Law Firm Accounts Receivable: AI-Drafted Chasers for Overdue Bills

A small law firm can use AI to chase overdue client bills by feeding it only an export of issued invoices (number, amount, due date, days overdue), drafting tiered reminder emails from approved wording, and having a partner approve each one. Client money in the trust or client account stays in your accounting system, never in the tool.

That separation is the whole safety case. The AI has no business knowing what sits in a client's trust balance, and a chaser that mentions it, or suggests drawing the fee from it, breaks the line your conduct rules draw between your money and theirs. What follows is a routine that keeps that line visible: what to export, how to tier the emails, which prompts to use and what a partner checks before anything is sent.

Follow me on Instagram@sagnikteaches

Why overdue bills are a good AI job, and a risky one

Chasing is repetitive and awkward. Fee-earners dislike it, so it slips; the bill ages from 30 days to 90, and the longer it ages the less likely it is to be paid in full. Drafting a polite, consistent reminder is a task AI does well, and an ageing report is a sort and a sum.

Connect on LinkedInSagnik Bhattacharya

The risks are specific to a law firm. Billing emails sit next to confidential matters. A client may be disputing a bill or have a complaint in progress. A chaser that is too sharp can turn a slow payer into a complaint, and one that is too soft teaches clients that late payment has no consequence. And any hint of drawing on client funds without the right authority is a conduct problem, not a collections problem. For a view of where AI fits in the wider practice, the first 90 days plan for a small law firm puts billing in context.

Subscribe on YouTube@sagnikteaches

What goes into the tool, and what never does

Put that table into your billing procedure. It also answers the first question a regulator or insurer would ask about AI in the billing process: what can the tool see? Use a business plan where the provider does not train on your content by default, and keep the export small enough that a leak would reveal little.

Step one: export and build the ageing report

Export from your billing system a list of unpaid invoices with: invoice number, client reference, issue date, due date, amount outstanding and fee-earner. No ledger movements, no trust data. Remove columns you do not need before it goes anywhere near a tool.

An ageing report sorts those invoices into bands, usually 1-30, 31-60, 61-90 and over 90 days past due. You can build it in a spreadsheet with a formula for days overdue, and ask AI to check the grouping and write the summary, or ask it to produce the report from the pasted rows. Either way, verify the totals against your billing system, because a model can miscount a column.

Below is an export of unpaid invoices from a small law firm. Columns: invoice no, client ref, due date, amount outstanding, fee-earner. Today's date is [date].
1) Add a column for days past due.
2) Group into bands: 1-30, 31-60, 61-90, over 90.
3) For each band give the number of invoices and total outstanding.
4) List the five largest outstanding invoices and the five oldest.
5) Flag any client ref with more than one unpaid invoice.
Do not add or infer any data not in the export. Show your arithmetic for the band totals.
Data: [paste]

A filled-in ageing summary

Say a five-lawyer firm exports 38 unpaid invoices. The illustrative output: 1-30 days, 17 invoices, $21,400; 31-60 days, 11 invoices, $18,900; 61-90 days, 6 invoices, $14,700; over 90 days, 4 invoices, $11,200. Total outstanding $66,200. Two client refs have three unpaid invoices each. The managing partner checks the bands against the billing system, finds the 31-60 total is $400 out because of a credit note, corrects the figure and notes that the model did not know about the credit note. That is the right division of labour: the AI saved the sorting, the human caught the thing only the system knew.

Step two: decide the tiers and who decides exceptions

Tiers keep the tone proportionate. Agree them once, as a partnership, and write them down.

1. Tier 0, before the due date (optional): a courtesy note with the invoice and payment details a few days before it falls due.
2. Tier 1, 1-14 days overdue: a friendly reminder assuming an oversight.
3. Tier 2, 15-45 days overdue: a firmer note stating the amount, the original due date and asking for payment or a call by a set date.
4. Tier 3, 46-90 days overdue: a partner-signed message offering a payment plan or asking about any dispute, and stating the next step the firm may take.
5. Tier 4, over 90 days: no AI-drafted email without partner discussion. Options include a call, a formal letter, stopping work where the engagement terms allow, or referral to your collections process.

The bands are illustrative; set yours from how your clients actually pay and from your engagement letter. Always write down exceptions: clients in a dispute, in the middle of a hearing, in hardship, or with a complaint open. Those clients are removed from automatic tiers and handled by the partner. The client status updates tutorial helps here, because clients who understand where their matter stands dispute fewer bills.

Step three: the chaser prompts, tier by tier

Save the rules in a ChatGPT or Claude Project, or a Gemini Gem, so every drafter uses the same boundaries. Google is replacing Gems with skills, so a firm on Workspace should expect that move in March 2027 (personal accounts switch in November 2026).

You draft payment reminder emails for a small law firm. A partner approves each email before sending.
Rules:
- Use only the facts given: client first name, invoice number, amount, original due date, days overdue, how to pay, fee-earner name.
- Never mention the matter's subject, advice given, trust or client account funds, or any other client.
- Never suggest taking payment from money held for the client unless I give you that wording, and I will not.
- Never threaten legal action, interest, credit reporting or stopping work unless the tier instructions say so.
- Plain, courteous, short. Under 120 words. No exclamation marks.
- If any fact is missing write [CONFIRM: what].
Tier: [1 / 2 / 3]. Tier instructions: [paste]
Facts: [paste]

For Tier 1 the instructions might read: "assume an oversight; attach the invoice; ask them to let us know if there is a query". For Tier 2: "state the amount and original due date; ask for payment or a call by [date]; say we are happy to discuss a payment arrangement". For Tier 3: "signed by the partner; ask whether there is a concern with the bill; offer a plan; say that if we do not hear from them by [date] we will need to consider next steps under our engagement terms". Keep the last phrase vague unless the engagement letter spells out what the firm can do.

Sample output and what to fix

A Tier 2 draft, illustrative: "Dear [first name], invoice 1042 for $2,850 was due on 3 September and is now 24 days overdue. I have attached a copy and our payment details. Could you let me know by Friday 2 October when we can expect payment, or call me if anything about the bill needs discussion? We are happy to talk about a payment arrangement. Kind regards, [fee-earner]." What the partner checks: the amount and dates match the system, the matter is not named, there is no mention of client funds and nothing reads as a threat. What they fix: if the model added "as you know, we have been holding funds on your account", delete it. It is false comfort at best and, at worst, a statement about client money that the billing prompt should never produce.

Keeping trust accounting outside the tool

Write this section into your billing procedure. It is short and it is the most important page.

• The AI never receives trust or client account data, exports or screenshots.
• No chaser mentions client money, balances, transfers or "paying from funds held".
• Any move of money from client account to office account follows your conduct rules, with the right authority and records, outside the AI process, by the people authorised to do it.
• The ageing report covers billed fees only. If a client's invoice could be settled from funds held for them, a partner or the account's authorised person decides whether and how, and records it in the accounting system.
• Staff are told in writing that a model must never be asked how to handle client money. Questions go to the partner responsible for accounts compliance.

Rules about client money differ between jurisdictions and regulators, so I am describing the principle rather than a rule. Read your own accounts and conduct rules, and ask your regulator's helpline or your compliance partner if in doubt. The safest design is that nothing the AI touches can possibly be mistaken for a client-money record.

A realistic mistake and how it showed up

Imagine a firm pastes a billing export that includes a "balance held" column from its practice management system, because the report was one click. The model then writes, in a Tier 2 chaser: "We note you have funds on account which could be applied to this invoice." The associate who approves it is busy and the sentence reads helpfully. It goes out. The client, who did not know the firm could apply funds, complains, and the firm now has a conduct question and a damaged relationship over a $900 invoice. The cause was an export with one column too many. The fix is to build a saved report that cannot include that column, and a partner rule: any chaser that mentions funds is rejected on sight.

The other tiers in full: sample drafts and edits

The Tier 2 example above is the middle of the range. Here are the two ends, labelled illustrative, with what a partner would change.

Tier 1 draft. "Dear [first name], a quick note that invoice 1042 for $2,850, issued on 20 August, was due on 3 September and I cannot see a payment yet. I have attached a copy in case it went astray. If you have already paid, thank you and please ignore this. If there is any question about the bill, just reply and I will help. Kind regards, [fee-earner]." The edit: nothing to cut. The draft assumes an oversight, offers help and makes no demand. Resist improving it.

Tier 3 draft. "Dear [first name], I am writing myself because invoice 1042 for $2,850 is now 58 days overdue and I have not heard from you since our last reminder. I would like to understand whether something about the bill concerns you. If it is a question of timing, we can agree a payment plan; tell me what would be manageable and I will confirm it in writing. If I do not hear from you by 13 November, I will need to discuss next steps with you under the terms of our engagement letter. Kind regards, [partner]." The edit: the last sentence must match what the engagement letter really allows. If the letter does not say what the firm may do, the line becomes "I would like us to speak before then".

Before and after: a harsh chaser softened

A fee-earner's own draft: "Your account is seriously overdue. Pay immediately or we will stop acting." After the style rules: "Invoice 1042 is now 58 days overdue. I would like to find a way forward with you, whether that is payment this week or a plan you can manage. Could you call me by Friday?" The second keeps the urgency and drops the threat, which a partner can still add by phone if the situation calls for it.

What your engagement letter should say, so the chasers have teeth

A chaser is only as strong as the terms behind it. Before you automate anything, read your standard engagement letter and check that it states: when invoices fall due, whether interim bills are issued, what happens if bills go unpaid (including whether work may be paused, subject to your professional obligations), whether interest or charges apply and how a billing query should be raised. If any of those are missing, fix the letter first. AI can help you draft clearer wording, but a partner or your insurer's guidance should decide it, and the text should be checked against your conduct rules. The tutorial on what an AI consultant does for a law firm touches on where such review sits.

A quick sum: what the routine is worth

Illustrative figures for a five-lawyer firm billing about $60,000 a month with 38 unpaid invoices at any time. Suppose chasing currently takes the office manager and partners roughly six hours a month, mostly drafting and finding the right words. A templated, AI-drafted routine with a partner reading each batch might take two hours. That is four hours a month saved, but the larger effect is on timing. If bills in the 31-60 band are chased at day 15 instead of day 45 and the average payment arrives ten days sooner, the firm carries roughly a third of a month less in debtors. For a firm billing $60,000 a month, ten days is about $20,000 less money waiting, which matters more to cash flow than the four hours. Measure your own figures before you quote any to partners.

Billing habits that make chasers rarer

• Send invoices promptly. A bill issued a month after the work lands as a surprise.
• Describe the work plainly. Ask AI to rewrite internal narratives into client-readable lines, with a partner check for confidentiality.
• Offer easy payment. A payment link and clear bank details on every invoice, held in your billing system, not in the AI tool.
• Update estimates. A short mid-matter note saying costs so far and the next stage's estimate heads off most disputes.
• Agree interim billing and deposits. Put the schedule in the engagement letter.

Handling replies, disputes and payment plans

Chasers produce replies. Plan for them. Most will say "will pay Friday" or "sorry, missed it", and a person can answer in a line. Others raise a dispute. That moves the invoice out of the chaser routine and into your complaints or fee-query procedure, with a partner owning it and the dispute noted against the invoice. Do not use AI to argue with the client about the bill. AI can help summarise the billing history for the partner, but the explanation to the client is the partner's.

For payment plans, agree the terms in writing and have the system, not the AI, track them. A simple email confirming amount, instalments and dates, signed off by the partner, is enough. Put the plan on the ageing report as "on arrangement" so it is not chased as though nothing had been agreed.

Fix the causes, not only the symptoms

Overdue bills often start earlier than the chaser. Clients who did not expect the figure; invoices issued months after the work; descriptions that mean nothing to a layperson; no payment link; no estimate update during the matter. Once a quarter ask AI to group the reasons recorded against late invoices, using only your own short notes, and see what comes up. If a pattern appears, change the process. Recovering unbilled time earlier, as described in the AI time capture tutorial, gets bills out sooner and reduces surprise, and a clear costs update mid-matter does more than any reminder.

How to check it worked

Measure four numbers before you start and again after three months. First, total over 60 days, as a share of total outstanding. Second, average days to payment. Third, the number of disputes or complaints about billing. Fourth, the partner's time spent on chasing. If total over 60 days falls and disputes do not rise, the routine is working. If disputes go up, your tone or your timing is off, and the usual culprit is a Tier 2 or Tier 3 message sent to a client who had never been told to expect the bill.

Also test the safety rules. Once a month have a partner read ten sent chasers and ask: any matter detail, any mention of client funds, any threat that the engagement letter does not support? One failure is a training moment; a repeat means the prompt or the export needs to change. If you would like help setting up the export, the prompts and the approval step around your billing system, that is what my AI implementation consultation covers, and the costs side is covered in the legal AI pricing tutorial.

Questions small firms ask about AI and overdue bills

Can AI see our trust or client account ledger?

It should not. Keep client money records in your accounting system and give the AI only an export of billed invoices, amounts and due dates. Mixing the two risks confusing fee income with client money, which professional conduct rules treat very differently. Check your own rules and keep the separation written down.

Should the chaser name the matter?

Usually only by reference number or a neutral description. A bill's subject can reveal confidential information, and emails get forwarded or read on shared devices. Say 'invoice 1042 for your matter' unless the client has agreed to more detail.

Can AI decide which clients to chase and how hard?

It can sort by age and amount, but a partner decides who gets a gentle note, a call or a pause. Some clients are mid-case, in distress or disputing the bill. A rule based only on days overdue will get those wrong.

Do the chasers need partner approval?

Yes at first, and for any client the partner marks sensitive. Once the templates are proven, routine first reminders can go out on a fixed schedule with a weekly partner review. Anything past the second tier should always have a person's eyes on it.

Further reads

• Family Law Client Communication With AI: Calm, Consistent Emails — Calm client emails for another sensitive area of legal work.
• How Much Does AI Cost a Small Accounting Firm? — How a similar-size firm budgets for AI.
• How Small Law Firms Use AI to Draft Letters and Routine Documents — Routine letters with the same partner-review step.
• ChatGPT or a Legal AI Tool: Which Should a Small Firm Use? — Whether a general chatbot or a legal tool suits the firm.
• AI Readiness Checklist for Accountants, Solicitors, Consultants — Check readiness before adding AI to billing.
• AI Legal Bill Review: Fix Time Narratives Before Invoices Go Out — A pre-bill routine for small firms: tidy narratives, screen for privileged detail, test client billing guidelines, then a partner signs off exceptions only.
• Client Money Checklist for Sole Practitioners: Monthly Controls With AI — A monthly client money routine for one-person practices, with exception lists and reminders AI can prepare and the checks only you can do.
• Trust Account Reconciliation for Law Firms: What AI May and May Not Do — The three-way trust reconciliation step by step, with exactly where AI may flag problems and which checks must stay with a named person.
• Credit Control Software for Small Distributors: AI Features to Compare — Chaser, Satago and built-in accounting reminders compared for a small distributor, with monthly costs, limits and a checklist of AI features to ask about.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: general professional-conduct principles on keeping client money separate and on confidentiality; check your own bar or regulator's accounts and conduct rules, which are the authority for this topic.