Key takeaways
- Decide who answers first by call type and time of day, not with one blanket rule.
- Hand off with a structured summary so staff never ask the caller to repeat their story.
- Lawyers with managerial authority stay responsible for nonlawyer assistance, and bar guidance applies the same thinking to AI[1][2].
- Most employed people work weekdays, so evening and weekend coverage is often where AI adds the most[3].
Start with the calls, not the tool
Before choosing who answers what, sort a month of calls into types: new inquiries, existing clients, opposing counsel and courts, vendors, and spam. Each type has a different best owner. New inquiries need speed and consistency. Existing clients need someone who knows the file. Courts and opposing counsel need a person every time.
Then look at timing. In 2025, 81% of employed Americans worked on an average weekday but only 30% on an average weekend day[3], and 15% of wage and salary workers usually worked an evening, night or other non-daytime schedule[4]. Many people call a lawyer when they are off work, which is exactly when the office is empty.
Source: U.S. Bureau of Labor Statistics, 2025[3]
Show the data
| Item | Value |
|---|---|
| Weekday | 81% |
| Weekend day | 30% |
Four coverage models
| Model | Who answers first | Fits firms that | Main risk |
|---|---|---|---|
| Human first, AI overflow | Staff; AI picks up after a set ring count or when lines are busy | Have steady daytime staffing but peaks they cannot cover | Inconsistent experience if the ring threshold keeps changing |
| AI after hours | Staff in office hours; AI on evenings, weekends and holidays | Lose most leads outside office hours | Urgent after-hours matters still need an on-call person |
| AI first for new inquiries | AI for unknown numbers; staff for clients and known contacts | Get high volumes of new inquiries from advertising | Existing clients calling from a new number |
| Split by practice area | AI on high-volume lines, staff on complex ones | Run volume practices alongside boutique work | Callers who land on the wrong line |
Most firms start with overflow or after-hours coverage, then move to AI first for new inquiries once call reviews are consistently clean. Moving one step at a time keeps the cause of any problem obvious.
Choose the model by asking where calls are lost today, not where the technology looks most impressive. If your phone logs show that daytime calls are answered well and evening calls go to voicemail, after-hours coverage solves the real problem. If calls are dropped at 9 a.m. on Mondays and after a television spot, overflow is the better fit. If staff spend most of the morning on new-inquiry screening and existing clients wait, AI first for new inquiries frees them for the clients who need them. Write the reason for your choice into the plan, so you can test whether it held.
Routing rules that keep people in the loop
- Existing clients go to the assigned staff member or team inbox, never into new-client intake.
- Courts, opposing counsel and other lawyers go straight to a person, with no automated questions.
- Urgent triggers, such as an arrest, custody emergency, protective order, hearing within days or serious injury, get a live transfer to the on-call person and a callback within your stated window if nobody picks up.
- Callers who ask for a person get a transfer or a booked callback at once; do not try to keep them in the flow.
- Spanish-speaking callers and speakers of other languages go to a bilingual staff member or a configured language flow, and the record notes which was used.
- Anyone in danger hears emergency numbers first; details come second.
The handoff, step by step
Capture the minimum
Name, callback number, matter type, other parties for the conflict check, key dates and how the caller found the firm.
Tell the caller what happens next
Who will call, roughly when, and from what number, so the caller recognizes it and answers.
Send a structured summary
The same fields in the same order every time, delivered to the intake queue rather than a personal inbox.
Warm transfer when live
The staff member sees the summary before the call connects, greets the caller by name and confirms details instead of asking again.
Close the loop
Staff mark the outcome: booked, declined, referred or no answer. Unmarked handoffs are reviewed the next morning.
An SI Employee such as Ava is built around this handoff: it answers, follows your intake questions, routes urgent calls and sends staff a clean summary, so the person who calls back starts where the caller left off.
What changes for staff
Hybrid intake shifts staff time from first answers to follow-up, conversion and client care. Receptionists in the US earned a median of $18.27 an hour in May 2025[5], and benefits made up 30.0% of private employers' compensation costs in June 2026[6], so the real question is where that paid time does the most good. Usually it is confirming booked consultations, chasing documents and helping existing clients who need someone who knows their file.
- Intake specialists own conversion: confirming consultations, sending engagement paperwork and following up on no-shows.
- Receptionists own existing clients and professional callers.
- One named person reviews AI summaries daily and logs errors in fixed categories.
- A supervising attorney reviews flagged calls weekly and approves script changes.
Talk to staff about these changes before the first AI call, not after. Explain which tasks move, which stay, and what the firm wants from the time that frees up. People who helped write the routing rules and urgent triggers tend to trust the handoffs, and they are also the first to spot when a rule is wrong.
Supervision and accountability
ABA Model Rule 5.3 requires lawyers with managerial authority to have measures in place giving reasonable assurance that nonlawyer assistance is compatible with the lawyer's professional obligations[1]. California's bar guidance applies similar thinking to AI, saying diligence and supervision should match a system's access and autonomy, with autonomous client intake named as an example[2]. In practice that means a written script, an escalation list, scheduled reviews and a record of every fix.
Callers want a say too. About six in ten Americans say they would like more control over how AI is used in their own lives[7]. Offering a person on request, and honoring it immediately, is the simplest way to give them that control.
Example: a weekly coverage map
| Time | New inquiries | Existing clients | Urgent calls |
|---|---|---|---|
| Weekdays, office hours | Staff first, AI overflow after four rings | Assigned staff | Staff, then on-call attorney |
| Weekday evenings | AI | AI takes a message for the team | AI live-transfers to the on-call phone |
| Overnight | AI | AI takes a message | AI transfers; callback first thing if unanswered |
| Weekends and holidays | AI | AI takes a message | AI transfers to the weekend on-call lawyer |
Example: the same fictional firm reviews the map monthly. When it sees that Monday mornings bring a backlog of weekend callbacks, it moves one intake specialist's start time earlier rather than adding more automation.
Run a weekly quality loop
- Pull every handoff marked unresolved and every call where the caller asked for a person.
- Listen to five AI-handled calls and five staff-handled calls using the same checklist.
- Compare booking rates by channel and time slot.
- Agree one script or routing change, make it and note the date.
- Recheck that change the following week before making another.
Comparing AI-handled and staff-handled calls with the same checklist matters. It stops the review from becoming a hunt for AI errors while staff mistakes go unnoticed, and it often shows where the human script needs work too.
Measure the hybrid model, not just the AI
Score the whole intake system, because callers experience the whole system. A fast AI answer followed by a two-day wait for a staff callback is still a slow firm. Track these weekly, split by who handled the first contact:
- Share of new inquiries reached by a person or the AI on the first attempt.
- Time from AI handoff to the first staff callback, during and outside office hours.
- Booked consultations per qualified inquiry.
- Handoffs where the caller had to repeat information, found by listening to callbacks.
- Requests for a person and how long each took to honor.
Where hybrid intake usually breaks
- The summary goes to one person's inbox, and that person is on vacation.
- Urgent transfers ring a phone nobody carries after 6 p.m.
- Staff rewrite the AI script informally on the phone, so callers hear two different processes.
- Nobody owns the error log, so the same routing mistake repeats for weeks.
- Existing clients get pulled into new-client questions because the number lookup failed.
Each of these is a process failure rather than a technology failure, which is good news: a named owner and a short checklist fix most of them within a week.
Related guides
For the human side of the handoff, see intake handoff from staff to attorney. Before widening AI coverage, check the hard limits in what AI phone tools should never do and the consent rules in voice AI and call recording consent. If you are just starting, the AI phone answering pilot plan shows how to test one model before committing.
Frequently asked questions
Should AI or staff answer first?
It depends on call type and time. Many firms keep staff first during office hours with AI overflow, and use AI first after hours and for new inquiries from advertising.
How do we stop callers repeating their story?
Use a structured summary that reaches staff before they call back or before a warm transfer connects, and train staff to confirm details rather than ask again.
Does using AI reduce our supervision duties?
No. Lawyers with managerial authority remain responsible for making sure nonlawyer assistance, including software, fits their professional obligations.
What should always go to a person?
Courts, opposing counsel, existing clients with active matters, callers who ask for a person and anyone describing an emergency.
References
American Bar Association (2026). Rule 5.3: Responsibilities Regarding Nonlawyer Assistance, ABA Model Rules of Professional Conduct (archived copy) (opens in a new tab) โ
The State Bar of California, Standing Committee on Professional Responsibility and Conduct (2026). Practical Guidance for the Use of Generative Artificial Intelligence in the Practice of Law (opens in a new tab) โ
U.S. Bureau of Labor Statistics (2025). American Time Use Survey - 2025 Results (opens in a new tab) โ
U.S. Bureau of Labor Statistics (2025). Job Flexibilities and Work Schedules (2024-25 data) (opens in a new tab) โ
U.S. Bureau of Labor Statistics (2025). Receptionists: Occupational Outlook Handbook (opens in a new tab) โ
U.S. Bureau of Labor Statistics (2026). Employer Costs for Employee Compensation - June 2026 (opens in a new tab) โ
Pew Research Center (2025). How Americans View AI and Its Impact on Human Abilities, Society (opens in a new tab) โ