Only 8% of solo practitioners have adopted AI "widely or universally" in their practice. That is the hard number from the 2025 Clio Legal Trends Report, against 35% for larger firms. The legal press usually frames this as a gap to close, a readiness problem, a sign that solos are "behind." (Source: Clio 2025 Legal Trends for Solo and Small Law Firms Report)
That framing is wrong, and if you are a solo attorney reading the 47th "AI for lawyers" pitch this quarter, you probably already know why. The reason AI adoption is at 8% among solos is not timidity. It is that most of what is marketed as AI for solo attorneys either does not actually work yet, costs more than the time it saves, or solves a problem that is not in your top five. The 8% of solos who have adopted AI widely did it by ignoring 80% of the pitches and picking two or three tools that do one job extremely well.
This is that list. Tool by tool, price by price, with the use cases where the thing genuinely saves you hours, the use cases where it will waste your afternoon, and the categories that should stay in the "check back in 2027" pile. Written for the solo in Massachusetts who is skeptical, which is the correct default when someone tries to sell you software.
What's actually working right now for solo attorneys
Three categories of AI have crossed the line from "impressive demo" to "a solo practice will genuinely recoup its subscription fee in month one." Everything outside these three is either optional or premature.
Document drafting inside Word (the clearest win)
If you make one AI purchase this year, it should live inside Microsoft Word and draft the language you are already writing by hand. This is the single highest-leverage AI category for solo attorneys and nothing else is close.
Spellbook is the category leader. It installs as a Word add-in, reads the document you are working on, and drafts clauses, redlines, and first-draft responses in the voice of your existing templates. As of April 2026 it is used by more than 4,000 legal teams and has reviewed over 10 million contracts. Pricing is roughly $180/user/month on the mid-tier plan. (Source: Spellbook.legal, Irys legal AI pricing landscape, April 2026)
Gavel (formerly Documate) is the alternative if your workflow is closer to document assembly than redlining. It is better for estate planning, immigration, and family law practices that draft from templates with conditional logic. Pricing starts around $83/user/month on the Essentials tier and climbs to roughly $113/user/month for AI-assist features. (Source: Gavel.io pricing)
Harvey is the tool BigLaw talks about, and yes, it is good. It is also priced for BigLaw. Harvey does not publish a self-serve price, but reported enterprise deals sit well north of $300/user/month with 50-seat minimums. For a solo, ignore it. (Source: Artificial Lawyer, Harvey pricing coverage)
What you actually gain, in hours: solos using Spellbook or Gavel on contract-heavy work consistently report cutting first-draft time on a standard agreement from 45-60 minutes down to 10-15 minutes. At $375/hour, recovering 20 hours a month pays the Spellbook bill roughly 40 times over. The tool does not replace your judgment on the final draft — it replaces the first 70% of typing.
Where it breaks: do not expect Word-native AI to be good at jurisdiction-specific nuance on a first pass. Massachusetts trust language, Chapter 7 bankruptcy forms, and immigration petitions all need a human editor at the end. The win is in the speed to a reviewable draft, not in eliminating review.
AI intake triage (the boring workflow win)
The second category that works right now is intake. Not "AI receptionist that qualifies leads like a human." That category is not there yet. The narrower win is AI that takes the free-text mess from your intake form or voicemail and turns it into structured matter data that writes into your case management system.
Lawmatics AI now ships conversational intake, auto-classification of matter type, and automated follow-up sequences tuned for legal intake. Pricing starts at $199/user/month for the Lite tier and runs to $349/user/month for the Pro tier that includes the AI features. (Source: Lawmatics pricing)
Clio Duo is Clio's AI layer, included in the Clio Manage Advanced tier at around $159/user/month. Its intake triage is solid if you are already on Clio. If you are not on Clio, it is not a reason to switch. (Source: Clio pricing)
The 2025 Clio Legal Trends Report found that 74% of everyday tasks performed by lawyers could be automated with AI, and intake is at the top of that list because it is rules-based, repetitive, and high-volume. A solo estate planning attorney who handles 8 new matters a month is doing intake data entry roughly 25-40 times per matter across calendar, CRM, matter, and billing systems. Automating the first capture alone is a 60-90 minute per week recovery. (Source: Clio 2025 Legal Trends Report)
Matter status summarization (with real guardrails)
The third working category is the least glamorous: using a general-purpose chat model to summarize a matter file, a deposition, or a stack of emails into a client-ready update.
Claude (Anthropic) and ChatGPT (OpenAI) both do this well. Pricing is $20/month per user for Claude Pro or ChatGPT Plus, or roughly $30/user/month for the Team tiers that include data-handling controls you actually want for client work. (Source: Anthropic pricing, OpenAI pricing)
For $20 a month, you get a tool that can turn 40 emails and a 12-page deposition summary into a two-paragraph client update in under a minute. The caveat that matters: do not paste privileged client information into a free consumer account. Use a Team or Enterprise tier, turn off training on your data in the admin settings, and treat the tool like any other third-party vendor under your ethics obligations. The ABA's 2025 TechReport puts the share of small-firm lawyers using generative AI at roughly 30%, and the single most common mistake it flags is consumer-tier use with privileged data. (Source: ABA 2024 Legal Technology Survey Report / 2025 TechReport coverage)
Legal research AI: worth it for solos, but pick one
Legal research was supposed to be the killer app for AI in law. In 2026, for solo attorneys specifically, the picture is more nuanced than the marketing suggests.
Vincent AI (vLex / Clio) is the strongest pure research AI for practicing attorneys right now. It sits on 1 billion+ editorially enriched documents across 110 jurisdictions. Bundled with Clio Advanced + vLex it runs in the $250-$400/user/month range depending on tier. If you do research-heavy work (litigation, complex corporate, appellate), it is worth it. If your practice is 80% drafting and 5% research, it is not. (Source: Clio press release on vLex acquisition)
Lexis+ AI (LexisNexis) integrates generative AI into the Lexis platform you may already use. Add-on pricing typically runs $125-$225/user/month on top of a standard Lexis subscription, putting the all-in cost north of $400/month for most solo subscribers. The interface is familiar if you are already a Lexis user, which is the main reason to pick it. (Source: [LexisNexis pricing, reseller quotes April 2026])
Westlaw Precision with CoCounsel (Thomson Reuters) is the most mature AI research product on paper, with the deepest Westlaw headnote integration. After the February 2026 Anthropic / Claude Cowork shock, which wiped 16% off Thomson Reuters' market cap in two trading days, TR has been repricing aggressively. Solo rates in early 2026 are landing around $275-$350/user/month all-in. (Source: Artificial Lawyer coverage of TR / Anthropic shock, February 2026)
The honest answer for most solos — pick one, not three, and pick based on the platform you are already paying for. The marginal value of a second legal research AI is close to zero. The 2025 Clio data shows 54% of solo AI users already have a legal research tool in their stack. Adding a second one is the most common overspend we see.
What's hype but usable if you are disciplined
Some AI categories are oversold but still have narrow, real value if you use them carefully and do not believe the pitch deck.
General chat models (ChatGPT, Claude, Gemini)
These tools are marketed as "your AI associate." They are not. What they are is an extremely capable writing and summarization assistant that will confidently make up case citations if you let it run unsupervised.
Works:
- Summarizing matter files, deposition transcripts, discovery productions
- Turning rough notes into a client-ready email
- Drafting non-legal correspondence, engagement letter revisions, marketing copy
- Brainstorming arguments and counter-arguments before you commit to a strategy
- Translating legalese into plain English for a client memo
Does not work:
- Citing cases (hallucination rate on legal citations in general-purpose models is still measurable in 2026; a Stanford study in 2024 pegged major-model hallucination on legal queries between 58% and 82%, and while foundation models have improved, unsupervised citation is still not safe)
- Jurisdiction-specific legal conclusions without a human check
- Anything you paste into a free consumer tier that includes client-identifiable information
(Source: Stanford HAI: Hallucinating Law)
AI contract review
"Upload a contract, get a full risk review" is the pitch. The reality in 2026 is more modest: AI contract review tools (Spellbook Review, LinkSquares, Ironclad) catch obvious issues like missing clauses, unusual indemnity language, and date inconsistencies at roughly 85% accuracy on standard commercial agreements. They miss the kind of subtle, jurisdiction-specific risk you are actually paid to spot. (Source: LegalOn Technologies benchmark studies, 2025-2026)
The disciplined use: run the AI pass first to catch the dumb stuff, then do a real human review. The undisciplined use, sending the AI output to a client as "your contract review," is how malpractice claims start.
What to ignore in 2026
A short list of categories that are either marketing theater or genuinely not ready.
"Autonomous legal research agents." The 2026 pitch is an AI that independently drafts, researches, cites, and files. The technology is not there. What actually ships under this label is a chat interface with a longer context window and a search tool. You still have to verify every citation. Pay for research AI, not for "autonomous agents."
"AI that replaces your paralegal." No tool in 2026 replaces a paralegal. Tools that reduce the volume of paralegal-style work (intake, document assembly, scheduling) absolutely exist and are covered above. Anyone selling you full paralegal replacement is selling a story.
"Legal GPT" wrappers with no real technology. A meaningful share of the "AI for lawyers" tools launched in 2024 and 2025 are thin wrappers around the OpenAI or Anthropic APIs with a legal-themed prompt. They charge $99-$299/month for something you can do in ChatGPT Plus for $20. Before you pay, ask the vendor: what model is under the hood, what proprietary data is it trained or fine-tuned on, and what happens to your data. If the answers are evasive, it is a wrapper.
Single-purpose "AI" features from your existing vendors that cost extra. QuickBooks, DocuSign, and half the practice management suites have bolted on "AI" features in the last 12 months and repriced upward. Most are cosmetic. Do not upgrade for an AI feature unless you can name the specific hour-per-week it will save you.
Where to start this month: the one concrete first move
If you take one action in the next 30 days, do this — install Spellbook (or Gavel, depending on your practice area) in Word and run it on the next five documents you draft from scratch.
Not a pilot. Not a committee. A thirty-day working trial on real client work. By day 15 you will know whether it fits your practice. The cost is $180/month for one month. At a $375/hour rate, the tool pays for itself the first time it saves you 30 minutes on a draft, which it will do on day one.
The reason to start here and not with a platform overhaul is structural. The 2025 Clio data shows document drafting is where 25% of AI-adopting solos have already landed their highest-ROI use case. It is also the category with the lowest switching cost, the clearest measurable outcome, and the smallest ethical exposure when done inside a tool designed for legal work.
After 30 days, if Spellbook or Gavel is in your workflow and saving you 5+ hours a week, then you have a foundation to add the second move, intake automation, without guessing. Without that foundation, every other AI purchase is speculation.
The practical bottom line
The 8% of solo attorneys who have genuinely integrated AI did not do it by buying a platform. They did it by picking one tool that does drafting extremely well, learning its limits, and then adding one more tool a quarter later when the first one stopped being the bottleneck. That sequence, narrow and specific and measurable, is the only one that has worked repeatedly across the Massachusetts solos we have talked to this year.
Everything else in the "AI for solo attorneys" market right now is either a bigger version of the same idea priced for someone else's firm, or a category that will be ready in 2027 and is not worth the risk of being the first customer today.
If you want to figure out where the highest-leverage first move is in your specific practice:
- Take the OI Index Assessment, a 3-minute diagnostic across 5 operational dimensions, scored against benchmarks for solo practices in your vertical. It tells you whether your next dollar belongs in drafting AI, intake, research, or somewhere else entirely.
- Read the attorneys page for how we actually embed with solo practices to wire these tools into an existing workflow, not as a vendor, but as the operations layer you would build if you had the time.
- Or, if you would rather just talk it through, book a working session. We map your stack, flag the two or three tools that are worth adding, and you walk away with a one-page ops audit either way.