Insights
Updated September 2026

AI and the Billable Hour: Evaluating Legal Pricing

How to evaluate pricing and complete review effort when introducing AI assistance into legal work.

6 min read

General information, not legal advice. Legal position as of . Limitations in the Legal Notice

Review status: legal and language review by a named human reviewer is pending.

In this article

Introducing AI assistance is a reason to examine legal pricing, not evidence that the billable hour is disappearing. The useful question is how the complete work, risk and client agreement change. Count verification and correction before translating an apparent speed gain into a fee promise.

What productivity evidence does and does not show

Nielsen and colleagues' 2024 experiment with law students found that AI-generated highlighting improved efficiency in the studied legal task, while summaries alone did not improve performance. That difference matters more for pricing than a generic claim that “AI saves time.” The study does not establish a saving for a particular firm's matters, insurance position or billing model.

Noy and Zhang's 2023 writing experiment also found productivity benefits in its professional-writing tasks. Those were bounded experiments, not a measurement of complete legal engagements. Use research to identify questions for a pilot, then measure the work your firm actually delivers.

Pricing models to compare

Compare fee options against the same scope

Hourly billing: define which actual work is charged, how technology-related costs are treated and what records support the invoice. Do not describe time the team did not work as lawyer hours worked.

Phased or fixed fees: scope the assignment, dependencies, exclusions and changes before quoting. AI might reduce some work while adding uncertainty or review. A fixed fee transfers some estimation risk to the firm; it does not make that risk disappear.

Subscriptions: define included services, volume assumptions, response expectations and exceptions. Recurring revenue is not evidence that unlimited work can be supported safely.

Value-related pricing: discuss the client's objectives and the services needed to achieve them. For lawyers who fall under the Swiss Lawyers Act (BGFA), Article 12 letter e prohibits agreeing, before a legal dispute ends, to a share of the proceeds in place of a fee, and prohibits undertaking to waive the fee if the proceedings end unfavourably. In BGE 143 III 600 (in German), the Federal Supreme Court held that a pactum de palmario, a success bonus in addition to the fee, is permissible within limits. The fee owed regardless of outcome must cover the lawyer's costs and allow a reasonable profit. Relative to the fee owed in any event, the success-dependent part must not be so high that it impairs the lawyer's independence and creates a risk of unfair advantage being taken of the client; it is too high in any case if it exceeds the outcome-independent fee. The agreement may be made at the start of the mandate or after the dispute has ended, but not while the mandate is running. Check these limits and the applicable independence requirements before choosing any outcome-related arrangement. The diagram presents options to compare, not a mandatory progression or a recommendation for every matter.

Build a matter-level cost record

Record scoping, document preparation, research, drafting, expert review, corrections, communication and delivery. Add relevant licence, integration, training and support costs. Compare matters with similar scope and complexity and report uncertainty where the sample is small.

For a client-supplied AI draft, identify the task as review, correction or replacement. A plausible draft may still need substantial source checking or reconstruction. Agree the work and the pricing basis before proceeding; neither the client's prompt time nor the length of the draft establishes the lawyer's required effort.

Test a phased engagement

An illustrative pilot could begin with a paid scoping stage, followed by a defined execution proposal and an agreed process for scope changes. Select a suitable client and matter, record the assumptions, and compare forecast with actual effort. Decide whether the arrangement works from quality, client understanding and economics together.

Expect the effect on variance and margins to differ from task to task. Track exceptions, missed issues and rework. A commercially attractive model still needs enough review capacity to deliver the agreed service.

Explain the actual use and price

Describe AI's role and the human checks accurately. The Swiss Bar Association (SAV) AI guidance (in German, French and Italian) identifies possible disclosure duties arising from supplier terms or expectations of personal mandate performance. It does not create a universal rule that every AI use is authorised by adding one sentence to an engagement letter.

Separate the pricing agreement from consents to disclose confidential information and other professional duties. Give clients a clear scope, fee basis and route for questions. A claim of savings should identify the comparison and include the work needed to verify the output.

Pricing pilot record

0/6

Key Takeaway

Price the agreed service using evidence about the complete work. AI speed alone does not determine the right fee model or prove a saving for the client.

Get in touch to discuss a pricing pilot for a defined practice area.

Align direction and organisational readiness

Align leaders, understand readiness and choose an owned 90-day roadmap.

You might also like

Need clearer footing for an AI decision?

Start with a focused conversation about a live AI use case, workflow bottleneck, training need, or governance gap.