision with real consequences. AI surfaces the recommendation. A person pulls the trigger.
Tone calibration when things get emotional. If a customer replies angrily to a reminder, the agent should immediately hand off to a human. Automated responses to angry customers almost always make things worse.

The best model is what leading companies already use: AI handles everything up to 45-60 days for standard accounts. Humans focus exclusively on high-value, high-risk, and disputed accounts. Your collectors stop being email machines and start being strategic problem-solvers.

Expected Time and Cost Savings

Let's put numbers on this, based on published benchmarks and real-world results from companies that have automated AR follow-up.

Metric	Before Automation	After Automation	Improvement
Time per invoice follow-up	8-15 minutes	1-2 minutes (human review only)	80-90% reduction
Accounts per collector	150-300	500-1,000+	3-4x capacity
DSO (Days Sales Outstanding)	45-65 days	30-45 days	20-35% reduction
Cost per collected invoice	$15-$40	$3-$8	70-80% reduction
On-time payment rate	60-70%	80-90%	15-25% increase
Bad debt write-offs	1.5-4% of receivables	0.5-2%	Significant reduction

For a company processing 500 invoices per month with an average value of $2,000, reducing DSO by even 10 days means $333,000 in accelerated cash flow. That's not savings — that's cash you already earned, arriving sooner.

The automation itself pays for itself within the first quarter for most companies. The ongoing ROI compounds as the system learns your customers' patterns and optimizes accordingly.

Where to Start

If you're drowning in AR follow-up work, here's what I'd do this week:

Go to Claw Mart and look at the pre-built AR automation agent templates. Don't build from zero if you don't have to.
Map your current dunning process — every step, every email, every rule. You can't automate what you haven't documented.
Pick your smallest, lowest-risk segment and set up a pilot. Invoices under $5,000, good customers, first and second reminders only.
Run it for a month, measure the results, then expand.

The technology for this isn't experimental. It's just underadopted. Only 23% of companies have meaningfully automated their collections process, which means the other 77% are leaving money on the table — literally.

Your AR team shouldn't be spending their days copying invoice numbers into email templates. Build the agent. Let it handle the routine. Let your people handle the work that actually requires a brain.

Ready to automate your accounts receivable follow-ups? Check out the AR agent templates on Claw Mart or reach out to our Clawsourcing team to have experts build a custom AR automation agent for your business on OpenClaw. We'll scope it, build it, and get it running — so your team can stop chasing payments and start focusing on the work that moves the needle.

Automate Accounts Receivable Follow-ups: Build an AI Agent That Sends Payment Reminders