Legal Tech and AI

Generative AI Use Cases for Small Law Firms

Practical generative AI use cases for small law firms, sorted by risk, with the verification each needs, confidentiality limits and billing rules.

AssistCall research teamPublished Reviewed by Faruque Amin, VP of Sales and Marketing6 min read

On this page
  1. Start with tasks, not tools
  2. Use cases sorted by risk
  3. Low-risk uses to start with
  4. Medium-risk uses that need a defined check
  5. High-risk uses that need strict controls
  6. Verification should match the task
  7. Confidentiality comes before convenience
  8. Billing for AI-assisted work
  9. How law firm professionals feel about it
  10. Example: a first-draft demand letter workflow
  11. Uses to avoid for now
  12. Who should own AI use in a small firm
  13. Getting started
  14. Frequently asked questions
  15. References
Legal Tech and AI: Generative AI Use Cases for Small Law FirmsArticle coverLEGAL TECH AND AIGenerative AI Use Casesfor Small Law FirmsAva for LawyersInsightsLegal Tech and AI: Generative AI Use Cases for Small Law FirmsArticle coverLEGAL TECH AND AIGenerative AI UseCases for Small LawFirmsAva for Lawyers

Key takeaways

  • Sort AI uses by risk: internal drafting from your own material is low risk; research citations and client advice are high risk.
  • Verification should match the task; ABA Formal Opinion 512 allows lighter checks only after a tool has been tested on that task[1].
  • Even legal research tools produced incorrect information more than 17% of the time in Stanford benchmarks[2].
  • Bill for time actually spent; AI efficiency should not be billed as if the work took longer[3].

Start with tasks, not tools

The most common mistake with generative AI in a small firm is starting with a tool and looking for something to do with it. Start instead with the tasks that take staff and lawyer time every week, then ask which of them involve turning material you already have into a first draft that a person will check. Those are the uses where generative AI tends to help and where errors are easy to catch.

Practitioners are already moving. Among legal professionals who use generative AI, the top uses are document review (74%), legal research (73%) and document summarization (72%)[4]. The list below goes further, sorting practical small-firm uses by risk and showing what checking each one needs.

Use cases sorted by risk

Risk tiers are a starting point; adjust for your practice and tools.
TaskRisk tierWhat verification looks likeClient data involved?
Rewrite your own client letter in plainer EnglishLowLawyer reads the final letterYes, if names are included; use an approved tool
Draft a marketing article outlineLowFact-check every claim and statisticNo
Summarize a transcript or record already in the fileMediumSpot-check against the source; check every date and numberYes
First draft of a routine letter from a template and factsMediumLawyer review of facts and toneYes
Build a chronology from documentsMediumCheck each entry against its source pageYes
Legal research and case citationsHighRead every cited authority in fullSometimes
Drafting arguments for a filingHighFull lawyer review; verify every authorityYes
Answering a client's legal question directlyNot for unsupervised useLawyer must give the adviceYes

Low-risk uses to start with

  • Plain-language rewrites. Paste a letter you drafted and ask for a version a non-lawyer can follow, then compare it with the original to make sure nothing legal changed.
  • Checklists from your own procedures. Turn a long internal procedure into a step list for staff, then have the person who owns the procedure check it.
  • Email triage drafts. Draft replies to routine scheduling or document-request emails that staff edit before sending.
  • Marketing drafts. Outline articles or newsletter items, then fact-check every claim and remove any statistic you cannot source.

Medium-risk uses that need a defined check

Summaries and chronologies are where small firms often see the largest time savings, and also where quiet errors hide. A summary can drop a qualifying phrase or swap two dates. Build a fixed check into the workflow: compare every date, amount and name in the summary with the source, and read the source in full for any passage the summary relies on for a key fact. Record who did the check.

Routine first drafts from your own templates are similar. The tool fills in facts and adjusts wording; the lawyer reviews the facts, the tone and anything the tool added that was not in the instructions.

High-risk uses that need strict controls

Legal research is the area with the most public failures. Stanford researchers found that Lexis+ AI and Ask Practical Law AI produced incorrect information more than 17% of the time on their benchmark, and Westlaw AI-Assisted Research more than 34% of the time[2]. If purpose-built legal tools err at those rates, general tools deserve even more caution. Treat AI research output as a list of leads to check, never as a source to cite.

  • Read every case, statute or rule cited, in full, from an authoritative source.
  • Confirm the authority says what the draft says it says, and check subsequent history.
  • Never file anything with a citation that a person in the firm has not personally read.

Verification should match the task

ABA Formal Opinion 512 says the appropriate amount of independent verification depends on the tool and the task. Its example: a lawyer using a tool to summarize many lengthy contracts would not necessarily have to review every document manually if the lawyer had first tested the tool on a smaller subset, compared the summaries with the documents and found them accurate[1]. That is a practical model for a small firm. Test a tool on a task with documents you already know well, record the error rate, and set your ongoing check accordingly.

Confidentiality comes before convenience

Before entering any client information, find out what the tool does with it. California's State Bar guidance says a lawyer must not input confidential client information into a generative AI tool that may present material risks to confidentiality or security without informed client consent[5]. ABA Formal Opinion 512 says informed consent is required before inputting information relating to a representation into a self-learning tool whose output could lead to disclosure, and that boilerplate engagement-letter language is not enough[1]. In practice, many small firms limit client data to business or enterprise accounts whose terms exclude training on inputs, and keep consumer accounts for tasks with no client information.

Billing for AI-assisted work

If a task that took two hours now takes 30 minutes, the hourly bill should show 30 minutes. California's guidance says hourly billing must reflect time actually spent and suggests flat fees for more predictable work[3]. For firms that bill hourly, that is an argument for moving routine, AI-assisted tasks to flat fees where the rules allow, rather than letting efficiency shrink revenue.

How law firm professionals feel about it

Attitudes are more positive than the headlines suggest. Thomson Reuters found 59% of law firm professionals believe generative AI should be applied to their work[6], half are excited or hopeful about it, and only 10% see it as a significant threat to revenue[7]. The resistance in most small firms is less about principle than about time to set it up safely.

Law firm professionals' views of generative AI, 2025Thomson Reuters Institute and Thomson Reuters, 2025 surveys
Law firm professionals' views of generative AI, 2025Bar chart. Believe it should be applied to their work: 59%; Excited or hopeful about it: 50%; See it as a significant threat to revenue: 10%.Believe it should be applied to their workBelieve it should be applied to their work: 59%59%Excited or hopeful about itExcited or hopeful about it: 50%50%See it as a significant threat to revenueSee it as a significant threat to revenue: 10%10%Law firm professionals' views of generative AI, 2025Bar chart. Believe it should be applied to their work: 59%; Excited or hopeful about it: 50%; See it as a significant threat to revenue: 10%.Believe it should be applied to theirworkBelieve it should be applied to their work: 59%59%Excited or hopeful about itExcited or hopeful about it: 50%50%See it as a significant threat torevenueSee it as a significant threat to revenue: 10%10%

Source: Thomson Reuters Institute, 2025[6]

Show the data
ItemValue
Believe it should be applied to their work59%
Excited or hopeful about it50%
See it as a significant threat to revenue10%

Example: a first-draft demand letter workflow

Example (fictional): a two-lawyer injury firm uses an approved tool to produce a first draft of demand letters. The paralegal assembles the treatment summary and bills from the file, then asks the tool to draft the narrative sections using the firm's template. The lawyer checks every medical date and amount against the records and rewrites the liability section personally. Over a month, drafting time falls by about half, but the log shows the tool twice described treatment that was not in the records. The firm adds a rule: every factual sentence in a draft must trace to a page in the file.

Uses to avoid for now

  • Letting a tool answer prospective clients' legal questions without a lawyer in the loop.
  • Filing anything with citations that a person has not personally read in full.
  • Entering restricted material, such as sealed records or information under a protective order, into any tool without written approval.
  • Using consumer accounts for client work when the terms allow training on inputs.
  • Generating client testimonials, reviews or case results for marketing.
  • Relying on a tool to decide a conflict of interest.

Most of these are not permanent bans. They are uses where the safeguards are not yet in place in a typical small firm. Revisit the list when your policy, training and tool terms are mature enough to support them, and write down what would need to be true before any of them moves into the allowed column.

Who should own AI use in a small firm

Give one lawyer responsibility for approving tools and uses, and one staff member responsibility for the practical side: accounts, settings, the usage log and the list of approved tools. In a solo practice both roles are the same person, but writing them down still helps, because it forces a decision about each new tool rather than drift. The owner should also keep a short file of the vendor terms checked and the date, so the firm can show what it reviewed if a client or insurer asks.

Getting started

  1. List five weekly tasks

    Pick the ones that involve turning your own material into a draft.

  2. Choose one and a tool

    Check the tool's terms for retention and training on inputs before any client data goes in.

  3. Test on known material

    Run the tool on a closed matter you know well and count errors.

  4. Write the check

    Decide what the reviewer checks every time and record it in a short policy.

  5. Review after 60 days

    Keep, adjust or drop based on time saved and errors caught.

Frequently asked questions

What is the safest first use of generative AI in a small firm?

Internal drafting from your own material, such as plain-language rewrites of letters you wrote, where the lawyer reads the final version.

Can I use AI for legal research?

As a source of leads only. Read every cited authority in full before relying on it; even legal research tools produced errors in benchmarks[2].

Do I need client consent to use AI?

It depends on the tool and the information. ABA Formal Opinion 512 requires informed consent before entering client information into certain self-learning tools[1]. Check your state's guidance.

How should I bill AI-assisted work?

Bill hourly work for time actually spent, or consider a flat fee for predictable tasks where your rules allow[3].

References

  1. American Bar Association, Standing Committee on Ethics and Professional Responsibility (2024). Formal Opinion 512: Generative Artificial Intelligence Tools (archived copy) (opens in a new tab) โ†‘

  2. Stanford Institute for Human-Centered Artificial Intelligence (HAI) (2024). AI on Trial: Legal Models Hallucinate in 1 out of 6 (or More) Benchmarking Queries (opens in a new tab) โ†‘

  3. The State Bar of California, Standing Committee on Professional Responsibility and Conduct (2026). Practical Guidance for the Use of Generative Artificial Intelligence in the Practice of Law (opens in a new tab) โ†‘

  4. Thomson Reuters Institute (2025). 2025 GenAI report: Executive summary for legal professionals (opens in a new tab) โ†‘

  5. The State Bar of California, Standing Committee on Professional Responsibility and Conduct (2026). Practical Guidance for the Use of Generative Artificial Intelligence in the Practice of Law (opens in a new tab) โ†‘

  6. Thomson Reuters Institute (2025). 2025 GenAI report: Executive summary for legal professionals (opens in a new tab) โ†‘

  7. Thomson Reuters (2025). Generative AI Adoption Nearly Doubles as Professional Services Reach Crossroads (opens in a new tab) โ†‘

Book a demo

See Ava handle your calls.

Book a short demo. Weโ€™ll walk through your calls and workflow, and agree what Ava should take care of.

Book a demo