Skip to content
be-human.ai
← Back to the journal

AI by industry

AI Intake for Bankruptcy Attorneys: Document Lists and Deadlines

How a bankruptcy practice can use AI for document checklists and deadline tracking, with attorney approval at every step and clear limits on legal judgment.

Be Human typographic cover with the title: AI Intake for Bankruptcy Attorneys: Document Lists and Deadlines
Conceptual illustration created with AI · Be Human

An AI agent can help a bankruptcy practice assemble document checklists, track the deadlines that appear in court notices, and follow up with clients about missing paperwork, as long as an attorney approves every checklist version and every date before it becomes a working commitment. The practical setup is narrow: the agent sorts, compares, drafts and reminds, while lawyers and paralegals decide what the rules require and what advice a client receives. This guide explains how to scope that work, which decisions stay with the firm, and how to tell whether the arrangement is working.

Key takeaways

  • Start with one intake path, such as the document list for a new consumer case, rather than the whole practice.
  • Every checklist version and every calendared date carries a source and an attorney confirmation.
  • The agent may collect, sort, compare and remind. It does not advise clients, choose strategy or sign filings.
  • Text inside a client's uploaded document is treated as material to read, never as an instruction to follow.
  • Review the agent on a recurring schedule against attorney corrections, so errors appear as a visible trend.

Why bankruptcy intake suits a bounded first agent

Bankruptcy intake produces a lot of repeated paperwork. Clients send partial sets of pay records, bank statements, tax returns, property documents and prior court papers, often out of order and sometimes as photographs taken at an angle. Staff spend time identifying what is missing, asking again and filing what arrives. That sorting and reminding is well bounded, which makes it a sensible place to start. The legal analysis that depends on those documents is not bounded in the same way, and the design should state that difference from the first day.

The same split appears in other legal intake work. Our article on family law intake, court dates and documents separates administrative sorting from attorney judgment in much the same way, and the process service logs guide shows how dated records can be kept accurate without anyone guessing at what they mean. Anthropic's guidance on building effective agents also favors simple, well-bounded patterns over broad autonomy, which suits a first engagement.

How do you build a document checklist the attorney can approve?

The checklist is the main artifact of the first engagement, because it defines what the agent is allowed to request. Written by the attorney, it turns general practice knowledge into a fixed list with reasons attached. A workable sequence looks like this:

  • Write one checklist for each case path the firm handles, and name the attorney as author of record.
  • Mark every item as required, conditional or reserved for attorney review, and write each condition in plain words.
  • Attach a one-sentence reason to each item so the client understands why it is being requested.
  • Give the agent only the approved version. It cannot add, remove or reword items on its own.
  • Stamp each checklist sent to a client with a version number, so later questions can be traced to the exact text that was sent.

The list also needs a maintenance rule. When a court changes a form or a trustee asks for a new category of document, the attorney revises the source, the version number changes, and clients who received the old list are told through the normal process. An agent that keeps using the old list will ask for the wrong items unless someone owns that update.

Which deadlines can an agent calendar, and which must stay with the attorney?

Deadline tracking is where the boundary matters most. A date printed on a court notice and a date that depends on a legal conclusion are different kinds of work. The agent can help with the first and should not settle the second. Use this guide to decide what happens to each deadline the agent encounters:

  • The date appears in a court notice or order already on file: the agent calendars it, cites the document, and an attorney confirms the entry.
  • The date must be computed from a rule: the agent applies only the rule text the attorney has approved, shows its arithmetic, and waits for attorney verification before treating the date as final.
  • The date depends on whether a rule applies, whether an extension is available or how a judge may read a filing: the agent flags it for attorney review and does not place it on the calendar as settled.
  • The source is missing, unreadable or contradicts another document: the agent escalates the conflict and makes no estimate.

Each calendar entry should show its source, the person who confirmed it and the time of confirmation. Those fields make a deadline auditable after the fact, and they are what a supervising attorney needs when a date moves. Our piece on property tax appeal firms applies the same source and confirmation pattern to assessment dates, which is a useful comparison when the firm designs its own fields.

What does the client see, and what must the agent refuse?

Clients should experience the agent as a clear, polite intake assistant. It explains the checklist, confirms what has been received and says when a person will respond. It can answer logistics questions, such as which file formats are accepted or where the upload area is, when the approved text covers the answer. It should not interpret the law for a client, estimate outcomes, recommend whether to file or comment on a creditor's position. Those questions go to an attorney with the conversation attached.

Uploaded documents need particular care. A client can submit a PDF containing sentences that look like instructions, such as a request to ignore earlier rules or to send the file to someone else. The agent must treat that content as data to read, not as a command. Access should be limited to what each task requires, following the authorization principles in the OWASP guidance, and outbound messages should wait for approval until the firm has enough history to justify a narrower rule.

What does this look like in a real intake?

Illustrative scenario: a new consumer client uploads six files from a phone on a Sunday evening. The agent matches four of them to the approved checklist, marks one bank statement as unreadable because the bottom part is cut off, and notes that a required item about prior filings is absent. It drafts a short message asking for a clearer copy and the missing item, then places that draft in the paralegal's queue instead of sending it. On Monday the paralegal approves the message with one wording change. The file reaches the attorney with a list of matched items, open items and the reason each open item is needed. Separately, when a court notice with a hearing date is uploaded, the agent places the date on the calendar with the notice cited, and the attorney confirms it the same morning.

How Be Human's managed approach handles this

Be Human scopes the first engagement around one role, here an intake and deadline assistant for a single practice area, and defines the memory and tools that role needs: the approved checklist library, the case file index and the calendar it may write to. Approval boundaries are written down before launch, so the attorney can see which actions require sign-off and which are read-only. The success criterion is observable, such as the share of checklists accepted at first attorney review and the time between a client upload and a human-confirmed status. Ongoing operations include monitoring of agent activity, recurring evaluations against attorney corrections, permission reviews of who can read each case file, maintenance of the calendar and document integrations, and incident response if a message or calendar entry goes out wrong.

How do you prepare for stronger assistants without giving up judgment?

More capable AI systems will be able to read more documents and draft more text, which makes clear boundaries more important, not less. The NIST AI Risk Management Framework is a useful reference for mapping risks, assigning owners and measuring performance over time. For a bankruptcy practice the preparation is concrete. Keep approved checklists and deadline rules in a controlled library, record who approved each change, log what the agent read and wrote, and decide in advance which tasks would send a design back for human review. Confirm the rules that apply in your jurisdiction, including your bar's guidance on supervision and confidentiality, before expanding any agent's reach.

A firm that writes down its intake procedures now builds a record that any later assistant can work from under the same supervision. If you want to find the first task worth handing to an agent in your practice, take the free Be Human AI diagnostic.

Frequently asked questions

Can an AI agent give a bankruptcy client legal advice? No. In this design the agent handles intake logistics, approved document checklists and reminders. Questions about eligibility, strategy, exemptions or the effect of a filing are routed to a lawyer, and the agent's instructions should say so in plain terms. Advice remains an attorney function under the rules that govern your practice.

Should the agent calendar deadlines directly? It can place a date on the calendar when the date comes from a court notice or order already on file, and the entry must cite that source. A date that requires computation or a legal conclusion should be flagged for an attorney instead of being treated as settled. Every calendar entry should also show who confirmed it.

What is the safest first task to automate? A document checklist for one case type is usually the most contained starting point, because the items are known, the output is easy to review and mistakes show up before any filing is affected. Choose a task where the attorney can compare the agent's output with their own work in a short review. The first task should also have a clear measure of success that someone checks.

How do we know the agent is working? Measure what the attorney can verify: how many checklists are accepted at first review, how often a person corrects a deadline entry and how long it takes from a client upload to a confirmed status. Review these figures on a recurring schedule and trace each correction to its cause. Improvements should be deliberate changes with a named owner, not silent edits to the agent's behavior.

Your next step

See where an AI agent fits in your company first.

Take the free Be Human AI diagnostic. A few questions about your work show which task to hand to an agent first, what it needs to know and where a person approves.

Be Human editorial guidance. Examples are illustrative unless explicitly attributed. Sources provide background; implementation choices depend on your workflow.

Less busywork. More room for you.

Bring us the work you wish you had time for.

Book a call to try a personalized AI agent free for 7 days ↗Or take the free AI diagnostic →