For finance teams at 100–5,000 employees

The whole cash cycle.
One system, one audit trail.

Outreach, payment plans, payment capture, bank-feed cash application, disputes, and analytics — with modelling that runs on your own history and shows its reasoning.

Model the impact

Published pricing. Integrate against your own stack with a documented REST API — no implementation partner, no rip-and-replace.

app.arpilot.ai
Total AR
$142,500
Collected (30d)
$87,200
DSO
28 days
Recent InvoicesView all
Acme Corp
14 days
$12,450.00overdue
TechFlow Inc
3 days
$8,200.00pending
Summit Labs
Today
$24,800.00sent
Vertex Digital
Paid
$6,750.00paid
Workflow Activity
Email reminder sent
2 min ago
Payment link opened
1 min ago
Follow-up scheduled
in 3 days
Collection progress67%

Visibility and control at the action layer

Every scheduled communication is inspectable before it executes. Preview the rendered content, edit it, reschedule it, pause it, cancel it — from a list or a calendar. Nothing reaches a customer that your team couldn't have seen and stopped.

  • Trigger-based sequences across email, SMS and voice, with escalation-contact routing and business-hours randomisation
  • Write-backs to your ledger queue for approval — nothing posts silently, whether it reaches us through the API or a connector
  • A payment recorded anywhere — bank feed, payment link, or your accounting system — cancels the pending reminders on that invoice
Upcoming messages
4 scheduled
Northgate Facilities
Reminder — 3 days before due
Tomorrow, 9:00am
Bellweather Construction
Nudge — 5 days late
Thursday, 9:00am
Kestrel Property Group
Follow-up — 15 days late
Friday, 11:00am
Aldridge & Co
Escalate to accounts payable
Mon 14th, 9:00am
Preview
To: accounts@northgate.example
Invoice INV-2841 — due Friday

Hi Dana,

Just a quick note that invoice INV-2841 for $4,820.00 is due this Friday. You can pay by card or bank transfer here:

View and pay invoice →

Thanks — Marcus

Sends tomorrow unless you change it

Coverage

The whole cycle, not one slice.

Reminders, payment plans, payment capture, bank-feed cash application, dispute intake, analytics. Point tools cover one or two of these, and your team feels every seam between them — a separate contract, a separate admin model, a separate security review, and a reconciliation project every time someone asks what collection efficiency actually is.

One
Data model across outreach, payments, reconciliation and disputes
One
Permission model, enforced in PostgreSQL
One
Audit trail covering every action and every operator
One
Security review instead of four vendor questionnaires
We use AI to give you superpowers

Like having a data scientist on the AR team

The analysis a finance-analytics hire would do on your receivables, running continuously, on your own data, with its work shown. Every recommendation carries a confidence score, the data points behind it, the current and proposed values, and the reason.

  • Nine dimensions tuned on your history: send timing, time of day, template choice, frequency, channel mix, step reordering and removal, payment-plan tuning, dispute prevention
  • Invoice-level modelling — candidate strategies ranked by likelihood, expected days to payment, plus a payment-plan recommendation and a dispute-risk prediction
  • Messages generated against each invoice's specifics, with tone selected from how far past due it sits, and template effectiveness fed back into the engine
  • It refuses to guess: the engine gates on data quality and says whether it is operating on sufficient, limited or insufficient history
  • Recommendations persist as pending. A workflow mutates only when an operator applies one — there is no autonomous self-tuning path

An optional shared pattern pool exists — anonymised, bucketed signals only, never amounts, names, or message content. It is opt-in, reversible, and the contributor pool is still small, so treat it as upside rather than as the reason the modelling works. Everything above runs on your data alone.

Suggested change
Pending

Send the second reminder earlier

Now
7 days late
Suggested
5 days late
Why

Across 214 of your own reminders, the ones sent on day 5 were answered noticeably sooner than the ones sent on day 7. Tuesday and Wednesday mornings did best.

Confidence
0.82
ApplyDismiss

Nothing changes until you apply it.

Cash application that shows its working

A Plaid bank feed lands in staged transactions, then each deposit is scored against candidate invoices on amount, date proximity, payer-name similarity and reference number. The thresholds are published, not proprietary.

  • Above 0.95 applies automatically and cancels that invoice's pending reminders
  • Above 0.70 goes to a review queue with candidates ranked, so the decision takes seconds
  • Below that it stays unmatched — a wrong auto-match costs more than an unmatched deposit
Bank deposit
$4,820.00ACH · NORTHGATE FAC LLC
INV-2841 · Northgate Facilities0.97
Exact amount · payer name match · reference on the remittanceApplied automatically
INV-2799 · Northgate Holdings0.74
Exact amount · similar payer name · no referenceTo review queue
INV-2705 · Northbridge Ltd0.41
Near amount onlyLeft unmatched
≥ 0.95 auto-apply≥ 0.70 reviewbelow · unmatched

Payment plans, disputes, and the reporting layer

The three things most often missing from a cadence tool, each measured rather than anecdotal.

Payment plans

  • Instalment templates with optional financing fee
  • Optional early-enrolment discount
  • Customer self-enrolment from a secure link
  • Card and ACH charged on schedule via Stripe
  • Default tracking as instalments are missed
  • Modelled plan-shape and completion-window recommendations

Disputes

  • Customer-initiated structured intake
  • Raising a dispute pauses collections on that invoice
  • IVR-initiated holds from an automated call
  • A cleanup job releases holds that go stale
  • Correspondence threaded to the dispute, not the invoice
  • Dispute risk modelled per invoice before it is raised

Analytics

  • DSO by several methods, with the method stated
  • Collection efficiency, bad-debt rate, aging
  • Payment-plan and dispute analytics as standing series
  • At-risk tiering from engagement, not days overdue alone
  • Human opens distinguished from automated ones
  • Every formula published, not proprietary

The problem with vendor dashboards is rarely that they are wrong — it is that nobody can explain how they were computed. Every figure above ships with its computation.

Integration

Built to be integrated, not migrated onto.

You are not going to move your ledger to suit a collections tool, and we would not ask. ARPilot sits on top of the system you already run: a documented REST API with an OpenAPI schema, scoped keys, and webhooks emitted by database triggers rather than by application code someone has to remember to instrument. Teams running NetSuite, Sage Intacct, Dynamics or SAP integrate through that surface; the ingestion core is source-agnostic with per-source adapters. QuickBooks Online has a native bidirectional connector today, and it is one supported path rather than the required one.

OpenAPI
Published schema, scoped keys, per-minute and monthly rate limits
Triggers
Webhooks fired by the database, so an event cannot be missed
7
Delivery attempts with backoff, 30s to 24h, with a full log
Adapters
Source-agnostic ingestion for the system you already run

Model it on your own numbers

What five days is worth to you

Your numbers, our arithmetic, every assumption visible and editable.

$
days

Assumption — adjust to your team.

Assumption — adjust to your team.

Nothing you type here leaves your browser.

Cash released per 5 days of DSO improvement

$548K

Tied up in receivables
$5.7M
Follow-up each month
200 hrs
Against your AR team
about half a full-time role

These are estimates computed from the inputs and assumptions above. They describe the size of the opportunity in your own numbers — they are not a forecast of results with ARPilot, and we do not publish outcome percentages we cannot source.

What you are probably comparing us against

Suites and portals

  • Enterprise suites: priced and implemented for the Fortune 1000. Same capability surface, live in days, no implementation partner
  • Buyer portals: their value depends on your customers adopting a portal. ARPilot works through the inbox and phone they already use

Point tools and builds

  • Cadence tools: the closest comparison. We add bank-feed cash application, auto-charging plans, dispute intake with holds, and a modelling layer
  • Native QuickBooks or Stripe dunning: static reminders, no escalation logic, no cash application, no disputes, no analytics
  • Building it internally: easy to start, expensive to own — deliverability, retries, consent, reconciliation, and the on-call rotation

Controls

What your security review will ask

Where is tenant isolation enforced?

Row-level security in PostgreSQL. Application-layer checks exist for user experience; the database is the boundary that matters.

How are credentials stored?

AES-256 via pgcrypto, with the passphrase held in a managed vault rather than in application config or environment variables.

Can support staff see our data?

Only through an impersonation path that is read-only and written to an audit trail.

Is MFA available?

Yes, including enforced MFA on privileged actions rather than MFA as an account-level preference.

Can the AI act on a customer without us?

It can send what your team scheduled — that's the automation you're buying. It cannot change your workflows, templates, or cadence on its own: those recommendations stay pending until an operator applies them. And every scheduled action is visible, editable, and cancellable before it runs.

What about TCPA exposure on SMS and voice?

A do-not-contact list and a consent audit log, with recordings tracked against consent.

Are you SOC 2 certified?

Not yet. We describe controls, not certifications, and we won't claim a report we don't have. If a completed SOC 2 is a gate for you, say so in the first call.

Published pricing, not a quote request

You can model the cost before you talk to anyone.

Professional

$299/month

For a finance team running collections as a process, with the API surface to feed your warehouse and BI.

  • 200 invoices a month
  • 25% unused invoices roll over
  • 10 follow-up sequences
  • 50 message templates
  • 10 users
  • 25 payment plans (0.2% fee)
  • Priority support
  • API access and webhooks

$3,000/year — about 2 months free

Enterprise

Custom

Unlimited volume with an SLA, custom integrations, and dedicated support. Priced per engagement.

  • Unlimited invoices a month
  • 50% unused invoices roll over
  • Unlimited follow-up sequences
  • Unlimited message templates
  • Unlimited users
  • Unlimited payment plans
  • Dedicated support
  • API access and webhooks

Annual billing works out cheaper. Stripe processing is standard pass-through.

Compare every plan →

See it against your own aging report.

A demo runs about thirty minutes. Bring a current aging export and we'll walk the action hub, the matching thresholds, and what the model says about your book.

Published pricing
No implementation partner required
Controls described, never certifications claimed