Phone answering at a law firm is not the same job as phone answering at a trades business, and most of the difference is about what the person answering must not say.
A plumbing receptionist who oversteps creates an awkward conversation. A legal intake that oversteps can create an impression of representation, an unrun conflict check treated as run, or an off-hand comment about a claim’s prospects that the firm then has to walk back. The constraints are the design, not an afterthought.
There is also a constraint running the other way, and it is the reason firms care about this at all: a missed call at a law firm can be a claim that expires. Limitation periods — statutes of limitations, in the US — do not pause because nobody picked up. A caller who reaches voicemail on the last week of a limitation period and does not leave a message has lost something the firm cannot restore later. That is a materially different stake than a rescheduled service call.
For transparency: Reign AI builds an AI receptionist on human-grade voice AI. This page is written so a firm that reads it and decides to keep a human intake coordinator has still got value from it — and for some firms, that is the right answer.
The three hard lines
Any configuration for a firm has to enforce these, and none of them happens by default:
1. No advice, ever. Not “you probably have a case,” not “that sounds like it’s within the deadline,” not “you’d normally get something for that.” Assessing merits is practising law. The agent collects facts and says a lawyer will review them. Callers will push — often, and sympathetically — and the configuration has to hold anyway.
2. No conflict check, and no implication that one has happened. The agent takes the names — caller, adverse party, any related company — so a real check can be run quickly. It does not tell the caller the firm can act. “Someone will confirm whether we’re able to help” is the honest sentence, and it needs to be in the script rather than left to the model.
3. No implied representation. The caller should end the call knowing that details were taken and that the firm will review them. Not that they have a lawyer. The technical position is that a relationship forms when the firm agrees to act — but the risk is the caller’s belief, and belief is created by wording, so wording is where it is managed.
What it can genuinely do
Within those lines, a configured agent handles the part of intake that is repetitive and consistency-sensitive:
- Answer on the first ring, at any hour, including the evenings and weekends when someone who has just had a bad experience actually starts calling firms.
- Capture a structured intake — the caller’s name and contact details, the other side’s name, what happened, when it happened, whether any deadline or court date is known, whether they have spoken to another firm.
- Ask the date question every single time. “When did this happen?” is the field most likely to be skipped by a tired human and most likely to matter, because it is what lets someone at the firm triage urgency the next morning.
- Take the call while the intake coordinator is on another one. Firms lose a real share of enquiries to a busy signal at 10am on a Monday.
- Route by matter type — sending an employment enquiry somewhere different from a commercial one, if the firm works that way.
- Say what happens next, accurately. “Someone will call you back today” only if that is true.
The consistency is the underrated part. An intake that captures the same fields at 9am and at 11pm produces a queue a firm can actually triage, rather than a set of notes of varying usefulness.
What it should not be asked to do
- Screen for merit. Even soft screening (“we usually only take matters over a certain value”) should be handled carefully and explicitly, because to a caller it sounds like an assessment of their situation.
- Give a fee quote beyond a genuinely fixed consultation fee, if the firm has one.
- Handle a caller in crisis without an escalation path. Configure a route to a human.
- Be the only line for an urgent deadline. If a caller says a limitation period or a court date is close, that should escalate immediately rather than sit in a queue until morning.
Confidentiality — ask these before you sign anything
These questions apply to a human answering service exactly as much as to an AI one, and most firms do not ask them:
- Where do transcripts and recordings live, and for how long? A retention period you cannot state is a retention period you cannot defend.
- Who receives the notification? If it lands in a shared inbox everyone at the firm reads, that is a decision, and it should be a deliberate one.
- Is the call recorded, and are callers told? In Canada, informing the caller at the start is the expectation — the Privacy Commissioner’s guidance for businesses is the place to start on what collecting that information commits you to. Requirements vary by state in the US and by what the firm has committed to. Either way, the greeting is where it belongs.
- Can you delete a specific matter’s records on request?
A provider who has no answer to these is not set up for legal work, whatever their website says.
When a person is still the right front line
Plainly, because it is true for a real share of firms:
- Callers are frequently distressed. Family, criminal, and some injury practices are full of people having one of the worst weeks of their life. The judgement a skilled intake coordinator brings there is not reproducible by configuration.
- Intake IS the sales conversation. In practices where the firm is chosen on the first call, handing that call to an agent gives away the moment the decision is made.
- Volume is low enough that a person catches everything. If the firm’s calls are handled and nothing hits voicemail, there is nothing here to fix.
For those firms, the useful shape is narrow: a person on the front line during the day, and AI answering only for overflow and after-hours — the calls that were already going to voicemail. Nobody is replaced; what was falling on the floor gets caught, with the three hard lines enforced on exactly the calls where nobody was listening anyway.
That shape raises a separate question this page does not answer: new matters are only a fraction of what rings a small firm, and the rest of it — existing clients chasing a status update, opposing counsel, court staff, process servers — is a routing problem rather than an intake one. That side is covered in virtual receptionist for a small law firm, which assumes the three hard lines above rather than restating them.
How do you evaluate it?
Before any of this, decide whether the problem is worth solving: the cost of missed calls sets out how to measure what unanswered intake calls are costing, and for a firm the per-matter value makes that arithmetic move fast.
Call it and behave like the caller you are worried about. Push for an opinion on your situation — it should decline and keep collecting facts. Ask whether the firm can take your case — it should say that will be confirmed. Say a deadline is next week — it should escalate, not schedule. Then read the notification that arrives and ask whether someone at the firm could triage from it cold, the next morning, with no other context.
If the answer is yes, it works. If the answer is “I’d have to call them back to find out what this is about,” it does not — regardless of how the conversation sounded.
Background on the mechanics is in how AI receptionists work; the honest trade-offs against a human answering service are in the four-way comparison.
Common questions
Can an AI receptionist run a conflict check?
No, and it must not imply that one has happened. A conflict check requires searching your matter management system against parties, adverse parties and related entities, and interpreting the result — that is professional judgement, not a lookup. The agent collects the names so a real conflict check can be run quickly, and says explicitly that the firm will confirm whether it can act. Any setup that lets an agent tell a caller the firm can take their matter has skipped the step that exists to protect the firm.
Does an AI answering a call create a lawyer-client relationship?
The relationship is formed by the firm agreeing to act, not by a phone call being answered — but the risk is the impression created, not the technical position. That is why the agent should be configured to say plainly that it is taking details for the firm to review and that nothing is confirmed until someone from the firm follows up. A caller who hangs up believing they are represented is the outcome to design against, and it is entirely a configuration decision.
What about confidentiality and privilege?
Treat everything a caller says as sensitive from the first word, because it usually is. Practically that means knowing where transcripts and recordings are stored and for how long, restricting who at the firm receives the notification, and telling callers the call is recorded where that applies. These are questions to ask any answering provider — human or AI — before signing, and a provider who has not thought about them is answering a different kind of business.
Should the agent tell a caller whether they have a case?
Never. Assessing merits is legal advice, and an agent that ventures an opinion is both wrong about the law and a liability the firm did not agree to take on. The correct configuration collects the facts, states that a lawyer will review them, and moves on — even when the caller pushes, which they will.
Can it handle a caller who is distressed?
Partially, and this is where the honest limit sits. It can be configured to slow down, avoid pressing for detail, and escalate to a person rather than continuing to work through an intake list. What it cannot do is exercise the judgement a good intake coordinator brings to someone in real difficulty. For firms whose callers are frequently in crisis, that gap is a genuine reason to keep a person on the front line and use AI only for overflow.