Your firm in two years. A thought exercise worth doing today.
Not an article about a distant future. A simple economic calculation about what's already happening - and what it means for who survives and who doesn't.
When your client already knows what AI can do
According to ABA and Thomson Reuters/Georgetown data for 2025, approximately 52% of large corporations have integrated AI into regular workflows. Large in-house legal departments report that AI capability has become a factor in selecting external firms - a trend first documented in industry surveys in 2024.
Legal technology investment grew 9.7% in the past year - the sharpest growth of any legal expenditure category. This is not a question of when the shift will arrive. It is a question of whether your firm is already answering it.
An estimated ~52% of large corporations have integrated AI into regular workflows (per ABA, 2025)
Large in-house legal departments report AI capability has become a factor in selecting external counsel (Thomson Reuters, 2025)
An estimated ~9.7% - annual growth in legal technology investment (per Georgetown Center for the Legal Profession)
This is not a future scenario. It is the conversation the market is already having.
Statistics are sourced from third-party research cited alongside each figure. ORCA has not independently verified this data.
An arithmetic exercise nobody wants to do
Take a task every litigation firm knows: preparing a statement of defense. A junior associate sits on it. Analyzes the claim. Breaks down causes of action. Researches precedents. Drafts. Revises. Sends to the partner. Partner returns comments. Another revision. On average - eight to ten hours of work.
Now imagine an AI system absorbs that mechanical layer. Not a generic draft - a statement of defense that breaks down every cause of action into its elements, generates tailored defense lines, maps evidence, and is ready for partner review. The partner still reviews. But reviews focused on judgment, not on rebuilding the draft.
Today: Junior attorney bills 8 hours × $110 = $880. Partner reviews 2 hours × $165 = $330. Client total: $1,210.
Tomorrow: ORCA absorbs the mechanical drafting layer. The junior moves to strategy and more cases; the partner reviews 1 hour × $165 = $165. Client total: $165.
The firm using ORCA: charges $330 - far below the competition - and takes four cases in the same time. Revenue: $1,320. More clients. More revenue. A price point no manual firm can match.
This is not a hypothetical. The technology already exists. The question is not if - but when your client discovers that someone else is already working this way.
The numbers aren't limited to statements of defense. Statements of claim, strategy memos, risk analysis - every task that's fundamentally about analyzing facts and drawing legal conclusions faces the same arithmetic.
The firm that restructures first captures the clients. The firm that waits loses them.
What's dictating the pace of change
A year ago, lawyers talked about AI as a theoretical threat. Today, according to Thomson Reuters and ILTA surveys, most firms worldwide already use AI tools in their work — a sharp rise since 2023. Not gradual. A jump.
In parallel, Alternative Fee Arrangements are projected to rise from 20% of firm revenue toward 70% and above in coming years. The implication: firms that bill by the hour and produce documents manually are losing twice. They're slower than competitors and squeezed on pricing.
The client already expects speed. The lawyer is the last one still asking for patience.
The pressure isn't coming from above. It's coming from the client who's used to speed in every other service in their life and doesn't understand why a legal document takes a week.
The equation is clear: firms that integrate AI will restructure pricing, handle more cases in the same time, and offer prices their manual competitors can't match. Those that don't will shrink, merge, or close. In my assessment, the process will take 3-5 years in Israel. Maybe less.
"I'll just use a general language model"
This is the most common response. And legitimate. Every lawyer who gave a general language model a case description and got back something that looks like a statement of claim thinks the problem is solved. It's not.
What a generic language model produces is text that looks legal. Not a legal document. The difference is enormous, and in court it's devastating. Here's what's missing:
Element analysis
A statement of claim requires breaking down every cause of action into its legal elements. A general language model writes a general paragraph about "breach of contract." A legal system checks: Is there a valid contract? Is breach proven? Is there causation? Is there damage? An unproven element - the cause fails.
Fact grounding
Language models fabricate. Not intentionally - structurally. They generate plausible text, not truthful text. A statement of claim containing one fact that wasn't in the input is a professional risk. A real legal system verifies: every word traceable to original input.
Filing structure
A statement of claim isn't free text. It's a document with rigid structure: caption, procedural opening, factual background, jurisdiction, causes of action, damages, relief, evidence map, signature. A general language model doesn't know what these ten sections are or how each is built for Israeli practice.
Quality control
A general language model doesn't check itself. Doesn't know if a cause is weak, if relief doesn't fit, if evidence is missing. A serious legal system runs dozens of quality checks - on legal content, not grammar.
The people who say "I'll just use a general language model for legal documents" are the same people who said "I'll just use Google Translate for legal translation." They'll try it. They'll get burned. And then they'll look for the professional tool.
A general language model is a smart pen. Not a lawyer. The difference seems small until you're standing before a judge.
What a firm that made the transition looks like
Not a firm without lawyers. The opposite - a firm where lawyers do more legal work and less assembly work. The new model doesn't replace the litigator. It frees them.
Same team. More cases. Quality control that doesn't depend on who's tired.
The partner who used to review a defense statement all day - reviews on judgment, not on rebuilding the draft. The associate doing basic research - works on strategy. The paralegal assembling evidence - manages cases.
The firm doesn't shrink. It handles more cases, in fewer hours, at the same quality. Even higher quality - because quality control is automatic and consistent, not dependent on who's assigned and who's tired.
From hourly strategy to volume strategy: Instead of charging $1,210 for one defense statement, the firm charges $330 and handles four cases in the same time. Revenue goes up. Cost per case goes down. The client is happy.
Fixed fees become profitable: When the mechanical draft cycle is absorbed and the partner's work is confined to judgment, you can price a Fixed Fee with confidence - and profit.
Consistent quality: every document runs the same deterministic checks, independent of which attorney drafts it or how busy the week is. And when a case is weak or a draft isn't ready, the system says so and refuses to mark it filing-ready - the floor tells the truth instead of flattering the output.
The transition doesn't require an internal revolution. It requires one tool that works. A tool that takes facts and returns a legal document - accurate, grounded, ready for review. Not a text generator. A decision system.
A litigator who wrote about the future - then built it
In August 2025 I published an article on AI and international arbitration. The central argument: arbitration, as a flexible consent-based forum, is the natural laboratory for integrating AI into law.
I described a 2032 scenario - an AI system conducting interviews, scanning documents, detecting contradictions, preparing analysis reports for arbitrators. I finished writing. I realized I was wrong about the timeline. Not 2032. Now.
AI and International Arbitration: Arbitration as the Natural Forum for Revolutionizing Jurisprudence
International arbitration can serve as a normative laboratory: a flexible judicial forum that experiments with new rules and practices before applying them on a larger scale.
The best person to write accounting software is not an engineer — it's a good accountant with a toolkit. Because they know the domain really well, and coding is the easy part. Knowing the domain is the hard part. I think this is obviously the future.
— Boris Cherny, Anthropic (Claude Code)
I left a salaried partnership. Built ORCA - a system that takes facts and produces legal documents ready for review, with quality control that shows no mercy. Not because the technology is exciting. Because I saw what's about to happen to firms that don't change, and knew I could build the tool that would help them.
ORCA was built by someone who wrote statements of claim. Not by someone who read about them.
Want to see what this means in practice?
ORCA is now open for early access. ₪100 per run - early access, not market price. When the product is ready, the price will reflect the value.
Try it on your next case What is ORCA →