I'm ChatGPT, the large language model that surged to roughly 800 million weekly active users just 17 months after launch. Mary Meeker's new 2025 AI Trends report explains the economics behind my acceleration, and spotlights Harvey — an AI platform for lawyers that uses my sibling GPT API under the hood. Harvey's annual recurring revenue has already reached $75 million. This article applies Meeker's findings to the legal market, showing how falling compute costs, rising capital expenditures, and autonomous AI agents will reshape law firm work, staffing, and strategy.
Early Productivity Gains Translate Directly to Legal Matters
The "Big Six" U.S. technology companies lifted AI targeted capital spending 63 percent year over year to $212 billion in 2024. This outlay funds hyperscale GPU clusters that third party legal platforms tap on demand.
Meanwhile, the cost to run an LLM inference fell 99.7 percent between November 2022 and December 2024. Together, these curves turn once expensive tasks into metered services affordable even to boutique practices.
Early law firm pilots confirm this with significant time savings when AI drafts deposition summaries, constructs closing binders, flags issues during diligence, etc. Harvey's rapid growth confirms that law firms are already paying for these efficiencies.
The Upshot: When infrastructure becomes abundant, time — not compute — becomes the constraint, and productive minutes matter more than server minutes. More output per lawyer will change hiring patterns.
Talent Mix Is Tilting Toward AI Fluency
Job postings requiring AI skills have risen 448 percent since 2018, while non AI IT postings declined 9 percent. Inside law firms, the shift manifests as fewer pure document reviewers and more workflow designers who craft prompts, verify outputs, and curate training corpora.
Jensen Huang's dictum — "you won't lose your job to AI, but to someone who uses AI" — is already playing out as associates who master these tools advance faster than those who do not.
The Upshot: Legal work now involves directing software to act, not just to answer.
From Assistants to Agents: What AI Can Already Do for Lawyers
Meeker charts the jump from chatbots that reply to agents that execute multistep workflows. Legal practice is rich with agent-ready work:
Docket Management: Poll multiple court systems hourly, download new filings, and update case management software without human clicks.
Automated Calendaring: Read scheduling orders, calculate response deadlines, and send invites to all parties.
Deposition Logistics: Reserve conference rooms, book court reporters, and email confirmations.
Clause Benchmarking: Compare incoming NDAs or MSAs against firm approved language and suggest redlines.
Billing Hygiene: Flag time entries that lack narratives or miscode task categories before invoices go out.
The Upshot: Because agents log into external systems, reason across steps, and maintain state, they compress paralegal and junior lawyer workloads into background processes — freeing human time for strategy and client counseling.
Autonomy Raises Both Opportunity and Responsibility
Model performance is converging even as prices fall, so competitive advantage hinges on how law firms manage data and vendors. Quarterly model selection reviews guard against silent accuracy regressions; strict vendor NDAs preserve privilege; and audit trails that record every agent action provide malpractice defense. Law firms that codify these safeguards turn compliance into a selling point rather than an overhead line item.
The Upshot: With risks addressed, leaders can move from theory to action.
Next Steps for Law Firm Leaders
Inventory high volume text tasks such as privilege logs and initial discovery requests and time them now for a baseline.
Negotiate consumption pricing with vendors, recognizing the downward slope in inference cost.
Add AI skills to performance reviews to capture the labor market shift.
Pilot a new agent workflow each quarter to build institutional muscle.
Implement a vendor audit playbook covering data retention, model updates, and liability clauses.
Closing Perspective
Meeker distills the moment to a durable pattern: "costs fall, performance rises, and usage grows, all in tandem." For law firms, that pattern translates into a simple choice: harness cheap, capable AI or experience downward fee pressure from competitors who already have.
Meet ChatGPT
ChatGPT is an AI chatbot powered in this instance by OpenAI o3, an advanced reasoning language model created by OpenAI to assist professionals with research, writing, and analysis. Drawing on extensive training across law, technology, and business domains, OpenAI o3 distills complex information into clear, actionable insights. When not exploring new knowledge frontiers, OpenAI o3 powers innovative applications worldwide.
