Ethical Billing When AI Does the Work
Applying Oregon Opinion 2025-205 and ABA guidance to practical billing scenarios.
When AI reduces a two-hour research task to fifteen minutes, what do you bill? The question sounds simple but implicates professional responsibility rules, client expectations, and the fundamental economics of legal practice.
Oregon's Formal Opinion 2025-205 and ABA Formal Opinion 512 provide the clearest guidance to date on AI billing ethics. This analysis translates their principles into practical billing frameworks.
The Core Tension
Legal billing has historically tracked two dimensions: time spent and value delivered. These often correlate—complex work takes longer and delivers more value.
AI disrupts this correlation. A well-crafted prompt might produce in seconds what previously required hours of research. The work product may be equally valuable, but the time investment differs by orders of magnitude.
Three billing models compete:
- Time-based: Bill for actual time spent, including AI-assisted time
- Value-based: Bill based on the value delivered to the client
- Hybrid: Some combination reflecting both inputs and outputs
None of these models is inherently unethical. All require transparency and reasonableness.
Oregon Opinion 2025-205: The Requirements
In February 2025, the Oregon State Bar issued Formal Opinion 2025-205 addressing AI tools in legal practice. The billing guidance is direct:
Disclosure of AI costs: Lawyers must inform clients, preferably in writing, if they intend to charge for the actual costs of using AI. This obligation parallels disclosure requirements for research database costs.
No duplicative billing: If AI produces significant time savings, lawyers may not engage in billing practices that duplicate charges or falsely inflate billable hours.
Learning time not billable: Time spent learning AI technology cannot be billed absent prior agreement with the client.
Cost recovery options: AI tool costs may be treated as overhead (absorbed by the firm) or passed to clients. If passed to clients, advance disclosure and informed consent are required.
ABA Formal Opinion 512: Additional Framework
The ABA's July 2024 opinion adds specificity:
General learning vs. specific learning: Lawyers may not charge for time spent learning a technology used for client matters generally. However, if a client specifically requests use of a particular AI tool, learning time for that specific tool may be billable with disclosure.
Informed consent for costs: A lawyer may charge for a portion of an expensive, proprietary AI tool or on a per-use basis if appropriate, provided everything is fully explained to the client in advance and informed consent is obtained.
Reasonableness standard: All fees must be reasonable under Model Rule 1.5. AI does not change this standard—it applies regardless of the technology used.
Practical Billing Scenarios
Scenario 1: AI-Assisted Research
Situation: Research that previously took 4 hours now takes 45 minutes with AI assistance—15 minutes of prompting and 30 minutes of verification.
Improper approaches:
- Billing 4 hours because "that's what it would have taken"
- Billing 4 hours plus AI tool costs
- Billing 45 minutes without disclosing AI use when client would consider it material
Proper approaches:
- Billing 45 minutes of actual time at standard rates
- Billing based on value delivered (if client has agreed to value-based billing)
- Billing 45 minutes plus disclosed AI costs (if client has consented)
Scenario 2: AI Tool Subscription Costs
Situation: Your firm pays $500/month for an AI legal research tool.
Options:
- Treat as overhead (no separate charge to clients)
- Allocate across matters (requires disclosure and consent)
- Charge per-use (requires disclosure and consent)
Key principle: However you handle the cost, consistency and disclosure are required. You cannot charge some clients and not others for the same service without explanation.
Scenario 3: Contract Review Efficiency
Situation: AI reduces a contract review from 6 hours to 2 hours.
The value question: The client receives the same work product. Does the billing reflect the 2 hours of attorney time or the value of 6 hours of traditional work?
Permissible approaches:
- Bill 2 hours at standard rates (time-based)
- Bill a flat fee agreed upon in advance (value-based)
- Bill 2 hours plus AI costs with prior disclosure (hybrid)
Impermissible approaches:
- Bill 6 hours claiming you "would have spent that time" without AI
- Bill 2 hours plus undisclosed AI costs
Scenario 4: Learning New AI Tools
Situation: You spend 5 hours learning a new contract review AI.
Not billable: The learning time cannot be billed to any specific client. This is professional development, not client work.
Exception: If Client A specifically asks you to use ToolX that you have not used before, you may bill learning time for ToolX to Client A with disclosure and consent.
Practical approach: Treat AI learning as a firm investment in capabilities, not a client-billable expense.
A Practical Billing Framework
Step 1: Establish Firm Policy
Before billing questions arise in individual matters:
- Decide how AI tool costs will be handled (overhead vs. pass-through)
- Establish time tracking protocols for AI-assisted work
- Create standard disclosure language for engagement letters
- Train timekeepers on proper recording of AI-assisted time
Step 2: Disclose at Engagement
Include in engagement letters:
- Whether AI tools may be used in the matter
- How AI costs are handled (overhead or passed through)
- Whether billing reflects time spent or value delivered
- Client's right to request or prohibit AI use
Step 3: Track Time Accurately
When recording time:
- Record actual time spent, including AI interactions
- Note when AI was used (for transparency, not to justify higher bills)
- Do not inflate time based on what tasks "would have taken"
- Maintain consistency across similar tasks
Step 4: Review Before Billing
Before finalizing invoices:
- Verify time entries reflect actual work
- Confirm AI cost allocations match engagement terms
- Check for apparent efficiency gains that should reduce bills
- Ensure compliance with reasonableness standard
The Reasonableness Standard
Model Rule 1.5 requires that fees be reasonable. Relevant factors include:
- The time and labor required
- The novelty and difficulty of the questions involved
- The skill requisite to perform the legal service properly
- The fee customarily charged in the locality for similar services
- The amount involved and the results obtained
AI affects several of these factors:
Time and labor: AI reduces both. Billing must reflect this reality.
Skill: Using AI effectively requires skill. This skill has value.
Customary fees: As AI adoption increases, "customary" fees for AI-assisted work will emerge. Early adopters should not exploit information asymmetry to overcharge.
Results: AI may improve results through more comprehensive analysis. Value billing can capture this when appropriate.
Disclosure Requirements Summary
| Situation | Disclosure Required? |
|---|---|
| AI used in research | When material to client |
| AI costs passed to client | Yes, in advance |
| AI used in drafting | When material to client |
| Flat fee arrangement | AI use not separately disclosed |
| Learning new AI tool | If billing learning time (with consent) |
"Material to client" means the client would consider it relevant to their engagement. When in doubt, disclose.
Key Takeaways
- Bill for actual time spent on AI-assisted work, not what the task would have taken without AI
- Disclose AI tool costs before charging them to clients—treat disclosure similar to research database cost disclosure
- Time spent learning AI tools generally cannot be billed to clients unless a specific client requested use of a specific tool
- AI efficiency gains should benefit clients through reduced bills, not be captured as hidden profit
- Establish firm-wide policies on AI billing before questions arise in individual matters
For a practitioner-oriented analysis of Oregon's approach, see NW Sidebar's overview and the Multnomah Lawyer Ethics Focus column examining the practical implications of Opinion 2025-205.