A small law practice usually needs three layers: a legal research tool grounded in a real database (CoCounsel or Lexis+ with Protégé), AI inside the practice-management system (Clio or Smokeball), and a general assistant such as ChatGPT for non-confidential work. Spellbook covers contract review in Word, and Lawmatics covers intake and follow-up. Each does a different job.
The mistake is shopping by brand rather than by job. A tool that drafts a lease clause beautifully will not answer a late-night enquiry, and a tool that books consultations will not check a citation. Match the tool to the job, run a trial on your own documents, and keep the lawyer's signature at the end of every chain.
How to read this list: seven tools, six jobs
The seven picks below are not ranked against each other, because they mostly do not compete. The table maps each to the job it covers and the size of practice it suits. Pricing is deliberately left out where vendors publish none: Smokeball, Spellbook, Lawmatics and CoCounsel sell through quotes or plans that change, and LexisNexis shows its small-firm prices only inside its online store, so the only honest figure is the one on your own proposal.
| Tool | Job it covers | Best fit | First thing to test |
|---|---|---|---|
| CoCounsel (Thomson Reuters) | Research and drafting grounded in Westlaw and Practical Law content | Solo to mid-size practices already on Westlaw | Does each answer link to a passage you can open? |
| Lexis+ with Protégé | Research, drafting and analysis in the LexisNexis database | Practices on Lexis, including plans for 1-3 lawyer firms | A memo on a question you already know the answer to |
| Clio Manage AI | Matter admin: deadlines, client updates, draft bills, questions about your own data | Practices that already run on Clio | Court-notice deadlines against your own calendar |
| Smokeball Archie | Drafting and file work inside Smokeball matters | Practices on Smokeball, mostly Word-centred | Whether the add-on is priced per user or per firm |
| Spellbook | Contract drafting and review in Microsoft Word | Transactional and commercial practices | Redline a contract you have already finalised |
| Lawmatics | Intake, lead scoring, follow-up and marketing automation | Practices that lose enquiries or rely on referrals | Response time on an out-of-hours test enquiry |
| ChatGPT | Plain-English explanation, structure, non-confidential drafting | Every practice, with a confidentiality rule | Which plan, and what the privacy settings say |
One caution applies to all of it. Vendor pages describe what a tool can do, not what it did on your files. Product names move too: Clio's assistant used to be called Clio Duo and is now Manage AI, and Clio has since grouped its AI across several products, so check the current names on the vendor page before you quote one to a partner. The comparison in ChatGPT or a legal AI tool: which should a small firm use? covers the first big decision in more depth.
1. CoCounsel for research and first drafts
CoCounsel is Thomson Reuters' legal assistant. Its own page describes it as research, analysis and drafting grounded in Westlaw and Practical Law content, and the company has a section pitched at solo practitioners and small firms. The reason to pick it is the grounding: when a tool answers from a curated legal database instead of the open internet, the answer is traceable to a source you can open.
Use it for the first hour of a research task: framing the issue, finding the leading authorities, getting a structured summary, and drafting an outline of an argument. A realistic example: a four-lawyer employment practice is asked whether a restrictive covenant in a signed contract is enforceable against a former sales manager. The associate asks CoCounsel for the governing test, the leading cases and the facts courts weigh. She gets an organised answer with links, opens each cited case, reads the relevant passage, and finds one authority that was later narrowed. That check took ten minutes and is the part you never skip.
How to start: Thomson Reuters' plans page shows CoCounsel Legal prices only after you enter your sector, number of lawyers and a plan length of one, two or three years, and it sends existing customers to sales for account pricing. If you already have Westlaw, ask your representative whether CoCounsel is included in your current subscription or sold as a tier above it, because packaging changes. Longer terms are pitched as cheaper, so price the one-year option before you sign for three. Request a trial on a real question you have already answered, so you can compare the output against what you know to be correct.
Watch-outs: it is a subscription on top of a research service you may already pay for, so count the total, not the add-on. And a grounded tool still produces summaries that sound more certain than the underlying cases. Read the passage, not the summary.
The Complete Excel Course with AI Integration
From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.
By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner
Buy now2. Lexis+ with Protégé, the other research route
LexisNexis sells the same kind of job under a different name: Lexis+ with Protégé, which it describes as research and drafting grounded in LexisNexis sources and your firm's own documents. Its small-firm page says practices of one to three lawyers can see prices and buy plans in the LexisNexis online store without going through a sales representative, and it offers a two-day free trial of Lexis+ with Protégé. That matters, because most legal AI marketing assumes a procurement department. One rename to know if your lawyers draft in Word: Lexis Create+ is now Protégé for Word and Protégé for Outlook.
The decision between this and CoCounsel is rarely about the AI. It is about which database your practice trusts and already knows how to search, which citator your lawyers believe, and what your existing contract looks like. If your team has used Lexis for ten years, the AI that sits on top of Lexis will be quicker to adopt than a new interface, and the reverse is true for Westlaw.
A fair test uses one question your practice has already answered, run on both tools if both will give you a trial; with a two-day window, have the questions written before the trial starts. Take a lease-assignment question, say, ask each for the governing principles and leading authority, and score them on four things: did the cited cases exist, were they current, did the summary match the passage, and was anything important missing. Mark the output against your own memo. The scorecard in AI app for lawyers: score accuracy, privacy and cost in a short trial gives you a ready-made sheet.
Watch-outs: do not buy both research suites. A small practice rarely needs two, and the second one mostly buys you a second set of habits to teach the team.
3. Clio Manage AI for the work around the work
Most of a small practice's non-billable time is not research. It is chasing deadlines, writing status updates, turning time entries into bills and finding a document. Clio's AI is aimed squarely at that. Clio's own descriptions of its AI include: dates pulled from an uploaded court notice into draft calendar entries with the source text beside each one; client updates written from recent matter activity for you to review and send; draft bills built from recorded time, with changes highlighted and routed to an approver; and the ability to ask questions about your own matters in plain language.
That list sounds dull and it is where the hours are. Say a six-lawyer litigation practice receives about thirty court notices a month. If a paralegal spends ten minutes reading each notice and keying deadlines, that is five hours. With draft deadlines pre-filled and a person checking them against the source, the checking might take three minutes a notice, so the saving is closer to 3.5 hours a month. Those figures are illustrative, not a promise, but they show the shape: the benefit is real and modest, and the control is the review step.
On data, Clio's page says client work follows the same security controls as the rest of the platform, is not used to train public AI models, and is processed under zero data retention agreements. Administrators can review 90 days of AI activity and turn AI features off for the firm. Read those statements on the current page and put a copy in your records.
The naming is the thing to watch. Clio now sells Clio Work (research, analysis and drafting), Clio Manage and Clio Grow, so "Manage AI" refers to the practice-management side. Ask the representative exactly which AI features are in your plan and which are separately priced. Whatever the tool drafts, a person should read each client update before it is sent.
4. Smokeball Archie for Word-centred practices
Smokeball is the other major practice-management system built for small firms, and its AI assistant is called Archie. On Smokeball's pricing page Archie is an add-on on every plan above the entry-level Bill plan (Boost, Grow and Prosper+). "Communicate powered by Smokeball AI" is included from Boost upwards, Intake is an add-on on Boost and Grow and included on Prosper+, and the same plans list Microsoft Word and Outlook integration.
Who it suits: a practice that lives in Word and email, wants documents filed to the matter automatically, and prefers one vendor for billing, trust accounting and AI. If you are already a Smokeball customer, Archie is the obvious first trial because the AI can see the matter file without export steps.
A walk-through for a three-lawyer estate practice: the client emails a scanned will and a list of assets. The paralegal files the email to the matter, asks the assistant to summarise the asset list into a table, and drafts a schedule of assets for the lawyer to check. The saving is the retyping. The risk is a misread figure, so the table is checked line by line against the scan before it goes anywhere.
How to start: ask which plan includes the add-on, whether it is charged per user or per firm, and whether you can try it on dummy matters first. Pricing is "get pricing" on the vendor site, so never rely on a price quoted by a reseller page.
Watch-outs: an AI that only works inside one practice-management system is a reason to buy that system only if you were going to anyway. Do not switch systems for an assistant.
5. Spellbook for contracts in Microsoft Word
Spellbook works inside Word as an add-in, which is why transactional lawyers like it: there is no new interface to learn. Its pricing page lists a suite with Review, Draft, Ask, Benchmarks and Playbooks, plus a separate multi-document agent called Associate. A Playbook is a saved set of instructions for how you want a particular type of document reviewed, which is the feature that turns a generic assistant into your house style.
Pricing is per team member and quoted, and the vendor offers a seven-day free trial for lawyers. It states that it operates under zero data retention agreements, preventing data use for training, and that it holds SOC 2 Type II compliance. As with any vendor claim, ask for the document behind it.
Here is a concrete use. A small commercial practice reviews a supplier agreement from a courier firm client. The lawyer builds a Playbook with five positions: liability capped at twelve months of fees, thirty-day payment terms, mutual termination for convenience after ninety days, data-return obligations, and no unilateral price rises. The tool flags the clauses that depart from those positions and suggests redlines. The lawyer reads every flag, rejects two that misread a defined term, and adjusts the others. Time falls from ninety minutes to perhaps forty, again illustrative, and the first review of a new contract type always takes longer because the Playbook has to be written.
The broader question of what a contract tool catches and misses is covered in AI contract review for small firms.
Watch-outs: a flag-and-redline tool is best at spotting departures from a stated position and worst at spotting a problem nobody told it to look for. Never let a clean report substitute for reading the contract.
6. Lawmatics for intake and follow-up
Intake is where small practices lose revenue quietly. An enquiry arrives at 9pm, nobody replies until ten the next morning, and the prospect has already booked a competitor. Lawmatics is a legal CRM and intake system with automation, forms, appointment booking, conflict checking, file requests and email campaigns. Its current homepage promotes a "Merlin" AI suite, including Merlin Qualify, which the company says scores incoming leads so you can sign the right clients faster. Two more Merlin products, Copilot and Engage, were labelled "coming soon", so buy for what is live today.
The decision rule: buy intake software when your problem is speed and follow-up, not when your problem is a lack of enquiries. A practice that gets five enquiries a week and answers within the hour has little to gain. One that gets thirty and answers a third of them next day has a lot.
Illustrative numbers for a five-lawyer family practice: forty enquiries a month, of which sixteen arrive outside working hours. Say the practice currently converts about 20 per cent of those. If an immediate acknowledgement and a booking link raise that to 28 per cent, that is roughly one extra consultation a month from the out-of-hours group, and the maths improves if the average matter is worth a few thousand dollars. Your numbers will differ, so run the sum with your own conversion and fee figures before you pay.
Clio Grow and Smokeball's Intake add-on do similar jobs inside their own systems. If you are already on one, test that first. The fuller design questions, including handling distressed callers and conflict checks, sit in AI client intake for law firms.
Watch-outs: an intake bot must never give legal advice, must say clearly that it does not create a client relationship, and must hand a person anything involving a deadline or distress. Write that script yourself and test it with ten awkward messages.
7. ChatGPT as the general-purpose assistant
A general assistant earns a place on the list because most of a law practice's writing is not legal analysis: a firm brochure, a letter to a landlord, a checklist for new matters, an agenda for a partners' meeting, a plain-English explanation to send a client. None of that needs a legal database. It needs a fast, tidy first draft.
The rule that makes it safe is a confidentiality line. Public chatbots are for work that contains no client identities or confidential facts, or for work you have anonymised properly. If the practice wants its team to use an assistant on real client material, move to ChatGPT Business, read OpenAI's data terms before anyone uploads a file, and confirm that workspace content is excluded from model training, which is OpenAI's default for business plans. ChatGPT Business starts at two seats, listed at $20 a seat a month on annual billing ($25 month to month) when last checked.
A useful prompt for non-confidential drafting looks like this:
You are helping a small law practice write a plain-English client explainer.
Topic: what happens after a landlord serves a notice to end a periodic tenancy.
Audience: a tenant with no legal training.
Length: 250 words, no jargon, short paragraphs.
Do not cite cases or statutes. End with three questions the tenant should ask their lawyer.
Mark anything that varies by jurisdiction with [CHECK].Illustrative output: a tidy explainer with a [CHECK] marker beside the notice-period sentence. What you fix: the marked sentence, any claim stated as universal, and the tone if it sounds like advice rather than information. Three minutes of editing replaces thirty of drafting.
Watch-outs: never paste privileged material into a consumer account, and never cite a case it gave you without opening the case. The routine in how to stop AI inventing case law is worth printing for the team.
A four-lawyer practice picks its first two tools
Consider a four-lawyer practice with two paralegals, mixed work across family, small-business and property matters, and a practice-management system already in place. The partners ask for "AI" without knowing what they want. The first step is a one-page audit of where unbilled time goes over one typical week.
| Job | Weekly hours (illustrative) | Tool that targets it |
|---|---|---|
| Chasing and updating deadlines | 4 | Practice-management AI |
| Answering client status questions | 5 | Practice-management AI |
| First-pass research | 6 | CoCounsel or Lexis+ with Protégé |
| Contract review for business clients | 5 | Spellbook |
| Replying to enquiries | 4 | Intake tool |
| Marketing, letters, internal documents | 3 | General assistant |
That is 27 hours a week of candidate work. The practice already uses a practice-management system, so the cheapest first trial is its AI, which targets nine of those hours. The second trial is the research tool the lawyers already know, aimed at the six research hours. Contract review and a standalone intake system wait for a second round. The decision rule is simple: trial the AI already inside your existing systems first, add a specialist tool only for a job that survives that trial as a real cost.
Total test plan: two weeks, two tools, five tasks each, scored on accuracy, time saved and how much checking the output needed. If a tool saves time but needs the same amount of checking as writing from scratch, it has not earned its seat. Roll the winner out over about ninety days, one team at a time, with the scorecard below as the exit test.
A scorecard you can copy for any of the seven
Use the same five questions for every vendor, so your comparison does not drift towards the best demo.
- Accuracy: on five questions you have already answered, how many answers were fully correct, partly correct or wrong? Anything citing a source goes into a separate check: did the source exist and say that?
- Checking time: minutes to verify one output. A tool is only faster if drafting plus checking beats doing it yourself.
- Data handling: written answers on training use, retention, subprocessors and who at the vendor can see prompts.
- Fit with your systems: does it work where your lawyers already work, whether Word, Outlook or your practice-management system?
- Cost per seat per month, in total: add-ons, minimum seats, implementation charges and the length of the contract.
A filled-in example for one trial: the practice scored a research tool at four fully correct answers out of five, one partly correct, eleven minutes average checking against thirty-five for manual research, written confirmation on training and retention, works in the browser only, and a quote with a twelve-month commitment. That reads as a good tool with one question mark, the annual commitment, which the partners then try to negotiate down to a trial period.
Where the money goes wrong
Three realistic mistakes show up often. The first is buying a tool for a job that does not hurt. A practice with no contract work buys Spellbook because the demo impressed the partner, and it sits unused after the trial. The second is paying twice for the same capability, for instance a standalone intake product alongside a practice-management system that already has intake. The third is skipping the review step because the output looked tidy: a draft that reads well is not a draft that is right.
There is a fourth, quieter problem: team habits. If one lawyer pastes client files into a consumer chatbot "just this once", the whole practice has a confidentiality problem. A short written rule fixes most of it. Say which tools are approved for client material, which are for anonymised or public work only, and who checks outputs before they leave the building. Start the "never" list with privileged strategy, filing deadlines and any advice to a client that no lawyer has reviewed.
Questions that decide the purchase
Before signing any of the seven, ask the vendor for answers in writing to these. Put the replies in a folder alongside the contract.
- Is our client content used to train any model, yours or a third party's?
- How long are prompts, uploaded files and outputs retained, and can we delete them?
- Which sub-processors handle our data, and where are they listed?
- Does every cited authority link to a source we can open?
- What happens to our data and our matter files if we cancel?
- What is the minimum term, and what is the price after the first year?
A tool whose vendor answers all six clearly has probably been bought by law firms before. A vendor that answers with marketing copy has told you something too. For pricing structures, the tutorial on how legal AI tools charge small firms explains what to expect. Reading those replies against your own list of jobs is part of what my AI implementation consultation covers, if you want a second opinion before signing.
Month by month for a practice with no AI yet
If you are starting from nothing, this order costs least and teaches most. Month one: set the confidentiality rule and trial the general assistant on non-confidential writing, because it costs little and teaches the team how to prompt and check. Month two: trial the AI inside your existing practice-management system on deadlines, updates and bills. Month three: trial one research tool on five real questions. Months four to six: add contract review or intake only if the audit shows a job that justifies it.
Keep a log of every trial: the task, the tool, minutes saved, minutes spent checking, and any error found. After ninety days you will have a record that tells you which tool earns its seat. That log is also what you show an insurer or a client who asks how the practice uses AI. The honest answer, with the rule and the log behind it, is the best marketing line a small practice can have on the subject.
Questions small practices ask before choosing
Do I need all seven tools?
No. Most small practices need one research tool, one practice-management system with AI in it, and a general assistant for non-confidential drafting. Contract review and intake automation are add-ons for practices that handle a lot of contracts or lose enquiries out of hours. Start with the job that costs you the most unbilled hours.
Is a general chatbot good enough for legal research?
Not as the only source. A general assistant can explain a doctrine or tidy a draft, but it is not tied to a current legal database, so every citation needs checking against the primary source. Use it for structure and plain-English explanations, and use a legal research tool when you will cite authority to a court or client.
How do I check a legal AI tool's answer?
Open every cited authority and read the passage the tool relies on. Confirm the case or statute exists, is current, and says what the summary claims. Treat the tool as a fast first draft from a junior who never tires but sometimes invents. The signing lawyer remains responsible for what is filed or sent.
What should I ask a vendor about client data?
Ask whether client content is used to train models, how long prompts and files are kept, who can see them, and whether the vendor offers a zero-retention arrangement with its model providers. Ask for the security documentation and keep a copy with your records. Anything the vendor will not put in writing, assume is not true.
Further reads
- AI Implementation Plan for a Small Law Firm: The First 90 Days — A 90-day plan for rolling out whichever tools you pick.
- AI for Small Law Firms: What to Automate First — Pick the first job to automate before buying anything.
- Which Legal Tasks Should a Small Firm Never Hand to AI? — The tasks to keep away from AI whichever tool you choose.
- How Small Law Firms Use AI to Answer Client Status Questions — Review rules for client updates drafted by AI.
- AI Voice Agent for Lawyers: Cost and Safe Handling of Intake Calls — Cost and safe handling when intake moves to voice.
- AI Paralegal Software: What a Small Firm Should Test Before Paying — What to test before paying for paralegal-style software.
- AI Billing Software for Lawyers: Which Features Earn Their Cost — Which billing features earn their cost in a small practice.
- AI for Attorney Research: A Citation-Checking Workflow — A repeatable way to turn AI research suggestions into checked authorities before a draft reaches the court.
- How to Use AI in a Law Practice: A Four-Week Starter Plan — A week-by-week rollout for a small law practice: write the policy, pilot one low-risk task, verify every citation, then settle client disclosure and review.
- Using AI in Law: Which Tasks a Small Practice Should Start With — A risk-ranked list of what a small law practice can hand to AI first, with supervision and confidentiality rules for each rung.
- Which AI Is Best for Lawyers in Australia? Choose by Task — A decision rule for choosing AI by task for an Australian law practice, comparing general assistants with the AI built into practice software.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: vendor product and pricing pages for Thomson Reuters CoCounsel, LexisNexis, Clio, Smokeball, Spellbook, Lawmatics and OpenAI, read in October and November 2026.