---
title: "A matter-level AI use log: the 8 fields that make AI work defensible"
url: https://thematteros.com/blog/ai-use-log-for-legal-matters
updated: 2026-10-27
author: Raghav Handa
---

# A matter-level AI use log: the 8 fields that make AI work defensible

A matter-level AI use log is a table in the client file that records, for each AI-assisted task, the matter, tool, task, data class, reviewer, verification method, outcome and any client-consent or court-disclosure flag. Eight fields are enough. It lets you show what AI did, who checked it and why you relied on it.

## Key takeaways

- Log at the matter level, not only firm-wide; that is where questions arise.
- Eight fields: matter, tool, task, data class, reviewer, verification, outcome, disclosure flag.
- The log records; it does not make a use compliant.
- Decide your engagement-letter position before you need it.

## Why log AI use at the matter level?

A firm-wide AI policy says what may be done. A matter-level log shows what was done on a given file, and that is what a client, insurer, court or regulator asks about. If a citation is challenged or a client asks whether AI saw their documents, you want the answer in the file, not in memory.

No rule we can verify requires this exact log, so treat it as good practice built on the general duties in ABA Formal Opinion 512 and your own bar's guidance (competence, confidentiality, supervision, communication, fees). Check local rules and any court AI orders.


## What are the eight fields?

| Field | What to record | Why |
| --- | --- | --- |
| Matter | Matter number and name | Ties the entry to the file |
| Tool | Product, tier, date of use | Terms differ by tier |
| Task | Summarise, draft, research, translate, extract | Risk differs by task |
| Data class | Public, anonymised, identified, sensitive | Links to your AI policy table |
| Reviewer | Named person who checked | Supervision evidence |
| Verification method | Authorities opened, facts matched to file | Shows reliance was reasonable |
| Outcome | Used as is, edited, discarded | Shows judgement was applied |
| Disclosure flag | Client told? Court certification needed? | Prompts a required statement |


## A worked example

Fictional entries for a supply-contract dispute, matter 2026-014:

| Date | Tool / task | Data class | Reviewer / verification | Outcome | Disclosure |
| --- | --- | --- | --- | --- | --- |
| 14 Oct | Business-tier assistant / summarise 60-page contract | Identified | B. Partner / read key clauses against original | Edited, used in note | Client informed in engagement letter |
| 15 Oct | Same / draft chronology from file notes | Identified | A. Lawyer / every date matched to source | Two dates corrected | None |
| 20 Oct | Research tool / find authorities on mitigation | Anonymised | A. Lawyer / all cases opened in database | One case discarded as inapplicable | Check court's AI order before filing |

Note the outcome column. 'Two dates corrected' is exactly the evidence that a human actually checked.


## Should you tell the client, and how?

Practice varies and bar guidance is not uniform. Commentary on ABA Opinion 512 discusses informing clients when AI use bears on confidentiality, fees or significant decisions. Bloomberg Law has reported firms struggling with how much to disclose. Vendors such as Polygraf publish example engagement-letter clauses in several variants, and Ezel publishes a client-disclosure template; treat any of these as drafting examples to adapt with your own advice, not as approved wording.

Three positions are common, from lightest to heaviest: a general sentence that the firm may use AI-assisted tools under supervision and confidentiality safeguards; matter-specific consent when identifiable data goes into a third-party tool; and no AI on this client's matter. Decide which applies by client and by data class, and record it in the Disclosure field.


## What about billing and court certifications?

Commentary notes a tension between billing time and AI-assisted speed: charging hourly for work a tool did in seconds raises fee-reasonableness questions, which is one reason flat fees are discussed alongside AI. Take advice on your jurisdiction's fee rules.

Some courts have added orders requiring certification or disclosure of AI use in filings; reports differ and some are unverified, so check each court's current orders before filing. The log's disclosure flag is the reminder.

- Flat fees in the AI era: https://thematteros.com/resources/flat-fees-in-the-ai-era
- How to use AI on a matter: a 5-step workflow: https://thematteros.com/blog/use-ai-on-a-legal-matter
- Law firm AI policy template: https://thematteros.com/blog/law-firm-ai-policy-template


## Where MatterOS fits

Last reviewed: 30 October 2026. We make MatterOS, a Notion template for lawyers, so weigh our view accordingly. This page is organisational information, not legal advice.

The log is a small table linked to the matter, which is the kind of structure MatterOS ($39 one-time) gives you in Notion: add it as a database related to Matters. MatterOS does not log AI use automatically, and Notion's page history is not a compliance audit trail. Our audit trail page explains the difference. The AI prompt pack helps with prompts, not with compliance.

- Audit log vs page history: https://thematteros.com/features/audit_log
- Data-hygiene prompts: https://thematteros.com/features/agent_hygiene
- MatterOS: the Notion operating system for lawyers: https://thematteros.com/product
- What is included and what it does not do: https://thematteros.com/facts
- Pricing: one payment, no subscription: https://thematteros.com/pricing


## FAQ

### What is an AI use log for a law firm?

It is a record, kept with the client file, of each AI-assisted task: tool, task, data class, reviewer, verification method, outcome and disclosure flags. It shows what AI did and who checked it. It supports a firm's AI policy but does not by itself make use compliant.

### Do lawyers have to disclose AI use to clients?

There is no single rule. Bar guidance generally focuses on whether AI use affects confidentiality, fees or important decisions, and many firms add a general engagement-letter sentence. Check your jurisdiction and treat sample clauses as drafting examples, not approved wording.

### How do I document AI use in a client file?

Add a table or database to the matter with one row per AI-assisted task. Record the tool, the task, the data class, the person who checked the output, how they checked it and whether it was used. Keep it with the file so it is retained and exported with it.

### Can I bill clients for AI-assisted work?

Fee rules still apply: fees must be reasonable and communicated. Time saved by AI raises questions for hourly billing, and flat or fixed fees are one response. Commentary on ABA Opinion 512 addresses this; confirm what your bar says.

### Do courts require disclosure of AI use?

Some do. Individual judges and courts have issued standing orders or certification requirements, and reports of specific orders vary in reliability. Always check the current rules and orders of the court you are filing in before relying on any general statement.

## Sources

- [Polygraf: engagement letter AI disclosure clause templates](https://polygraf.ai/ai-compliance-library/engagement-letter-ai-disclosure-clause-templates-for-law-firms-polygraf-ai/)
- [Ezel: attorney AI tools client disclosure](https://ezel.ai/templates/attorney-ai-tools-client-disclosure)
- [Bloomberg Law: firms wrestle with telling clients about AI](https://news.bloomberglaw.com/us-law-week/law-firms-wrestle-with-how-much-to-tell-clients-about-ai-use)
- [ABA Business Law Today on Formal Opinion 512](https://www.americanbar.org/groups/business_law/resources/business-law-today/2024-october/aba-ethics-opinion-generative-ai-offers-useful-framework/)