AI in Legal Billing and Time Entry

Legal billing software can use machine learning or generative AI to suggest task codes, organize activity, draft narrative descriptions, flag missing fields, or summarize time records. These features may reduce clerical effort, but an entry must remain a truthful record of professional work. A calendar block, document edit, or sequence of keystrokes is evidence of activity, not conclusive evidence of the legal service delivered, who performed it, or whether that activity is chargeable under the client agreement. A human familiar with the matter should verify each proposed entry before it reaches an invoice. The American Bar Association’s Formal Opinion 512 applies existing professional obligations to generative AI use. It discusses competence, confidentiality, communication, supervision, candor, and reasonable fees. On fees, the opinion points lawyers to Model Rule 1.5: charges must be reasonable, and a lawyer may not bill a client for time that was not actually spent. The opinion also cautions that a lawyer generally may not charge clients for learning a tool the lawyer will use across matters. State rules and client engagement terms control in practice; the ABA Model Rules are influential models, not binding law everywhere. Automation can create errors in both directions. A system may omit a short task, assign activity to the wrong client, turn internal training into client work, or inflate a duration from idle computer time. A fluent narrative may add a legal task that never occurred. Firms should distinguish measured activity from inferred work, preserve edits and approvals, and test whether proposed entries comply with the relevant billing arrangement. If a tool processes matter content, assess confidentiality, retention, access, and vendor terms before use. Clients may have questions about how technology affects staffing, efficiency, or charges. Clear communication is especially important if the engagement requires approval for a tool, limits data processing, or uses a fee method tied to actual hours.