When Voice or Chat AI Helps Intake, and When It Quietly Costs Cases

By Bobby Sawyer · Published 4 November 2026 · Last updated 1 November 2026

By Bobby Sawyer · Published November 4, 2026 · Last updated November 4, 2026

AI handles the simple, repeatable parts of intake well: routing a caller to the right practice area, answering basic questions about the firm, collecting a name and number after hours, and booking a consult once a person has qualified the case. It quietly costs cases when it does the parts that need judgment: deep qualification, callers who are scared or angry, facts that raise conflict or confidentiality issues, and complex document gathering. The test is simple. If a wrong answer would lose a good case or expose the firm to an ethics problem, a trained person should own that step. Measure AI the same way you measure people, by signed cases, not by calls handled.

For the record, Law Ops Forge answers intake calls with live people and does not use AI voice agents. Read this with that in mind. The aim here is to help you decide where AI fits in your own intake, not to argue it out of every step.

Which intake tasks can AI handle well?

• Routing. Sending a caller or chat visitor to the right practice area or office.
• Basic firm questions. Office hours, location, practice areas, whether the firm offers consults.
• After-hours capture. Taking a name, number, practice area and a short reason for the call, then handing it to a person first thing.
• Scheduling a qualified lead. Offering open consult times once a person has decided the case fits.
• Reminders. Consult reminders and document requests the firm has already approved, sent with the right consent.
• Summaries for staff. Drafting a call summary that a person checks before it goes into the file.

What these have in common is that the right answer is known in advance and a mistake is cheap to fix.

Which intake tasks still need a trained person?

• Deep qualification. Deciding whether a case fits takes follow-up questions that depend on the last answer. A missed question can turn a good case into a decline.
• Upset or frightened callers. Someone just arrested, injured, served or fired needs a person who can slow down, reassure and still get the facts. A caller who feels processed hangs up and calls the next firm.
• Conflict and confidentiality risk. Under ABA Model Rule 1.18, a lawyer who learns information from a prospective client “shall not use or reveal that information,” even when no client relationship follows. Intake has to stop at the right point, run a conflict check and avoid collecting more than it needs. That takes judgment.
• Deadlines and documents. Spotting a court date, a filing deadline or a missing record, and flagging it to an attorney the same day.
• Exceptions. Anything the script did not expect.

What ethics and compliance points apply to AI in intake?

The ABA’s first formal opinion on generative AI, Formal Opinion 512, says lawyers using these tools must “fully consider their applicable ethical obligations,” according to the ABA’s July 29, 2024 announcement. The announcement names competence, confidentiality, communication and fees. It also notes that other model rules extend similar protections to “former and prospective clients’ information.” Intake data is prospective client information, so where an AI vendor stores and uses it matters. Each state bar adopts its own version of these rules, so have your attorney check the states you serve.

Outbound calls are a separate issue. The FCC ruled that calls made with AI-generated voices are “artificial” under the Telephone Consumer Protection Act, per the FCC’s February 8, 2024 release. Its consumer guide on robocalls and texts puts it plainly: “AI-generated voice calls are illegal unless the consumer has agreed to receive them or the caller is exempt.” If an AI tool calls leads back, have your attorney confirm consent first. State rules can add more.

Supervision still applies. Whoever owns intake should be able to explain what the tool says, what it records and when it hands off to a person.

How should a firm test AI before full deployment?

1. Start with one narrow task, such as after-hours capture or scheduling for qualified leads.
2. Run it alongside people, not instead of them, for a set trial period.
3. Read every transcript in the first weeks. Look for wrong answers, missed handoffs and callers who gave up.
4. Test the hard calls on purpose. Have staff play an upset caller, a caller with a court date tomorrow and a caller who names the other side.
5. Check the handoff. Every AI conversation should reach a named person with the source, contact details and reason for the call intact.
6. Decide in advance what failure looks like, and turn the tool off if it shows up.

How do you measure whether AI is helping or hurting conversion?

Compare the same numbers before and after, split by whether AI touched the lead:

• Abandonment: callers or chat visitors who leave before giving contact details.
• Speed to first response from a person.
• Qualification rate and consult booked rate.
• Consult to signed rate.
• Signed cases and cost per signed case by source.

Calls handled or chats completed are not the goal. If AI handles more conversations and signed cases fall, it is costing you cases. Also check Local Services Ads responsiveness. Google’s About ad rankings page lists “Your responsiveness to customer inquiries and requests” among the factors in how likely your ad is to result in a lead.

What is a sensible way to start?

Pick the lowest-risk task where your firm loses leads today, usually after-hours capture or scheduling. Give AI that one task, keep a person on qualification and on anyone who is upset, and review the numbers monthly. Expand only when signed cases hold or improve.

Frequently asked questions

Which intake tasks can AI handle for a law firm?

Routing callers to the right practice area, answering basic questions about the firm, capturing contact details and a short reason for the call after hours, scheduling consults once a person has qualified the case, sending approved reminders, and drafting call summaries that a person checks.

Which intake tasks should a person handle instead of AI?

Deep qualification, upset or frightened callers, anything that raises conflict or confidentiality issues for a prospective client, spotting deadlines and missing documents, and any exception the script did not expect.

How do you know if AI is hurting intake conversion?

Compare abandonment, speed to first response from a person, qualification rate, consult booked rate, consult to signed rate, and signed cases by source for leads AI touched versus leads it did not. If AI handles more conversations and signed cases fall, it is costing cases.

If you want a clear view of which intake steps at your firm need a person and which do not, start with a free intake and sales audit.

Related reading

• Automated Legal Intake: What AI Handles Well, and What It Doesn’t
• After-Hours and Weekend Intake Coverage
• Criminal Defense Intake: The Qualification Questions
• Live Agents: Real People Answering Your Firm’s Phone