Clients Push Big Law to Prove AI Cuts Costs

Clients Are Forcing Big Law to Show Its AI Work

Large corporate clients are asking a pointed question before they hire: How are you using generative AI to lower cost and speed up delivery? That question is moving from curiosity to a gatekeeper in RFPs-and it could pressure firms to rethink the old model without killing the billable hour.

Legal leaders want measurable cost savings. As one GC put it: "We are done waiting." The buy-sell dynamic is shifting toward proof of efficiency, not promises.

What Clients Are Asking

• Spell out your AI stack and where it fits in the workflow (research, drafting, diligence, ediscovery, contract review).
• Show hard numbers: time saved, error rates reduced, and impact on staffing.
• Explain guardrails for data privacy, confidentiality, and hallucination control.
• Outline how AI affects pricing-fixed fees, phased fees, or discounts tied to AI-enabled efficiency.

These questions are showing up in RFPs and panel reviews. Rates have climbed, and clients see AI as a way to contain spend without losing quality.

Inside the Firms: Early Moves and Tensions

Some clients now require AI assistance on tasks usually handled by junior associates. That's pushing firms to rethink training without hollowing out the pipeline.

One partner described using Harvey to draft an issues list from a markup, and the firm rolled out a license to scale that use. Others report clients asking firms to help apply AI inside the client's legal ops and business functions, not just on outside counsel work.

The pattern looks familiar. Predictive coding in ediscovery was once suspect, then became the default. We may be at a similar moment with generative tools. For context on that shift, see technology-assisted review (TAR).

Key Voices

• "More general counsels are pushing their law firms to use the technology to deliver better, faster services… They are looking for measurable cost savings." - Mark Smolik, DHL Supply Chain Americas
• "We have certainly started to see clients asking those questions." - Hilary Strong, Latham & Watkins
• "More clients are requiring some work traditionally done by junior associates be assisted by AI… Figuring out how we are going to continue to develop our talent is the biggest challenge." - Laurie Mims, Keker, Van Nest & Peters
• "Some of our clients have started requiring the use of AI in certain situations… We are in the early days of this transition." - David Cunningham, Reed Smith
• "They used to spend 80% of their time collecting data and information and 20% doing strategic analysis. They want to flip that by using AI." - Bob Couture, Harvard CLP

What Firms Should Do Now

• Define your AI policy: approved tools, data handling, confidentiality, human-in-the-loop review, and matter eligibility.
• Run pilots that map to real client pain: first-draft memos, document summaries, diligence checklists, privilege review, and deposition prep.
• Measure outcomes: cycle time, hours saved, accuracy deltas, and rework. Tie those metrics to pricing options.
• Refactor staffing models: identify AI-assisted tasks for juniors, shift human time to analysis and strategy, and set coaching checkpoints.
• Prepare RFP-ready answers: tech stack, playbooks, QA process, security posture, and sample case studies.
• Train your teams: prompt standards, citation checks, red-team testing, and escalation rules for sensitive matters.

Pricing and Value Strategy

• Offer AI-adjusted fixed or phased fees on scoped tasks (e.g., doc review, contract summary) with clear SLAs and QA.
• Use "save and reinvest" framing: reduce time on rote work, increase time on strategy and risk mitigation.
• Share a quarterly AI value report: matters touched by AI, hours saved, quality measures, and lessons learned.

Metrics That Matter

• Time to first draft and cycle time per deliverable.
• Hours by task type (pre-AI vs. post-AI) and staffing mix shift.
• Defect rates: citation errors, hallucination incidents, and QC catches.
• Client satisfaction scores tied to speed, clarity, and outcomes.

Risk and Governance

• Data security: enterprise licenses, private endpoints, no training on client data, and access controls.
• Model risk: document sources, citation requirements, and mandatory human review for external deliverables.
• Matter selection: exclude highly novel legal issues or sensitive facts; apply stricter review on regulated data.
• Audit trail: prompts, outputs, versions, and reviewer sign-offs stored in the matter file.

Talent Development Without Losing the Farm

The concern is real: if AI takes the first pass, do juniors learn the craft? Solve that with structure.

• Two-track learning: juniors produce a blind first draft on key tasks, compare to AI output, then reconcile with partner feedback.
• Rotations: alternate AI-assisted matters with manual reps to build fundamentals.
• Skill-based scorecards: writing, issue spotting, judgment, and client comms-tracked across matters.

What Client-Side Legal Ops Can Do

• Update RFP templates to capture AI policies, metrics, and pricing models.
• Set a minimum bar: security requirements, human review, and tool certifications.
• Pilot with a small matter set and quarterly business reviews focused on outcomes.
• Ask for co-development: playbooks and prompts tailored to your templates and risk profile.

Training Resources

If you are building firmwide or department training on prompt quality, review workflows, and tool selection, consider structured curricula. A practical starting point: AI courses by job function that can be adapted for legal teams.

Worth Your Time

• Leadership change: Cleary Gottlieb elected Jeff Karpf as managing partner.
• On AI: Justice Amy Coney Barrett said the tech is correctly predicting questions asked at Supreme Court oral arguments.
• Investigations: Munger, Tolles & Olson retained by the New York Attorney General's office for federal subpoenas.

The Bottom Line

Clients aren't asking if you use AI-they're asking how it lowers their bill and improves outcomes. Firms that show proof, manage risk, and protect talent development will win the next wave of panel spots and premium work.