AI receptionist for law firms in Canada: what it should answer, what it must escalate, and how LSO guidance applies

What intake calls a law firm actually gets

Most law firm intake falls into six call types, and each needs a different first response. When all of them share one queue, urgent matters wait behind routine ones.

In our experience, after-hours and overflow calls are the usual gap: a receptionist covers business hours, and the rest of the week goes to voicemail. We compare the two options in AI intake vs. hiring a receptionist.

What an AI receptionist should handle

An AI receptionist should handle the administrative first contact: answering, identifying, collecting and booking. These are the tasks a well-trained non-lawyer receptionist does today, done instantly and at any hour.

• Respond to web forms, email, chat and missed calls 24/7 and confirm receipt in plain language.
• Text back every missed call within one to two minutes with a booking link.
• Identify the practice area and jurisdiction from a short set of questions.
• Collect the caller's name and the names of other parties, so the firm can run its conflict check.
• State published facts only: office hours, practice areas, consultation fee and format.
• Book a consultation into the right lawyer's calendar and send a confirmation.
• Flag urgency markers — custody, detention, a hearing date, a limitation deadline the caller mentions — for immediate human review.
• Log every interaction with source, timestamps and owner.

What it must escalate to a person

Anything that requires professional judgment goes to a person. The Law Society of Ontario puts it directly: "Generative AI should not be relied on to perform duties only a licensee can perform."

• Legal advice in any form: "Do I have a case?", "Is it too late to file?", "What will I get?"
• The conflict decision. The AI collects names; the firm runs the check and decides whether it can act.
• Accepting a matter or confirming representation. Only the firm creates a retainer.
• Urgent liberty and safety matters: detention, protection orders, imminent hearings. For emergencies the system directs the caller to 911.
• Distressed callers, where tone matters more than speed.
• Calls from the other side and any question about the content of an existing file.

LSO generative AI guidance and the Legal Aid Ontario confirmation

The Law Society of Ontario treats generative AI like assistance from a non-licensee employee: the lawyer delegates, supervises and stays responsible. Its resource "Generative AI: Your professional obligations", dated April 10, 2024, lists six duties. The right-hand column is our reading of what each duty means for intake design; the LSO text itself is broader.

Legal aid work adds a formal check. Legal Aid Ontario announced that starting January 2026, roster lawyers confirm in the annual Lawyer Self-Report that they have read and comply with three LSO resources: "Generative AI: Your professional obligations", "Practice tips for using AI" and "Licensee use of generative artificial intelligence". The same notice says lawyers must not enter any LAO business information into AI platforms and are responsible for privacy or data breaches arising from their use of these tools. Firms with roster lawyers should keep legal aid file details out of the intake system and route those callers to a person.

Privacy: PIPEDA today, Bill C-36 in Parliament

Today, an Ontario law firm's intake data falls under PIPEDA, the federal law for personal information collected in commercial activity. According to the Office of the Privacy Commissioner, Alberta, British Columbia and Quebec have their own substantially similar private-sector laws, and PIPEDA applies whenever personal information crosses provincial or national borders.

The next layer is Bill C-36, which would enact the Protecting Privacy and Consumer Data Act. It had first reading on June 15, 2026, and LEGISinfo lists it at second reading in the House of Commons. Provisions of the first-reading text that touch AI intake:

• Personal information includes information inferred about an individual; health information and biometric information that can uniquely identify a person are named as sensitive.
• Organizations must make available, in plain language, a general account of their use of any automated decision system to make predictions, recommendations or decisions that could have a legal or similarly significant effect on individuals (s. 62(2)(c) of the proposed Act).
• On request, they must explain such a prediction, recommendation or decision: the type of personal information used, its source, and the reasons or principal factors (s. 63(4)–(5)).
• Individuals can make written representations to an employee able to review the decision (s. 63(6)).

The same bill text sets a maximum administrative monetary penalty of $10 million or 3% of the organization's gross global revenue, whichever is greater (s. 114), and, for the offences listed in s. 145, fines on indictment of up to $25 million or 5% of gross global revenue, whichever is greater. The Act would come into force on a day fixed by order of the Governor in Council (s. 147). Whether automated intake triage reaches the "significant effect" threshold is a question for privacy counsel. The practical preparation is the same either way: keep a record of what the AI collects, what it decides, and where a person steps in.

Measuring the leak: missed calls and response time

The intake leak is the share of enquiries that never reach a consultation because nobody answered, answered late, or never followed up. Five numbers from your phone system and inbox are enough to size it:

1. Missed-call rate, split by business hours and after hours.
2. Median time from first contact to first human reply.
3. Share of enquiries that arrive outside office hours.
4. Enquiry-to-consultation rate.
5. Consultation-to-retainer rate.

Multiply the enquiries lost to missed calls, slow replies and after-hours gaps by your consultation-to-retainer rate and average matter value, and you have the monthly cost. The Revenue Leak Map estimator runs this calculation with rough numbers. Our piece on speed to lead explains why response time matters as much as answer rate.

Setup plan: two weeks for one intake line

The fastest path for one intake line is the two-week plan below. In our experience, a focused implementation usually takes two to five weeks from audit to live system, depending on the firm's tools, practice areas and how quickly wording is approved.

A step-by-step version of this plan for any service firm is in how to fix your intake chain in two weeks.

Why law firms work with IXIA

IXIA is an operator-led implementation team for intake and follow-up at service businesses with expensive leads.

• Human review and escalation paths are part of every AI Intake Desk build.
• The system runs on your existing tools; in most cases no new CRM is required.
• You own the playbook, the templates and the system.
• We take 2–3 new engagements per month to keep implementation quality high.
• Self-serve plans start at $99/mo (Pro $299/mo) with a 14-day free trial and no card — see pricing.

For the legal-specific setup, see AI intake systems built for law firms and our earlier analysis of what works in legal intake. Related news: Google Gemini now calls businesses for its users.