"If my law firm uses AI to review our contracts in a fraction of the time, why is my bill the same?" More business owners are asking their outside counsel that question, and it's the right one. AI is now doing work that used to eat weeks of associate time: sorting discovery, pulling key clauses out of thousands of pages, drafting first-pass agreements, and flagging risk in vendor paper. The tools are real, they're in production at large firms, and they're changing what a fair invoice looks like.

For a company that spends five or six figures a year on legal work, that shift matters. Knowing where the work got faster, and what to ask for, is the difference between paying old rates for faster output and actually sharing in the savings.

What Is AI Actually Doing Inside a Law Firm Right Now?

The honest answer: the boring, expensive parts. Firms are pointing machine learning and large language models at the tasks that used to keep junior lawyers at their desks until midnight. A few specific areas have moved fastest:

• Document review and discovery. Software sorts massive document sets by relevance, privilege, and topic, then surfaces the handful of items a human should read.
• Contract review and redlining. AI compares an incoming agreement against a firm's playbook, flags off-market terms, suggests fallback language, and drafts a first-pass markup in minutes rather than hours. A buyer's guide from Thomson Reuters walks through what to look for in this category of tool.
• Drafting and summarization. Models produce first drafts of NDAs, engagement letters, memos, and deposition summaries. A lawyer still edits and signs, but the blank page problem is largely gone.
• Legal research. Purpose-built legal tools pull case law and statutes with citations attached, so an associate spends the time analyzing rather than hunting.

None of this replaces judgment. It replaces typing, reading, and searching, which is a meaningful chunk of what firms have historically billed you for.

How Much Faster Are We Really Talking About?

Fast enough that the old economics don't cleanly hold. For pattern-heavy work like vendor contracts, NDAs, employment agreements, leases, and standard M&A due diligence, the time savings can be an order of magnitude. Novel litigation and nuanced negotiation strategy do not compress the same way. Push hard on efficiency for the first bucket, and expect to pay for judgment in the second.

Are Clients Actually Seeing the Savings?

Mostly, no. Not yet, and not automatically. The tools are being adopted; the invoices aren't reflecting it. In some markets, top-tier firm rates have kept climbing.

That mismatch is forcing the negotiation. Companies are writing AI expectations directly into outside counsel guidelines: use it on lower-risk tasks, disclose where it was used, and pass through the efficiency. Firms that resist are watching commodity work drift to smaller shops and to in-house teams that have brought review software in-house.

What Should You Put in Your Next Engagement Letter?

The engagement letter is where all of this becomes real. A few specific asks separate a serious client from a hopeful one:

• Disclosure. Ask the firm to tell you which tools they use, on what categories of work, and what human review sits behind the output.
• No billing for tool overhead or training time. Firms should not bill clients for time spent learning generative AI tools, and general subscription fees are overhead. Put that in writing.
• Fixed fees on repeatable work. Price NDAs, vendor contracts, and standard employment paper as flat or capped fees, not hourly. If the firm is faster, you benefit; if they're slower, that's their problem to solve.
• Confidentiality controls. Confirm your data is not being used to train a public model and that the firm's AI setup keeps privileged material inside a controlled environment.
• A quality standard. Faster isn't the point if the work is worse. Ask how the firm audits AI-assisted output and who signs off.

What Do You Need to Understand Before You Walk In?

You don't need to become a technologist, but you do need enough vocabulary to tell marketing from substance. There's a wide distance between a firm that plugged a chatbot into its intranet and one that has deployed private legal models, retrieval systems tied to its own precedent, and human approval gates on every output. For a useful primer before your next rate conversation, a a Law.co analysis of legal AI analysis of legal AI walks through how these systems actually get built for law firms and where the real capability sits.

Two questions cut through most of the fog. First: where does our data live when your team runs it through a tool? Second: who is checking the output before it comes to us? A firm that can answer both cleanly has thought about this. A firm that can't is still figuring out its own posture, and you should not be paying premium rates to underwrite that learning curve.

Bring the topic up. Ask the questions. Rewrite the engagement letter.

The best firms are already having these conversations with their sharpest clients and adjusting how they price, staff, and deliver. The rest will follow, or they'll lose the work to someone who did.