Not long ago, someone with a legal problem opened Google, typed a rough version of their question and clicked into a directory or a firm's landing page. Down the hall, an associate spent the afternoon reading discovery, drafting a memo and marking up a contract by hand.
Both of those workflows are being rewritten at once, by the same technology and firms are absorbing the shift on both sides of the house simultaneously.
That's the awkward part. Billable work is being reshaped by AI in the back office and the way a client even finds a lawyer is being reshaped by AI in the front. AI is impacting law firm marketing as much as it is law firm processes is hitting law firm marketing as hard as it's hitting law firm processes.
Most firms are staffed and budgeted to fix one problem at a time. This is two.
Why are both sides changing at the same time?
Generative tools crossed into serious professional use fast and the legal industry got no grace period on either front. Adoption inside firms is climbing: Thomson Reuters reports that the share of legal organisations incorporating generative AI nearly doubled between 2024 and 2025, with a meaningful portion of users now reaching for it multiple times a week.
The consumer path to a lawyer moved at the same time. People now ask a chatbot the first version of their question before they ever open a search tab. The pipeline is being reshaped upstream while the work being sold is being reshaped downstream. Firms that treat these as separate initiatives, run by separate teams, tend to fall behind on both.
What actually changes inside the practice?
The casework side of the change is concrete. It shows up in the tasks associates used to own and in the parts of a matter that used to eat a week.
Little of this replaces the lawyer outright. It changes what the lawyer's hour is spent on and, over time, what a firm can defensibly bill for it.
What changes about how clients find a lawyer?
The acquisition side is what most firms underestimate. When a prospective client asks an AI assistant what to do about a workplace injury, a landlord dispute, or a bad contract, the assistant answers. It may recommend categories of counsel, name firms, or produce a shortlist without ever surfacing a traditional search results page.
That shift is the subject of an Attorney at Work analysis arguing that lawyers now have to compete for visibility inside the AI answer itself, not only inside the ten blue links. The starting line of client acquisition has moved upstream and firms that only optimise for classic search are showing up after the decision has effectively been made.
Can a firm really rebuild both at once?
Yes, but not by treating them as one project. The mistake is handing both to the same partner on the same quarterly plan and hoping the vendor demos line up. The two efforts share a topic and almost nothing else: different stakeholders, different risks, different budgets, different failure modes.
A cleaner split usually looks like this.
Where do most firms get stuck?
Two places, mostly. The first is governance. Individual lawyers experiment with consumer AI tools, sometimes with client information, before the firm has written down what's allowed. By the time a policy exists, the habits are set and hard to reverse.
The second is measurement. On the casework side, firms rarely track where the saved hours actually went, so efficiency gains disappear into busier calendars instead of turning into capacity or lower cost-per-matter. On the acquisition side, firms keep reporting search rankings while the buyer's journey has moved into a channel those reports don't cover. Fix the measurement and the case for further investment writes itself.
What should a managing partner do this quarter?
The firms handling this well aren't the ones with the flashiest tool stack. They're the ones treating casework and client acquisition as two connected rebuilds happening in parallel, with clear owners, a shared policy and honest numbers. That's a harder plan to sell to a partnership than a single vendor demo. It's also the one that holds up.
