An AI teammate should not give legal advice, decide which matters your firm accepts, quote fees, or judge what a client's privilege covers, because those decisions belong to your attorneys and a well-configured teammate routes them to a person. The useful work sits on the other side of that line, where a teammate can sort intake messages, draft replies for approval, organize office notes, and keep follow-up from slipping through the cracks. The rest of this post explains where that line falls and how a Fort Wayne or Northeast Indiana law office can define it in writing before anything goes live.
Why does a law firm need a stricter boundary than most small businesses?
A landscaping company can let a teammate answer a routine question about service areas with comparatively little risk, because the worst outcome is usually a slightly awkward reply. A law office carries a different kind of exposure, since a reply about a deadline, a fee, or the strength of a claim can shape a person's decisions about money, family, or liberty. A careless message can also disclose confidential details, suggest that a lawyer has already accepted a matter, or cause someone to overlook a date that could affect their rights.
Because the stakes are higher, the boundary has to exist before the teammate touches a single message. Hey Button starts from the same principle it applies to every business, which is that access is set per task rather than through one global switch. A law office therefore does not have to choose between letting a teammate do everything and letting it do nothing. It can grant read-only access to a shared intake inbox, draft-and-approve access for replies to prospective clients, and no authority whatsoever over fees or case decisions.
That structure matters because a firm's risk is unevenly distributed across its inbox. Most intake messages are administrative in nature, a small share are urgent, and a few are legally sensitive in ways that require careful handling. A good setup treats each group differently. Administrative messages can be drafted quickly and approved in batches, while urgent and sensitive messages should reach a person directly without waiting in a draft queue. If you run a firm, see how Button works for law firms.
Can an AI teammate give a client legal advice?
No, and a responsible setup turns that limit into a written rule rather than leaving it to hope. Legal advice requires a licensed attorney's professional judgment about a specific person's facts, jurisdiction, and goals, and that judgment cannot be delegated to software. A teammate can gather those facts carefully and completely. It cannot replace the lawyer who weighs them, and it should never attempt to do so.
Consider an illustrative inquiry from a prospective client who writes, "Can my landlord keep my deposit if I broke the lease?" A useful teammate does not answer that legal question at all. Instead, it recognizes that the message asks for a legal conclusion, records the lease details, the move-out date, and any deadline the sender mentions, and then drafts a neutral acknowledgment. The draft explains that an attorney will review the question, and it neither predicts an outcome nor suggests that the firm will take the matter.
A staff member reads that draft before it reaches the sender. Under draft-and-approve, the teammate performs the sorting and prepares the first draft, while a person decides what the client actually reads. That division of labor is the core of the design, because the teammate keeps the work moving while the firm keeps the judgment where it belongs.
Can an AI teammate decide whether the firm takes a case?
No. Accepting or declining a matter depends on conflict checks, the practice's current capacity, the responsible attorney's experience with the issue, and a considered judgment about whether the firm can competently help. Those checks rely on records and professional relationships that a teammate should not weigh on its own, regardless of how capable it appears.
What a teammate can do is collect the details your intake process already requires. Those usually include the names of the parties, the identity of the opposing party, the general type of matter, the relevant jurisdiction, and the date of any deadline the prospect mentions. The teammate can then place the inquiry in a queue with a short written summary, so the person who makes the decision does not need to reread the entire message from the beginning.
One rule deserves particular care. The teammate should never promise a consultation time, a callback window, or a result unless a staff member has confirmed that commitment in advance. A draft stating "we will call you tomorrow" creates an expectation the office may be unable to meet. A draft stating "our team will review your message and follow up" is accurate, and it is the wording a firm should approve during setup.
What about client confidentiality and privilege?
This is the area where law firms should slow down and proceed deliberately. In July 2024, the American Bar Association issued Formal Opinion 512, its first ethics guidance addressing lawyers' use of generative artificial intelligence tools. The opinion explains that lawyers must fully consider their applicable ethical obligations before using those tools, including the duties of competence and confidentiality. Lawyers remain responsible for understanding how a tool uses data and for implementing adequate safeguards against unauthorized disclosure.
The opinion also indicates that lawyers should generally obtain informed client consent before placing client confidences into a generative AI tool, and that boilerplate consent language inside an engagement letter is unlikely to be sufficient. That point matters for any firm considering a teammate. A general statement that the office uses modern technology does not satisfy the duty, and a firm should not assume that it does.
In practical terms, a firm should be able to answer four questions in writing before a teammate reads anything sensitive. Where is the data stored, and who at the vendor can access it? Is the information used to train models? How long is it retained, and how is it permanently deleted? Ask Hey Button for those answers in writing, and review them with your ethics counsel before any privileged file reaches the teammate. The decision to share client information belongs to the firm and its attorneys, not to a configuration screen.
Treat the intake inbox as confidential from the first day of setup. Prospective clients frequently describe their circumstances in considerable detail in a first message, so intake correspondence can contain sensitive material even though no engagement exists yet. Privileged case files deserve a separate and later decision, made together with counsel, and they should never be loaded simply because a setup step appears to require it.
Which law firm tasks can an AI teammate handle safely?
The safest starting tasks are administrative and easy to verify. A teammate can perform several of them well when it operates within rules the firm has written down in advance:
- Sort the intake inbox by practice area and by urgency, with the firm reviewing those sorted results at read-only access first.
- Draft acknowledgments and lists of missing intake items, such as dates, addresses, or the names of other parties, for staff approval before sending.
- Track prospects who have not responded to a request for information, then prepare a follow-up draft that a person approves.
- Summarize an attorney's meeting notes into an action list, which the attorney then reviews, corrects, and adopts.
- Organize non-privileged office material, such as blank forms, mailing lists, and the calendar of scheduled office closures.
Each of these tasks carries a limit. A teammate can organize only what it is permitted to read. If intake arrives as walk-in visits, paper forms, or handwritten notes that never reach a shared inbox, the teammate has little to work with until those records are captured somewhere it can access. Summaries also present a particular trap, because a tidy summary of a deposition that no one has reviewed creates a new risk precisely because it looks authoritative. Every summary should name a reviewer, and that reviewer should be the lawyer responsible for the matter.
Begin with one task rather than all of them at once. Hey Button's own guidance is to start with the repetitive task that interrupts the week most, prove that it performs well on real examples from your office, and expand only after those results hold up under review. A narrow starting point makes both the successes and the mistakes far easier to evaluate honestly.
What should always go to a person right away?
Some messages should never sit in a draft queue, and their urgency should never depend on a teammate's judgment about whether they are urgent. Write these categories into the approval rules during setup, and route them directly to a staff member or an attorney:
- Any message that mentions a court date, a hearing, a filing deadline, or a statute of limitations.
- Any message from someone describing an emergency, a threat, a safety concern, or a custody situation. A person should direct that sender to call 911 or the appropriate emergency line, and staff should be alerted immediately.
- Questions about fees, retainers, payment plans, refunds, or billing disputes.
- Requests for a status update on an existing matter.
- Any message from opposing counsel, or from someone who says they are not a client but want legal help.
- Any message that asks directly whether the sender has a viable case.
That list may feel long for a small office, but each item marks a place where a draft that reads well can still be substantively wrong. Keeping those categories out of the ordinary draft queue costs far less than correcting a mistake after a client has relied on it. Routine messages will still require a person's approval as well, so the rule concerns speed for urgent matters rather than a general slowdown across the office.
How should a Fort Wayne law office set the rules before go-live?
Hey Button describes setup in three stages. Stitch is the intake conversation, in which the team maps how work reaches the office and which tools it already relies upon. Press is go-live, in which the approval rules are written down and the teammate is observed handling real examples. Mend is the ongoing tuning phase, in which the rules are revisited as the practice changes. For a law office, each stage requires meaningful involvement from an attorney, not only from an administrator.
During Stitch, list every written channel that brings in work, including the website form, the shared inbox, any text line, and any online scheduling page. Note which channels reach a person first and which ones already serve as the firm's record of a prospective client. During Press, sort the rules into three columns: topics the teammate may answer with approved text, topics it may only draft for review, and topics it must route to a person without drafting a substantive response. Then compare the teammate's drafts with what a staff member would have written. Each difference becomes either a rule to tighten or a draft to approve.
Local context matters as well. An office in Fort Wayne, Auburn, or elsewhere in Allen County often serves people who know its staff by name, so a message that sounds warm and familiar carries real value. A teammate can draft that tone, but the office must decide what the tone commits the firm to. A friendly sentence that quietly implies a particular result is still a commitment, and it deserves the same review as any other sentence the firm sends.
Where do the boundaries become hard to draw?
Some messages do not fit neatly into a single category. A prospective client may describe a legal problem and a scheduling need within the same message. An existing client may ask whether a particular filing has gone out, and answering accurately depends on notes that are privileged. A sender may describe a genuine emergency in calm language, or describe an inconvenience in urgent language. These gray areas are where a teammate is most likely to misjudge the situation, so the default should always be a person.
Defaulting to a person carries a cost. Replies will take longer, and some prospective clients will contact a competing office that responds faster. For a law practice, that trade is usually worth making for the sensitive categories, but it should be measured rather than assumed. Track how long staff take to review drafts and how long urgent items wait in the queue. If the review queue becomes the bottleneck, the usual remedy is to move low-risk categories to draft-and-approve, not to remove human review from the sensitive ones.
The honest summary is straightforward. A teammate can make an intake inbox easier to manage, keep prospective clients from going cold, and turn scattered notes into a clear task list. It cannot decide the law, choose which cases the firm accepts, or carry the professional responsibility that comes with a license. Those limits are not a flaw in the setup. They are the setup.
If you are weighing an AI teammate for your practice, begin with the intake questionnaire at start your Hey Button questionnaire, where the team maps your channels and asks which messages must always reach a person. Then read how Hey Button's setup works to see how access levels, approval rules, and review steps fit together before you decide anything.
Frequently Asked Questions
- Can an AI teammate answer questions from people visiting a law firm's website?
- Yes, within limits. A teammate can draft replies to website inquiries for staff approval, and it should send only wording your firm has approved. It should not address a specific legal problem, and it should route anything beyond general office information to a person.
- Does an AI teammate decide which cases a law firm accepts?
- No. Accepting or declining a matter depends on conflict checks, the firm's capacity, and an attorney's judgment. The teammate gathers the intake details your process requires, such as parties, matter type, jurisdiction, and deadlines, then places a short summary in a queue for the people who decide.
- Is it ethical for a lawyer to use an AI teammate?
- A lawyer may use generative AI tools, but must meet the duties of competence and confidentiality while doing so. ABA Formal Opinion 512, issued in July 2024, explains those obligations. Because state rules differ and guidance keeps changing, confirm current requirements with your state bar and your ethics counsel.
- What should a law firm ask before sharing client files with an AI teammate?
- Ask where the data is stored, who at the vendor can access it, whether it is used to train models, and how long it is kept before deletion. Get the answers in writing, review them with your ethics counsel, and obtain informed client consent where your obligations require it before any privileged file is shared.
- Which messages should always reach a person immediately?
- Messages mentioning a court date, hearing, filing deadline, emergency, threat, fee dispute, or a request for a case update should reach a person immediately. Write these categories into the approval rules during setup, so urgent matters never depend on a draft queue or on the teammate's own judgment.
Hey Button
Want a teammate that handles this for you?
Answer a few questions about your business and Hey Button will show you what an AI teammate would take off your plate, with you approving anything that goes out.




