Billing by Conversation: How Conversational AI Is Rewriting Invoicing and Dunning
The fastest billing interface is not a better form. It is a sentence. See, and hear, how conversational AI creates invoices, runs dunning ladders, and collects overdue money while you do the work you actually get paid for.
Every invoicing tool ever built has asked you to learn its shape. Which menu holds recurring invoices, which tab hides the tax setting, which of four buttons actually sends the thing. You did not start a business to become fluent in somebody else’s software, and yet the price of getting paid has always been an afternoon spent clicking.
Conversational AI collapses that. When the interface is language, there is nothing to learn, because you already speak it. You say what you want and the system does it: create the invoice, chase the client, tell me who owes me money. This article is about what changes when billing works that way, with a two minute conversation you can listen to, the exact dunning ladders it sets up, and an honest account of where the guardrails sit.
The Real Cost of Billing Software Is Not the Subscription
Ask any freelancer or agency owner what they hate about invoicing and almost nobody says the price of the tool. They say the friction. The invoice that sat in drafts for nine days because opening the app felt like a chore. The reminder they never sent because they could not find the words. The month they billed late because the work was more interesting than the billing. That friction has a number attached, and it is much larger than any subscription.
Where the money actually leaks
- Invoices sent late, because sitting down to build one is a task in itself.
- Overdue invoices that nobody chases, because chasing feels rude and takes emotional energy.
- Billable hours that never make it onto an invoice, because logging them at the time was inconvenient.
- Follow-up that stops after one polite email, exactly when persistence would have worked.
- No clear answer to the simplest question in business: who owes me money right now?
None of these are knowledge problems. Every one of them is an interface problem. The work required to express what you want has always been larger than the thought itself, and so the thought does not get acted on. Conversational AI removes exactly that gap.
What Conversational AI Means in a Billing Context
Conversational AI is not a help chatbot bolted onto a settings page. A support bot answers questions about the product. A conversational billing assistant operates the product: it has authenticated access to your real invoices, clients, and time entries, and it takes actions on them on your behalf. The difference between the two is the difference between a manual and a colleague.
The AI Billing Assistant in Invoice My Clients works this way. It runs on Claude, it understands plain language by text or by voice, and every action it takes lands on your actual account data in real time. Ask it to invoice a client and an invoice exists. Ask who is overdue and it reads your live receivables. Say "chase them" and a real dunning campaign starts running. See how the AI billing assistant works.
Three properties that make it different
- It acts, rather than advises. The result of a request is a created invoice or a scheduled reminder, not instructions for doing it yourself.
- It holds context. Say "those invoices" or "that client" and it knows which ones you mean, so you never have to restate the situation.
- It works by voice as well as text, with speech recognition and lifelike speech back, so billing is possible with your hands full and your laptop shut.
Example One: An Invoice, In One Sentence
The clearest way to see the shift is to compare the two paths to the same outcome. On the left is what a conventional invoicing tool asks of you. On the right is the entire interaction when the interface is language.
- 1Open the app, find New Invoice
- 2Search for the client, or create them
- 3Add a line item, type the description
- 4Enter quantity, then the rate
- 5Set the issue date and the due date
- 6Pick the tax treatment
- 7Apply the early-payment discount
- 8Preview, then find the send button
Invoice Acme Studio for six hours of design at $120 an hour, net 14, and send it.
Created INV-0482 for Acme Studio. Six hours of design at $120 an hour, $720, due in 14 days. Your 2% early-payment discount applies if they pay within 10 days. Sent.
The same invoice, either way. One takes about two dozen interactions, the other takes one sentence.
Pro Tip
The moment that matters is not the typing you saved. It is that the invoice went out the same day the work finished, instead of waiting nine days for you to feel like opening the app. Getting paid is mostly a function of when you asked.
Listen: Two Minutes of Running a Business by Talking
Reading a transcript of a conversation is not the same as hearing one. Below is a real recording of the kind of exchange this article is describing: a freelancer clearing their receivables at the start of the day. It covers the four things that actually matter in billing, which are knowing what you are owed, deciding who to chase, setting the rules for how hard to chase them, and getting a new invoice out the door.
The assistant voice is the same one the product uses. Play it, follow along in the transcript, and notice how little of it sounds like operating software.
Example conversation, produced for this article. The figures are illustrative, not a customer’s real account.
What just happened in those two minutes
- A full receivables position was read out loud, aged by client, without opening a dashboard.
- Three dunning ladders were configured and started, across email and SMS, with quiet hours and a reminder cap respected.
- The dispute and failed-payment rules were confirmed, so nothing embarrassing gets sent to a client who is already unhappy.
- A new invoice was created, priced, discounted, and sent.
- Last month’s recovery and days-to-payment were reported back as plain numbers.
All of that works hands-free too. See voice invoicing.
Dunning Is an Emotional Problem Disguised as a Workflow
Dunning is the structured sequence of reminders that sits between an invoice going overdue and anything as serious as collections. Every business textbook agrees on the mechanics: remind early, remind repeatedly, escalate the tone gradually, keep it professional. Almost nobody does it. That gap is not laziness and it is not ignorance.
The reason people stop chasing is that each reminder costs something to write. You are asking someone you like, and probably want to work with again, for money they already owe you. Every message is a small negotiation with your own discomfort: too soft and it gets ignored, too firm and you have damaged a relationship over an invoice that was probably just forgotten. So the second reminder does not get sent, and the invoice quietly ages.
The asymmetry that decides who gets paid
Most overdue invoices are not refusals. They are an email that dropped below the fold in somebody’s inbox. The businesses that get paid quickly are not the ones with better clients, they are the ones whose follow-up survives past the first polite nudge. Persistence is the whole strategy, and persistence is precisely what a human is worst at supplying.
This is why dunning is the single best fit for automation in the whole billing stack. A machine has no discomfort to negotiate with. It sends the fifth reminder as evenly as the first, at the hour you specified, in the tone you set, and it never once decides that today is not a good day to bring it up.
Example Two: What "Chase Them, But Keep It Friendly" Builds
One instruction, in one sentence, produces a complete multi-step campaign. Here is the ladder the assistant set up in the recording above, exactly as it runs on an invoice that has gone twenty six days past due.
A short, warm reminder to the billing contact with the invoice and a payment link. Assumes it was simply missed.
References the first message, restates the amount and the original due date, and asks for a payment date.
Switches channel, because an unanswered email is often an inbox problem rather than a payment refusal.
Automated follow-up stops and the invoice is handed to you with its full history, ready for a phone call.
If the client opens a dispute or a card payment is declined, the ladder halts immediately and the invoice moves to your escalation queue. Nothing is more damaging than a cheerful automated reminder landing in the inbox of a client who has just told you the work was wrong. Follow-up resumes on its own once the issue is settled.
A single dunning ladder. It stops the moment the invoice is paid, disputed, or a payment fails.
Pro Tip
Set your quiet hours and your reminder cap once, at the start. Dunning that respects evenings, weekends, and a sane ceiling on messages reads as professional persistence. Dunning without those limits reads as harassment, and it will cost you the client you were trying to bill.
Why Conversation Is the Right Interface for Collections
Automated dunning is not new. What has always been wrong with it is the setup. Configuring a reminder sequence in a traditional tool means a rules builder, a template editor, a cadence screen, and a set of decisions you have to make in the abstract before you have ever seen the thing run. It is enough work that most people never turn it on, which means the feature that would have collected their money sits switched off.
Describing the same policy out loud takes eight seconds. "Chase them, but keep it friendly" carries the cadence, the tone, and the escalation in one clause, because that is how humans have always specified this kind of thing to an assistant. The reason conversational AI matters for dunning is not that the follow-up is smarter. It is that the follow-up actually gets switched on.
What the conversation is really carrying
- Tone, without a template editor. "Keep it friendly" and "be firm, they are 60 days out" produce genuinely different copy.
- Exceptions, in passing. "Not Vertex, they are a good client and they always pay in the last week."
- Scope, without a filter builder. "Everything over 30 days" or "just the design retainers."
- Policy changes at any time. "Stop chasing Northwind, I spoke to them" takes one sentence and one second.
- Answers on demand. "Did that work?" returns the recovery figure and the change in days-to-payment.
Five Arguments for Running Your Billing This Way
Plenty of tools now advertise AI. Very few of them let the AI actually operate the business. These are the specific reasons Invoice My Clients is built the way it is, and what each one is worth to you.
The billing gets done, because asking is free
Every unsent invoice and every unsent reminder is a decision that lost to friction. When the cost of acting falls to a single sentence, the decision stops being a decision. Invoices go out the day the work finishes and reminders go out on the day they were scheduled, which between them is most of what separates a business that gets paid quickly from one that does not.
The follow-up is relentless in a way you never will be
Autonomous dunning runs sustained, multi-step campaigns across email and SMS on every overdue invoice, escalating the cadence and tone as the debt ages and adapting to how each client responds. It respects your quiet hours and your reminder caps, and it hands you only the cases that genuinely need a person. You are not being replaced in the awkward conversations. You are being spared the eleven polite ones that should never have needed you. See how automated collections works.
It knows when to stop, which is the hard part
Anyone can build something that sends five emails. The valuable behaviour is restraint: an opened dispute or a declined card pauses that invoice automatically and routes it to your escalation queue, and follow-up resumes on its own once the issue is resolved. Automation that cannot stop is a liability with your clients. This one is designed around the stopping.
You can ask the business questions and get answers
Behind the conversation sits a live accounts receivable view: days sales outstanding, aging buckets, collection rate, and a probability-weighted cashflow forecast. You do not have to build a report to reach any of it. "Who owes me money", "what is my average days to payment", and "show revenue this quarter" are complete queries, and the answer comes back in the same breath as the question. See AI-boosted reports.
Voice means billing survives contact with a real workday
The assistant works in full voice, with speech recognition and lifelike speech back. If your work happens on a job site, in a van, between appointments, or anywhere that is not a desk, this is the difference between logging the hours and losing them. You finish the job, you say what you did, and the invoice is written before the tools are packed.
Underneath all five is one idea: the assistant is not a layer on top of the product, it is a way of driving the product. Anything you can click, you can ask for instead. And if you would rather build your own workflows on the same foundation, the MCP server exposes it to your own tools and agents. See the MCP server.
What It Costs to Try It
The AI billing assistant is included with every account, free tier included, so you can test the whole idea before deciding anything. AI is metered as a monthly message allowance that comes from your plan rather than being sold separately.
- Free, $0. 20 AI messages a month. Enough to see whether talking to your billing feels as different as it sounds.
- Starter, $15 a month, or $12 billed annually. 150 AI messages a month, for a solo freelancer.
- Professional, $29 a month, or $24 billed annually. 750 AI messages a month, and this is the first plan with autonomous AI dunning.
- Business and Agency, $59 a month, or $49 billed annually. 2,500 AI messages a month for teams.
Free and Starter run on Claude Sonnet. Professional and Business run on Claude Opus for the most capable responses. Autonomous dunning is available from Professional upward, which is the plan most people move to the moment they see what consistent follow-up does to their receivables.
See full pricingThe Honest Part: What It Will Not Do
Software that touches money and client relationships should be specific about its limits, so here they are. Being clear about this is not a caveat, it is the reason the automation is safe to leave running.
- It is not a human collections agency. It is an automated engine inside the app. Cases that need judgment are escalated to you rather than pursued.
- It will not chase a disputed invoice. Disputes and failed payments pause follow-up automatically and wait for a person.
- It will not exceed the limits you set. Quiet hours and reminder caps are hard boundaries, not suggestions.
- It does not reach across accounts. Every action runs in an authenticated session tied to your account, and data is never shared between tenants.
- It shows its work. Actions are executed in view as they happen, and invoice numbers link straight back to the invoice so you can check anything it did.
Getting Started Takes About Ten Minutes
There is no configuration phase to survive first. Create an account, open the assistant, and start by asking it something you already know the answer to, so you can see it working against your real data.
- 1Create a free account and import or add a couple of clients.
- 2Open the assistant and ask "who owes me money" to see it read your live receivables.
- 3Create your next invoice by describing it in one sentence instead of filling in the form.
- 4Turn on dunning with a single instruction, and set your quiet hours and reminder cap.
- 5Try voice mode once, hands-free, and decide whether you ever go back to typing.
Pro Tip
Start dunning on invoices that are already overdue rather than waiting for new ones. Most businesses have several thousand dollars sitting in invoices that stopped being chased after one email, and the first ladder you run is usually the one that pays for the plan several times over.
Stop clicking. Start asking.
Create a free account, open the assistant, and ask it who owes you money. It takes about a minute, and the answer tends to be motivating.
Free tier included with every account. No credit card required.
Conversational AI Billing FAQs
What is conversational AI in invoicing?
Conversational AI in invoicing is an assistant that operates your billing through plain language instead of forms. Rather than answering questions about the software, it takes real actions on your account: creating and sending invoices, logging billable time, adding clients, running reminder campaigns on overdue invoices, and answering questions about your revenue and receivables. In Invoice My Clients it works by text or by voice and runs on Claude.
Can AI handle dunning and chase overdue invoices for me?
Yes. Autonomous AI dunning runs sustained, multi-step follow-up campaigns on every overdue invoice, across both email and SMS, with the reminder copy written by AI and the cadence escalating as the invoice ages. It stays inside the quiet hours and reminder caps you set, pauses automatically when a client disputes an invoice or a payment fails, and escalates the cases that need a human to your queue. Autonomous dunning is available on the Professional plan and above.
Is AI dunning going to annoy my clients?
It should not, and the guardrails are what prevent it. You control the cadence, the tone, the quiet hours, and the maximum number of reminders per invoice. Follow-up halts the moment a client raises a dispute or a payment fails, so nobody receives a cheerful reminder about an invoice they have already complained about. Consistent, professional persistence within limits reads very differently from unrestricted automation.
Can I create an invoice just by talking?
Yes. Voice mode gives you full hands-free operation with speech recognition and lifelike speech back. You describe the client, the work, and the amounts in normal speech, and the assistant builds the invoice in real time and can send it. It is the fastest way to bill when you are on a job site or between appointments.
Which AI model powers the assistant?
It runs on Claude. Free and Starter plans use Claude Sonnet, while Professional and Business plans use Claude Opus for the most capable responses. AI is included in every plan as a monthly message allowance rather than being sold as a separate product.
Is my billing data secure when I use the AI assistant?
Yes. The assistant operates inside authenticated sessions tied to your account, and it never shares data across tenants. Every action it takes is shown as it happens and is traceable back to the invoice or record it changed, so you can audit anything it did.
References
Federal Reserve Banks. Small Business Credit Survey: Report on Employer Firms. https://www.fedsmallbusiness.org/survey
U.S. Small Business Administration. Manage your finances: invoicing and cash flow. https://www.sba.gov/business-guide/manage-your-business/manage-your-finances
Federal Trade Commission. Debt collection practices and the Fair Debt Collection Practices Act. https://www.ftc.gov/legal-library/browse/rules/fair-debt-collection-practices-act-text
Consumer Financial Protection Bureau. Debt collection communication rules. https://www.consumerfinance.gov/rules-policy/final-rules/debt-collection-practices-regulation-f/
Anthropic. Claude model family documentation. https://docs.anthropic.com/en/docs/about-claude/models
Related Articles
AI Invoice Generator: How Artificial Intelligence Is Transforming Small Business Invoicing
Discover how AI-boosted invoice generators are helping small businesses create invoices faster, reduce billing errors, and get paid sooner - outperforming QuickBooks, FreshBooks, and Wave.
How to Handle Overdue Invoices Like a Pro
Effective strategies for following up on overdue invoices while maintaining client relationships and protecting your cash flow.
Cash Flow Management in Business 101
Essential strategies for cash flow management in business every owner needs to know to stay profitable and grow sustainably.
Why Time Tracking and Invoicing Belong in One Application: The Business Case for Integration
Discover why freelancers and small businesses that unify time tracking and invoicing in a single platform capture more billable hours, get paid faster, and gain real-time profitability insights - backed by research.