Ethical Billing When AI Does the Work

When AI turns a four-hour task into forty-five minutes, what do you bill? The answer is simpler than the profession wants it to be.

---

Let me be direct about something the legal profession finds uncomfortable: if AI reduces your research time from four hours to forty-five minutes, you bill for forty-five minutes.

Not four hours. Not "what the task would have taken." Not some creative hybrid that captures the value of the faster result. Forty-five minutes.

I know this is unwelcome in many quarters. The hourly billing model has structured the economics of legal practice for decades, and AI efficiency gains threaten the arithmetic that underpins it. But the ethical guidance is clear, and pretending otherwise creates liability.

What the Rules Actually Say

Oregon's Formal Opinion 2025-205, issued February 2025, is direct: lawyers may not engage in billing practices that duplicate charges or falsely inflate billable hours when AI produces significant time savings.

ABA Formal Opinion 512 reinforces this: fees must be reasonable under Model Rule 1.5. Time spent learning AI technology cannot be billed to clients absent prior agreement. AI tool costs can be passed through to clients only with full disclosure and informed consent.

The principles are not complicated:

• Bill for actual time spent
• Do not inflate hours based on what the task "would have taken"
• Disclose AI-related costs before charging them
• Do not bill learning time to clients

If you follow these four principles, you will not have a billing ethics problem.

The Real Scenarios

Scenario: AI-assisted research. Research that previously took four hours now takes forty-five minutes — fifteen minutes of prompting and thirty minutes of verifying the output against primary sources.

The proper billing: forty-five minutes of attorney time at your standard rate. If your firm passes through AI tool costs, add those with prior disclosure. If you treat AI tool costs as overhead, do not charge separately.

The improper billing: four hours because "that is what the research was worth." Value-based billing is permissible, but it must be agreed upon with the client in advance, not applied retroactively to capture efficiency gains the client did not authorise.

Scenario: contract review. AI reduces a review from six hours to two hours. The client receives the same work product.

If your engagement is hourly, bill two hours. If your engagement is flat-fee, bill the flat fee — the efficiency gain is your operational advantage. If you are billing value-based by prior agreement, bill accordingly. But you cannot bill six hours when you worked two.

Scenario: AI tool subscription costs. Your firm pays EUR 500 per month for an AI legal research platform.

You have three options: treat it as overhead (absorbed by the firm, not separately charged), allocate it across matters (with disclosure and consent), or charge per use (with disclosure and consent). All three are permissible. None can be hidden. And whichever you choose, apply it consistently.

Scenario: learning a new AI tool. You spend five hours learning a new AI research platform.

That time is not billable to any client. It is professional development — a firm investment, not a client cost. The single exception: if a specific client requests that you use a specific tool you have not used before, you may bill learning time for that tool to that client, with disclosure and consent.

The Value Question

Here is where the conversation gets interesting — and where I think the profession needs honest reckoning.

Some argue that AI-assisted work should be billed at a premium because the lawyer's skill in prompting and verifying adds value that the raw time does not capture. The twenty minutes you spent crafting the right prompt reflected years of legal expertise. The verification you performed required professional judgment that a non-lawyer could not provide.

This argument has merit. And there are billing models that accommodate it. Flat fees, value-based fees, and blended rates can all capture the value of expertise applied efficiently. These models are permissible under Model Rule 1.5 as long as the resulting fee is reasonable and the arrangement is transparent.

What is not permissible is retroactively inflating hourly billing to capture value that the time-based model does not reflect. If you are on an hourly engagement, you bill for hours worked. If you want to capture the value of efficiency, negotiate a different billing arrangement.

The Disclosure Framework

Oregon's opinion establishes a useful rule of thumb: disclosure of AI costs should parallel disclosure of research database costs. If your firm would disclose Westlaw or Lexis charges to clients, the same standard applies to AI tool charges.

Here is what must be disclosed:

• Whether AI tools may be used in the matter (when material to the client)
• How AI-related costs are handled (overhead versus pass-through)
• If costs are passed through, the basis for allocation
• The client's right to discuss or object to AI use

"Material to client" means the client would consider it relevant to their engagement. When in doubt, disclose. The cost of unnecessary disclosure is trivial. The cost of inadequate disclosure is substantial.

A Practical Framework

At engagement: Include AI disclosure language in your engagement letter. State whether AI tools may be used, how costs are handled, and whether billing reflects time or value. Get this right at the beginning and you avoid arguments later.

During the matter: Track time accurately. Record actual time spent, including AI interactions. Note when AI was used — not to justify higher bills, but for transparency. Do not inflate time entries based on pre-AI benchmarks.

At billing: Review entries before finalising invoices. Verify that AI-assisted entries reflect actual work. Confirm cost allocations match engagement terms. Check for apparent efficiency gains that should reduce bills. Apply the reasonableness standard honestly.

The Bigger Picture

The ethical billing question is really an economic transformation question. AI is compressing the time required for many legal tasks. The hourly billing model, which has been the dominant pricing mechanism for decades, is poorly suited to a world where expertise can be applied in a fraction of the previous time.

The firms that navigate this well will be those that evolve their pricing models — moving toward value-based, flat-fee, and subscription arrangements that capture the value of expertise without depending on time as a proxy.

The firms that navigate this badly will be those that try to preserve hourly billing economics by inflating time or hiding AI efficiency gains from clients.

Clients are not stupid. They will discover the discrepancy. And when they do, the trust damage will far exceed whatever billing premium was gained.

Bill honestly. Disclose fully. And start thinking about pricing models that work in a world where AI makes expertise faster to deliver.

The alternative is not sustainable — professionally, ethically, or commercially.