Disclosure: DROZlegal publishes this guide and builds a practice-automation product for Canadian law firms. The adoption and court-guidance statistics below are sourced directly from 8am's and Canadian Lawyer magazine's own 2026 published research, not ours.
How lawyers are actually using AI right now
Ask a lawyer whether they "use AI" in 2026 and the honest answer is almost always yes — but for most, that means a chat window, not a workflow. The clearest picture of that gap comes from 8am's 2026 U.S. Legal Industry Report, published March 5, 2026 from a survey of more than 1,300 legal professionals fielded across the United States in September and October 2025.
The numbers below are U.S. data — 8am's survey did not include Canadian respondents — but the shape of the gap is the one Canadian firms are living through too:
- 69% of U.S. legal professionals personally used generative AI for work in 2026, more than double 2024's 27% and up from 31% in 2025.
- 42% used legal-specific AI tools, double the 21% recorded a year earlier.
- Firm-wide adoption of legal-specific AI tools sits at just 34% — individuals are moving far faster than the organizations around them.
- 54% of firms report giving no training on responsible generative-AI use, with no plans to start.
- Only 9% of firms have a written AI-use policy that's actively enforced.
Source: 8am, "2026 Legal Industry Report" (survey fielded September–October 2025 among U.S. legal professionals; published March 5, 2026).
Read that list together and the pattern is: individual use is nearly universal, firm-level adoption lags a long way behind it, and governance lags behind that. A lawyer opening a chat tab is not the same thing as a firm running AI as part of how work actually gets done — and the gap between the two is where most of the risk sits.
Canadian courts are writing AI rules faster than most firms are catching up
Canada doesn't have a directly comparable national adoption survey with 2026 field dates, but there's a sharper, more practical Canadian data point: the courts themselves are moving.
Canadian Lawyer magazine contacted all 51 Canadian courts about their generative-AI policies for a survey published June 10, 2026. 21 responded directly, and counting those responses together with publicly posted practice directions, the magazine counted 21 courts that now have some documented genAI guidance for lawyers and litigants.
Eleven of those courts frame their approach around the Canadian Judicial Council's guidelines; only three have separate rules for law clerks using generative AI. The magazine named the BC Court of Appeal, the Quebec Court of Appeal, the Superior Court of Quebec, and the Yukon Court of Appeal as having the most developed policies. Source: Canadian Lawyer, "Canadian Lawyer survey: How Canada's courts are regulating, using, and evaluating generative AI," published June 10, 2026.
That's a second layer of obligation on top of the Law Society's. The Law Society of Ontario's existing guidance on generative AI — competence, confidentiality, supervision — applies to every matter regardless of which court it lands in; see our breakdown of LSO's generative AI guidance for what that requires in practice. But if the matter is heading to court, a firm now also needs to check that specific court's rules on AI-assisted filings before it drafts anything with a tool.
A chat window and a workflow are two different tools
Most day-to-day AI use at a law firm today looks the same regardless of firm size: type a prompt, get a draft, copy it into a document or email, and start over for the next matter. Nothing persists between prompts, nothing tracks what happened, and the task isn't finished when the chat ends — a person still carries the output the rest of the way, every time.
Practice automation is a different shape of tool. Instead of a conversation you drive step by step, it's a named job that runs the actual multi-step task on a queue and stops at a real decision point for a human to sign off — not because it ran out of things to say.
DROZlegal's bulk_intake agent is a working example: it ingests a batch of inbound matters — from a portal, an email folder, or a CSV — triages them, runs conflict checks against the firm's CRM, and opens each matter with an engagement letter already drafted, ready for the lawyer's sign-off. The letter is drafted, never sent on its own: engagement approval is one of six actions DROZlegal's agents treat as a permanent, human-only gate, alongside trust money movement, court filing, settlement, commencing litigation, and agent-initiated email send. For the intake side specifically, see our AI legal intake software guide; for the revenue data behind why chat-only adoption doesn't show up on the bottom line at small firms, see AI practice automation for small Canadian firms.
If your firm's AI use today is mostly a chat tab open next to your matter files, that's the exact gap the DROZlegal waitlist conversation is for — seeing what the same task looks like as a job that finishes, with an approval step you actually control.
Five questions before your firm adopts an AI tool
Whatever tool a firm is evaluating — DROZlegal or otherwise — these are the checks that separate a defensible AI rollout from a liability waiting to surface:
| Question to ask | Why it matters |
|---|---|
| Where does the data live, and does the vendor train on it? | Ask for the AI-processing terms in writing. "No training, deleted within roughly 30 days" is a specific, checkable claim; "zero retention" usually isn't a real arrangement unless separately negotiated. |
| Does the court I file in have its own AI rules? | 21 of 51 Canadian courts already publish genAI guidance for lawyers and litigants (Canadian Lawyer, June 2026) — on top of, not instead of, Law Society obligations. |
| Does it finish the task, or hand me the next draft? | A chat answer you still copy into a letter isn't automation — it's a faster typewriter. |
| Is there an audit trail I could show a regulator? | If the only record is a chat history in someone's browser, that's a supervision and confidentiality gap waiting to surface. |
| What can it never do without me? | Trust money movement, court filing, settlement, commencing litigation, engagement approval, and agent-initiated email send should always stop for a human, not just "usually." |
DROZlegal's free Lawyer AI Academy walks through the competence and confidentiality obligations behind the first and fourth questions in plain language, grounded in LSO guidance rather than vendor marketing. For a fuller vendor-safety checklist beyond these five questions, see is AI safe for law firms?
Frequently asked questions
How many lawyers are actually using AI in 2026? A lot, but mostly for chat-style tasks rather than firm-wide workflows. In the United States, individual generative-AI use among legal professionals reached 69% in 2026 — more than double 2024's 27% — while firm-wide adoption of legal-specific AI tools sits at just 34%, and only 9% of firms have a written, actively enforced AI policy (8am, 2026 U.S. Legal Industry Report). Canada-specific adoption surveys aren't identical, but the same pattern shows up in how fast Canadian courts are having to write their own AI rules to keep up with individual use.
Do Canadian courts have their own rules for AI-assisted legal work? Yes, and that list is growing. Canadian Lawyer magazine contacted all 51 Canadian courts about their generative-AI policies for a survey published June 10, 2026; 21 responded directly, and counting those responses together with publicly posted practice directions, the magazine counted 21 courts with some documented AI guidance for lawyers and litigants — including the BC Court of Appeal, the Quebec Court of Appeal, the Superior Court of Quebec, and the Yukon Court of Appeal, cited as having the most developed policies. A firm needs to check the specific court's rules, not just the Law Society's guidance.
What's the real difference between using AI and automating a workflow? Using AI usually means someone typing a prompt into a chat tool and copying the answer into an email or document, one matter at a time, with no record left behind. Automating a workflow means a named job — like DROZlegal's bulk_intake agent, which triages inbound matters, runs conflict checks, and drafts an engagement letter — runs the multi-step task on its own and stops at a defined point, such as engagement approval, for a lawyer to review and approve before anything is finalized.
Not ready to subscribe? Join the DROZlegal waitlist instead.