automationfinanceinvoicingaccounts-receivable

How to Automate Invoices | 2V Automation

How to automate invoices from generation to cash applied - invoicing rules, AR workflows, dunning sequences, and payment reconciliation that actually works.

SF
Sergey Furman Partner, 2V Automation
·
Jump to a section

To automate invoices end-to-end, build four connected workflows: invoice generation from CRM or billing system events, multi-channel delivery with audit trails, an aged-AR follow-up engine with escalation rules, and bank-feed-to-invoice payment matching. The platforms that do this well are billing systems (Stripe Billing, Chargebee, Maxio, NetSuite SuiteBilling), an accounting system as system of record (QuickBooks, Xero, NetSuite), specialist AR tools (Chaser, Upflow, Versapay), and a workflow engine to glue them together (n8n, Make, or Workato). Done right, you’ll cut DSO by 10-25 days and reclaim 60-80% of finance team hours spent on collections.

This post walks through what actually happens manually today, what automates cleanly vs what stays human, the tool categories with named examples, a concrete workflow, the ROI math, and the failure modes we see most often.

The manual invoicing process today

Most B2B companies between $5M and $100M run a version of these steps:

  1. Trigger - sales books a deal in the CRM, a milestone gets hit in the project tool, or a subscription renews.
  2. Manual invoice creation - someone in finance opens QuickBooks/Xero/NetSuite, copies deal data, picks the right line items, applies the right tax rate, generates the invoice.
  3. Internal review - depending on size, a manager approves.
  4. Delivery - emailed (sometimes as a PDF, sometimes via a billing portal link), occasionally posted to a customer-portal upload like Coupa or Ariba.
  5. Customer confirmation - sometimes they reply, often they don’t.
  6. Aging starts - 30, 60, 90 day buckets in the AR report.
  7. Reminders - controller emails the customer at 35, 50, 65 days. CCs the AE. Calls at 75. Escalates to legal at 120.
  8. Disputes - customer claims the PO number was wrong, the line item doesn’t match, the contact has changed. Email threads ensue.
  9. Payment lands - wire, ACH, check, credit card. Hits the bank account.
  10. Cash application - controller manually matches the deposit to the open invoice. Closes it in QuickBooks.
  11. GL update and revenue recognition - recorded against the right accounts.

The pain shows up in steps 2, 7, 8, and 10. A single accounts receivable analyst at a $20M company typically spends 70-80% of their week here.

What automates end-to-end vs what needs humans

Fully automatable:

  • Invoice generation triggered by closed-won, renewal, or milestone events
  • Tax calculation (Avalara, TaxJar, Stripe Tax)
  • Multi-channel delivery (email, portal upload, customer billing portals)
  • Aged AR reporting and Slack/email digests
  • Polite reminder sequences at 7/14/30/45/60 day buckets
  • Payment receipt confirmation
  • Bank feed import and probabilistic matching
  • Sub-ledger updates to the GL
  • Subscription renewals, upgrades, and prorations
  • Failed-payment retry logic (smart dunning)

Needs humans in the loop:

  • Pricing decisions for non-standard deals
  • Dispute investigation
  • Credit memos and refunds beyond a threshold
  • Escalation calls and difficult conversations
  • Reconciling ambiguous payments (multi-invoice deposits without remittance detail)
  • Anything customer-facing that involves a strategic account

Should not be automated even if you can:

  • Aggressive automated tone with strategic customers
  • Auto-suspending service to a top-10 account
  • Anything that looks like a robot escalating to legal - that’s a human conversation

The honest part: bank reconciliation gets 85-95% automatable, never 100. Some payments arrive without a clear invoice match. Build for the long tail of exceptions, don’t pretend they don’t exist.

Tool categories that fit

Layer 1: System of record

Your accounting system is the source of truth.

  • QuickBooks Online - best for sub-$50M companies, strong integration ecosystem.
  • Xero - strong outside the US, clean API.
  • NetSuite - for $50M+ companies, complex multi-entity, multi-currency situations.
  • Sage Intacct - strong mid-market alternative to NetSuite.

Layer 2: Billing engine

Where the invoice actually gets built.

  • Stripe Billing - best for SaaS, subscription-first, excellent API.
  • Chargebee - strong subscription billing with revenue recognition.
  • Maxio (formerly SaaSOptics + Chargify) - built for SaaS revenue ops.
  • Recurly - solid subscription billing.
  • Zone Billing / NetSuite SuiteBilling - when you’re already on NetSuite.

Layer 3: AR / collections specialist

These bolt on to handle the follow-up motion.

  • Chaser - automated dunning, customer-friendly tone.
  • Upflow - modern AR ops platform.
  • Versapay - collaborative AR with customer portals.
  • YayPay (Quadient AR) - predictive AR, common at mid-market.
  • Tesorio - AR + AP forecasting.

Layer 4: Workflow engine

Connects everything to your CRM, project tool, and Slack.

  • n8n - handles webhooks from Stripe, calls QuickBooks, pushes Slack alerts, runs scheduled aged-AR digests. See What is n8n and the n8n automation guide.
  • Make - visual scenarios for finance ops teams without engineers.
  • Workato - enterprise option, audit logs, strong governance.

Layer 5: Payment processing and reconciliation

  • Stripe, Adyen, Braintree - card and ACH processing
  • Modern Treasury, Moov - bank ops and reconciliation infrastructure
  • Plaid, Codat - bank feed connectivity

A concrete workflow recipe

Here’s the AR backbone we deploy for B2B SaaS clients in the $5M-$50M range. Stripe Billing + QuickBooks + n8n + Chaser.

Trigger 1: Deal moves to closed-won in HubSpot

  1. n8n webhook fires from HubSpot stage change.
  2. n8n calls Stripe Billing API: creates the customer (if new), creates the subscription with the right plan, applies the start date and any negotiated discount.
  3. Stripe generates the first invoice automatically.
  4. n8n posts confirmation to a Slack #new-invoices channel with a one-line summary.
  5. Stripe webhook fires when the invoice is finalized → n8n updates the HubSpot deal with the Stripe invoice URL.

Trigger 2: Stripe invoice sent

  1. Customer gets the email with a hosted invoice page (PDF + pay link).
  2. If the customer requires portal upload (Coupa, SAP Ariba, Tradeshift): n8n watches a Notion or Airtable list of “portal customers” and pings the AR analyst to upload manually - automate this step only if you have proper portal-integration tooling.
  3. Internal CRM updated with “invoice sent” timestamp.

Trigger 3: Aged invoice (cron)

  1. Daily, n8n queries Stripe + QuickBooks for all unpaid invoices past due.
  2. For each: pulls customer data, last contact, days overdue, total open.
  3. Sends a 7-day-overdue gentle reminder (templated, signed by the AR analyst’s name)
  4. At 21 days: escalates - AR analyst gets a Slack DM with the invoice, suggested message draft, and “send” button.
  5. At 45 days: AE for the account is looped in.
  6. At 60 days: controller and AE get a meeting auto-booked to discuss.
  7. At 90+ days: legal escalation flag, never auto-sent.

Trigger 4: Payment received

  1. Stripe payment webhook → n8n.
  2. n8n marks invoice paid in QuickBooks.
  3. Updates HubSpot deal with payment date.
  4. Posts to Slack #payments-received.
  5. Cancels any pending dunning steps.

Trigger 5: Bank feed payment without Stripe ID (wire/ACH/check)

  1. n8n pulls bank transactions daily via Plaid.
  2. For each deposit: extracts amount, sender name, memo.
  3. Searches QuickBooks open invoices for a match by amount + customer name fuzzy match (Levenshtein ratio).
  4. Confidence > 0.85 → auto-apply, mark paid.
  5. Confidence 0.5-0.85 → suggest to AR analyst with a one-click “apply” button.
  6. Confidence < 0.5 → human investigation queue.

Realistic 6-month outcomes: DSO down 12-22 days, AR analyst hours/week from 35 to 8, percentage of invoices auto-applied 80-92%.

ROI math

Use the ROI calculator with your own numbers. The model for a $20M ARR B2B SaaS with 500 active invoices/month:

Before:

  • 1.5 FTEs in AR + collections at ~$70K loaded = $105K/year
  • DSO at 52 days on $20M = $2.85M in tied-up cash
  • Cost of capital at 8% = ~$228K/year of opportunity cost

After:

  • 0.4 FTE workload, redeploy 1.1 FTE
  • DSO drops to ~34 days → $1.86M tied up → $74K savings on cost of capital
  • Faster, cleaner invoice generation drops billing errors and disputes ~50%

Implementation: $20K-$60K depending on portal-upload complexity and number of payment channels. The workflow cost calculator models ongoing tooling spend.

Common pitfalls

Auto-sending aggressive dunning to your top customers. A polite reminder to a $2K-MRR customer is fine. The same automated tone to a $50K-MRR enterprise customer breaks the relationship. Segment by ARR/strategic flag, and route high-value accounts to human-sent reminders by default.

Bank reconciliation that auto-applies low-confidence matches. A wrong cash application creates ghosts in the GL that take months to untangle. Always require human review below your confidence threshold.

Forgetting tax automation. US sales tax nexus rules and EU VAT rules are nontrivial. Wire in Avalara or Stripe Tax from day one rather than retrofitting.

No exception report. Without a daily exception queue, the long tail of weird cases (multi-invoice deposits, foreign currency rounding, voided invoices) silently breaks the books.

Treating the AR system as set-and-forget. Customer payment behavior shifts. Tone templates need updating. New billing models (usage-based, hybrid) break old assumptions. Review the system quarterly.

Skipping the audit trail. Every automated email, status change, and cash application needs a logged event. Auditors will ask. See how to monitor AI automation performance for the right pattern.

Implementation phasing

Phase 1 (weeks 1-3): Clean source-of-truth setup. Stripe Billing (or equivalent) wired to QuickBooks, customer records deduped, plans and SKUs cleaned up.

Phase 2 (weeks 4-7): Triggered invoice generation. CRM closed-won → invoice fires. Renewals automated. Portal-upload customers flagged.

Phase 3 (weeks 8-12): AR follow-up engine. Tiered dunning, ARR-aware tone, escalation rules, Slack-based AE loop-in.

Phase 4 (weeks 12-18): Payment reconciliation. Bank feed integration, fuzzy matching, exception queue.

Phase 5 (weeks 18-24): Reporting and continuous improvement. Cash forecasting, AR aging dashboards, exception trend analysis.

The full finance automation backbone lands around week 24.

Connecting to the rest of finance

Invoice automation is one half of a clean close. The output feeds directly into automated financial reports. New invoices fire from the sales process automation backbone. Customer success teams need invoice and payment status visible during onboarding and renewal motions. Run a baseline assessment with the efficiency scorecard before you build anything. For the broader design pattern, the complete guide to business process automation is the right primer.

Frequently asked questions

How long does invoice automation take to implement?
A typical mid-market engagement is 12-18 weeks for the full backbone (generation + delivery + AR + reconciliation). Phase 1 (generation and delivery) can ship in 4 weeks. DIY without prior finance-automation experience usually takes 2x that.
Can I automate invoicing without changing my accounting system?
Yes. QuickBooks, Xero, and NetSuite all have strong APIs. Most automation work happens around them, not by replacing them. Replacing the accounting system is the most expensive, most disruptive change available - don't do it unless you have a real reason.
What about customers that require portal uploads (Coupa, Ariba, SAP)?
Portal uploads are partially automatable via vendor connectors (e.g., Tradeshift, OpenInvoice integrations). For the long tail of less common portals, automate the prep (PDF, line-item CSV, alert) and leave the upload click as a human task. Don't try to brittle-RPA every portal.
How accurate is automated bank reconciliation?
Modern fuzzy matching against open invoices hits 80-95% auto-application rates after a tuning period. The remaining 5-20% needs human eyes. The goal isn't 100% - it's making the human-touched ones a small, prioritized queue.
What's the difference between dunning and AR follow-up?
Dunning historically meant escalating reminder sequences for failed credit card payments (subscription churn rescue). AR follow-up is the broader process - reminders for any unpaid invoice, regardless of payment method. Modern tools handle both. The principles are the same: escalating tone, human handoff at a threshold, never break trust.
Should I use my billing platform's AR features or a specialist tool?
For sub-$10M ARR with mostly card payments, Stripe Billing's dunning is enough. Above that, especially with multi-channel payments (wire, ACH, check), a specialist (Chaser, Upflow, YayPay, Versapay) earns its keep through better segmentation, customer portals, and team workflows.
How do I handle disputes and chargebacks?
Build a dispute workflow as a peer to the AR workflow, not an afterthought. Tag invoices in dispute so dunning pauses. Route to an owner (AE or CSM). Set an internal SLA on resolution. The dispute rate itself is a signal - if it's rising, the issue is upstream (in scoping or onboarding), not in collections.
What if my pricing is highly custom per deal?
Use a CPQ tool (Salesforce CPQ, DealHub, PandaDoc) to standardize the quote. The quote becomes the invoice input. The fewer hand-typed line items in your invoicing flow, the fewer bugs in your GL.