AI Intake for Law Firms: Conflict Checks and Ethics

Short answer

Yes, a law firm can use an AI receptionist for client intake, but the lawyer stays responsible for everything it says and collects. The Oregon State Bar’s Formal Opinion 2026-208, written specifically about chatbots and AI agents used for marketing and intake, answers “Yes, qualified”: the lawyer must understand how the tool stores information, protect what prospective clients share, enter them promptly into the conflict system, make sure the agent cannot be mistaken for a lawyer or create a lawyer-client relationship by accident, and supervise the vendor. Canada’s Model Code points the same way: a duty of confidentiality is owed to anyone who seeks a lawyer’s help, even without a retainer, and the lawyer remains fully responsible for delegated work. In practice that means an intake agent should collect the parties’ names before the story, never give advice, introduce itself as an AI assistant and hand anything sensitive to a person. This guide walks through each stage of intake, the rules behind it and the questions to ask a vendor. It is general information, not legal advice.

What regulators say so far

Bar guidance on generative AI has mostly addressed drafting and research. Intake is different: the AI talks directly to the public, often before anyone at the firm knows the person exists. Oregon’s Formal Opinion 2026-208, approved by the Board of Governors in February 2026, addresses exactly that situation and concludes that lawyers may use chatbots and AI agents for intake, provided they keep competence in the technology, implement safeguards and meet their professional obligations.

The opinion applies Oregon’s Rules of Professional Conduct: RPC 1.1 (competence), 1.6 (confidentiality), 1.18 (prospective clients), 5.3 (supervision), 7.1 (communications) and 7.3(a) (solicitation). Lawyers elsewhere should check their own jurisdiction’s rules and opinions. In Canada, the Federation of Law Societies publishes a Model Code and an interactive version that links each rule to the matching or related rule of every law society; in Quebec, the Barreau du Québec’s practical guide on generative AI says AI can improve intake for new or prospective clients without replacing legal advice.

Intake stage by stage: what the AI may do

Stage	What an AI agent can do	What stays with the lawyer or staff
First contact	Answer calls, texts and web messages around the clock; introduce itself as the firm’s AI assistant; say it cannot give legal advice	Approving the greeting and disclaimer wording
Conflict information	Collect the caller’s name, the other parties’ names and the practice area, before any details	Running the conflict check and deciding whether to proceed
Matter details	After the check, collect basic facts and documents with closed questions by practice area	Assessing the matter, its merits and any deadline
Consultation	Book in the right lawyer’s calendar, send reminders and the document list	Interviewing the prospective client before taking the matter
Fees and engagement	Answer from the firm’s published fee schedule	Quoting the matter, the retainer, and any engagement letter
Distress or danger	Follow the firm’s written instruction and hand off immediately; direct anyone in danger to 911	Judgment, and any decision to contact the person

Grid by ZeniTech, drawn from Oregon Formal Opinion 2026-208 and the FLSC Model Code. Adapt it to your jurisdiction.

Conflict checks: names first, story later

Prospective clients are protected even if they never hire you. Oregon’s opinion explains that, under RPC 1.18(b), lawyers may not use or reveal information a prospective client communicated through a chatbot except as permitted for former clients, and that under RPC 1.18(c), receiving disqualifying information through a chatbot can prevent the lawyer from acting against that person in the same or a substantially related matter. Its advice: enter prospective clients into the conflict management system promptly.

Canada’s Model Code reaches the same place by another road. The commentary to rule 3.3-1 says a lawyer owes a duty of confidentiality to anyone seeking advice or assistance on a matter, even without a retainer, and should be cautious about accepting confidential information on an informal or preliminary basis, because it may prevent the lawyer from acting for another party in the same or a related matter. Rule 3.4-1 bars acting where there is a conflict of interest, except as the Code permits. In Quebec, the Code of ethics of advocates protects the prospective client’s right to choose a lawyer (s. 24) and assesses conflicts across all clients of the firm (ss. 71 and 72).

The design consequence is simple: the intake agent asks for the parties’ names and the practice area first, and holds back the detailed story until your team has cleared the conflict check. That is the single most useful setting in an AI intake script.

Don’t let the agent sound like a lawyer

Accidental relationships: Oregon notes, citing the Florida Bar, that a chatbot that sounds like a lawyer could lead a person to believe a lawyer-client relationship has formed, with the duties that follow. It recommends clear disclaimers and that a responsible lawyer be notified and promptly review the conversation.
Identify the bot: the opinion notes that more than one jurisdiction has opined that a chatbot that does not identify itself as a chatbot could be a misleading communication.
No advice, no promises: lawyers must make sure the chatbot does not mislead about expertise, experience or past results, and does not attempt to give legal advice.
Vulnerable callers: the opinion flags solicitation rules where a person’s physical, emotional or mental state may be discernible to a human but not to a chatbot, and recommends that lawyers review such conversations and interview each prospective client before proceeding.
Canada: marketing must be demonstrably true, accurate and verifiable, and not misleading (Model Code rule 4.2-1).

Supervising the vendor and the agent

Oregon treats the duty as supervising the service provider rather than treating the AI itself as an assistant. Before using a chatbot, lawyers must confirm that information communicated through or stored by it is reasonably secure, which may include encryption, keeping client information out of training data, isolating each client’s information, and checking the other software products and vendors the tool relies on. They must also be able to monitor the chatbot’s communications and modify its responses.

Autonomous AI agents raise the stakes: an agent that can draft and send an email, a letter or a text on its own could, for example, send an engagement letter the lawyer never intended. Oregon’s advice is to subject autonomous actions to lawyer approval.

In Canada, the Model Code makes the lawyer fully responsible for all business entrusted to them and requires direct supervision of staff and assistants to whom tasks are delegated (rule 6.1-1). Under PIPEDA, the firm stays responsible for personal information sent to a vendor for processing and must protect it by contract (Schedule 1, principle 4.1.3). In Quebec, the Barreau’s guide says entering information protected by professional secrecy into an open, publicly accessible AI system is itself a breach.

An intake script that follows these rules

1. Greeting: “Hi, I’m [firm]’s AI assistant. I can take down your request and book a consultation. I can’t give legal advice.”
2. Practice area, with a closed list matching what the firm actually does; anything else is politely referred elsewhere and logged for a person to review.
3. Names: the caller’s, and the other parties’. No details yet.
4. Pause: the consultation is held subject to the conflict check, and the caller is told someone will confirm.
5. After clearance: basic facts with closed questions, deadlines the caller already knows about, and the document list.
6. Booking in the right lawyer’s calendar, with a neutral confirmation text that does not name the matter.
7. Hand-off rules: distress, danger (911), a deadline that looks imminent, or any request for advice goes to a person right away.
8. Review: a responsible lawyer reads new conversations promptly at first, then on a set schedule, and corrects the script.

Ten questions to ask an AI intake vendor

1. Are our conversations or recordings used to train your models or anyone else’s?
2. Which other vendors (telephony, text messaging, language models, hosting) receive the data?
3. Where is the data stored and processed?
4. Can we set what the agent asks, and stop it from collecting details before the conflict check?
5. Can we read every conversation and change the agent’s answers ourselves?
6. Does the agent introduce itself as AI, and can we approve the exact disclaimer?
7. Can the agent take any action on its own (sending emails, documents, engagement letters), and can we require approval?
8. How are prospective clients’ names exported to our conflict system?
9. How long is data kept, and how do we get it back and have it deleted when we leave?
10. How quickly will you notify us of a security incident?

What ZeniTech does

Max is ZeniTech’s AI intake agent for law firms and notaries. According to its page, it answers by phone, text, email and on your website, day and night, collects facts and documents by practice area, books consultations in the right lawyer’s calendar, prepares a short summary before each consultation and sends reminders. It never gives legal advice, conflict checks stay with your team, it applies your confidentiality and professional secrecy rules, and every conversation can be reviewed and corrected. Your data is never used to train a third party’s model.

ZeniTech does not claim any compliance certification; whether a given setup meets your jurisdiction’s rules is your firm’s call. Pricing is a one-time setup, then a monthly plan; the full price list is on zenitech.dev/en/pricing.

Features from zenitech.dev/en/ai-agents/law-firms, read on October 2, 2026. Prices are published on zenitech.dev/en/pricing.

Frequently asked questions

Can a law firm use an AI receptionist for client intake?

Yes, with conditions. The Oregon State Bar’s Formal Opinion 2026-208 answers “Yes, qualified”: lawyers must understand how the tool works and stores data, protect what prospective clients share, supervise the vendor and monitor what the agent says. Check your own jurisdiction’s rules; in Canada, the Model Code’s duties of confidentiality and supervision lead to the same precautions.

Does information given to an AI intake chatbot create conflicts?

It can. Oregon’s opinion explains that disqualifying information a prospective client shares through a chatbot can prevent the firm from acting against that person in the same or a substantially related matter, and recommends entering prospective clients into the conflict system promptly. Canada’s Model Code warns lawyers to be cautious about accepting confidential information on a preliminary basis for the same reason. Collect names first, details later.

Can an AI intake agent run the conflict check?

It can gather what the check needs, such as the parties’ names and the practice area, and pass them to your conflict system. The check itself, and the decision to take the matter, should stay with your team. Max, ZeniTech’s intake agent, works that way: conflict checks stay with the firm.

Do we have to tell callers they are talking to an AI?

It is the safe choice. Oregon’s opinion notes that more than one jurisdiction has treated a chatbot that does not identify itself as one as potentially misleading, and Canada’s Model Code requires marketing to be true and not misleading. A short greeting such as “I’m the firm’s AI assistant and I can’t give legal advice” covers it.

Can an AI chatbot accidentally create an attorney-client relationship?

Oregon’s opinion says it is conceivable: a chatbot that sounds like a lawyer could lead a person to reasonably believe a relationship has formed. It recommends clear disclaimers and prompt review of each new conversation by a responsible lawyer, and approval before any autonomous action such as sending an engagement letter.

Is client information safe with an AI receptionist?

Only if the firm checks. Oregon lists what to verify: how the tool stores information, whether prompts are used for training, encryption, isolation of each client’s data and the other vendors involved. In Canada, PIPEDA keeps the firm responsible for information it sends to a vendor, and Quebec’s Barreau says entering secret information into an open AI system is a breach.

What should a law firm AI receptionist never do?

Give legal advice, assess the merits of a matter, quote fees for a specific matter, send an engagement letter on its own, collect a detailed story before the conflict check, or handle a person in distress without handing off. It should introduce itself as AI and send anything sensitive to a person right away.

How much does an AI intake agent cost for a law firm?

At ZeniTech, Max is billed as a one-time setup, then a monthly plan with reasonable usage included; the full price list is on zenitech.dev/en/pricing. A version that answers and captures enquiries goes live in 1 to 2 weeks; connected to your CRM, calendars and follow-ups, count 3 to 4 weeks.

Free consultation

Get a fixed price for your project in 30 minutes.

We look at what the project has to achieve, tell you what it costs, and tell you when you do not need it.

Book a call →+1 581-748-7017

Sources

Market ranges compiled on October 2, 2026. They are not a ZeniTech quote.

AI intake for law firms: conflict checks and ethics