No. iCapability
78–81%
vs. 71% human baseline
AI legal research now beats the human-lawyer baseline. The capability question is settled. The supervision question is not.
↳src. Vals Legal AI Research benchmark, October 2025.
For practicing legal professionals
What Am I for?
Artificial Intelligence (AI) now does in minutes what took your associate hours. So what is the attorney for? The answer is a workshop, not another demo.
It’s 11:43 PM. There are 47 tabs open.Not one of them is a client matter. The blocker isn’t the tools — it’s the question underneath them.
Field stamp · 26 · V · 26 | reading time ~4 min | revision r.0014
The blocker beneath the blockerFour months of research.
Four months of research.
Zero implementation.
You did the work. You read the McKinsey deck, demoed Harvey, scrolled three Reddit threads, started a Notion doc, asked a partner what they thought. Nothing moved.
The real blocker isn’t the tool stack. It is the identity question underneath it — if Artificial Intelligence (AI) can do what an associate does in an afternoon, what am I for?
AI replaces
- Tasks
- Drafts
- Summaries
- First-pass review
AI never replaces
- Judgment
- Accountability
- Candor
- The client relationship
That line is the entire workshop. Everything else — the tools, the pricing, the verification gate — follows from it.
The research, audited
What the research actually says.
Verified figures only — sourced, not synthesised. Five numbers that argue the conversation is overdue, and that the answer is a workflow, not a product purchase.
No. iiRisk
≈ 43%
Hallucination rate of paid legal tools
The model isn’t the risk. Skipping verification is. Lexis+ AI ≈ 17%, Westlaw AI-Assisted Research ≈ 33%, GPT-4 ≈ 43%.
↳src. Stanford, independent testing, 2024–25.
No. iiiConsequence
1,459
Hallucination cases in courts worldwide
A growing database of sanctioned filings and on-the-record consequences. It keeps growing because the gate keeps being skipped.
↳src. Damien Charlotin database, early 2026.
No. ivAdoption gap
79%
/43%
/43%
Lawyers using AI vs. firms with no policy
Adoption has outrun governance. The associates are already using it. The firm hasn't said what's allowed.
↳src. Clio 2025 Legal Trends; 8am 2026 Legal Industry Report.
No. vOutcome
3.9×
Firms with formal AI strategy
… are nearly four times more likely to see critical benefits. A strategy is not a policy memo. A strategy is a codified workflow with a verification gate.
↳src. Thomson Reuters Institute, 2025.
The Showcase · Free Workshop“What Am I For?” —
“What Am I For?” —
Three days. One workflow built live.
Not a webinar. Not a demo reel. Three working sessions that resolve the identity question, codify your judgment, and end with a Claude workflow you actually own. Built for practicing legal professionals only.
FreeVirtual3 daysPracticing legal professionals only
01Day 01
The Internal Operating System
Resolve the question before you buy the tool.
Tasks vs. judgment, fully drawn. We break the four-month research loop in one session by separating what AI can do from what only you can sign your name to.
- Identity audit — what only you are for.
- The tasks-vs-judgment partition, drawn for your practice area.
- Why “more research” is the resistance, and how to recognise it.
02Day 02
Influence + The Amplifier Stack
The human skill AI multiplies but can never replace.
Codify your real workflow into a playbook a model can run. Build the verification protocol — mapped to American Bar Association (ABA) Formal Opinion 512 — that turns AI output into work you can defend.
- Codify a real intake, demand, or motion workflow.
- Wire it into Claude with the cite-check gate in line.
- Influence as the multiplier the model can never own.
“Influence is the multiplier the model can never own.”— Day 02, working note
03Day 03
The Stack
The honest 2026 tool landscape, then we build live.
The real prices, the real trade-offs, and a live build on Anthropic's Claude — including the cite-check gate that keeps your work out of the hallucination case database.
- Enterprise tools vs. the small-firm-accessible stack.
- Live build: a Claude workflow with cite verification.
- What to ship Monday morning.
Bring one real workflow you keep meaning to fix. Leave with it codified and running on Claude, with the verification gate in line.
Reserve Your SeatDBAI · The Amplifier Stack
Three movements. The order is non-negotiable.
Skip Codify, jump to Automate— buy a tool and bolt it onto an undefined process — and you get faster chaos. That is why four months of research moved nothing.
Codify
Extract your real judgment into an explicit, written playbook. Tone, edge cases, exclusions, decision criteria, what counts as “done.” The model can’t know what you’ve never said out loud.
Automate
Wire the playbook into Claude-powered workflows that run the repeatable 80% — with verification baked in, not bolted on. The model becomes the associate who actually reads your style guide.
Amplify
You spend your hours only where mastery is required — judgment calls, candor with the client, the one paragraph that wins the motion. Everything else runs.
“You don’t need more tools. You need your judgment, written down once,running every day.”
The two-tier reality
Stop researching Harvey.
The legal-AI market is split. One tier is built for AmLaw 100. The other tier is available to you today, at a fraction of the seat price, running on the same underlying model.
Tier I · Enterprise / BigLaw
The Harvey tier
Built for AmLaw 100 procurement, with the seat minimums and floor pricing to match. Mostly inaccessible to firms under 50 attorneys.
- Harvey20–25 seat min · ~$288K/yr floor$1k–$2k /seat/mo
- LuminanceEnterprise procurementEnterprise
- IroncladEnterprise procurementEnterprise
Tier II · Small-firm accessible
The tier most firms overlook
Real tools, real pricing — available to a two-attorney shop on a credit card today.
- Claude Team$25 /seat
- Microsoft 365 Copilot$30 /seat
- Spellbook$99–$199
- Paxton AI$159–$500
- vLex Vincentfrom $69
- Clearbrief$300
- Clio DuoElite tier$159
Most legal Artificial Intelligence (AI) runs on Anthropic’s Claude. Harvey runs Claude Opus 4.7 — 90.9% on its own BigLaw Bench. The same engine inside a $1,200/seat tool is available direct at $25. The differentiator isn’t the model. It’s your data, your integrations, and your codified workflow. Which is exactly what the workshop builds.
Safe by design
We build thegate.
Every workflow we ship carries a human-in-the-loop cite-check gate that fails closed — if a citation can’t be verified, the work doesn’t leave the system. The gate is mapped, line by line, to American Bar Association (ABA) Formal Opinion 512.
“Johnson v. Dunn sanctioned a firm that had an AI policy and a committee. They didn’t have a gate. We build the gate.”— Editor’s margin note, Day 02
Cautionary cases
- i.Johnson v. Dunn
Firm had a policy and a committee. No gate. Sanctioned.
- ii.Mata v. Avianca
The canonical fabricated-citations cautionary tale.
- iii.Noland v. Land of the Free
Hallucinated authority, sanctions on the record.
The four guardrails
Reference.ABA Formal Opinion 512, July 2024 — Generative AI & the Model Rules.
Competence
Understand the tool you’re putting your name on.
Confidentiality
Know where the data goes. Default-deny on training.
Candor
What the tribunal sees came from a verified source.
Supervision
A human signed every citation before it left the building.
After the workshop · The on-ramp
The AI Opportunity Audit for law firms.
Quoted in conversation
Engagement memo · sized to a small firm
The AI Opportunity Audit
A fixed-scope engagement, sized to a small firm. We sit with the partners and the ops lead, walk the actual work, and come back with a roadmap real enough to act on Monday. No 80-page PDF. No abstractions.
No. 01
Top 3 workflows mapped, scored for payback.
No. 02
Prioritised Amplifier Stack roadmap (Codify → Automate → Amplify).
No. 03
One-page AI governance checklist your insurer will recognise.
No. 04
Direct hand-off path: the audit fee credits toward a build.
FeeDiscussed in the conversation, not on a marketing page.
hello@digitalboutique.ai →Solo + small-firm engagements only · we don’t take BigLaw work.
Anticipated objections, answered straight
The six questions we get every week.
- Q. 1On price
Is it really free?
Yes. No upsell mid-session, no card to register. The workshop is a sample of how we think — we’d rather you decide on real work than on a marketing page.
- Q. 2On audience
Who’s it for?
Practicing legal professionals at solo and small firms (2–50 attorneys), and the operations leads inside them. If you’ve been researching Artificial Intelligence (AI) for months without shipping anything, you’re the audience.
- Q. 3On the work
Will AI replace my associates — or me?
AI replaces tasks. Judgment, accountability, candor, and the client relationship are not tasks. The shift is from doing the task to setting the judgment criteria the task runs against. That’s the workshop.
- Q. 4On ethics
What about confidentiality and the ethics rules?
The verification gate we teach is mapped, line by line, to American Bar Association (ABA) Formal Opinion 512: Competence, Confidentiality, Candor, Supervision. Default-deny on training data. Human signs every citation. The gate fails closed.
- Q. 5On skill
Do I need to be technical?
No. If you can write a memo describing how you decide something, you can build with us. Codify first. The wiring is our job.
- Q. 6On the next step
What happens after the workshop?
You leave with one Claude workflow you actually own and the verification gate to ship it safely. If you want to go further, the AI Opportunity Audit is the on-ramp; the fee credits toward a build.
Reserve a Seat3 days.
3 days.
No demo theater.
3 days · virtual · cohort capped
Open to practicing legal professionals only.
We’ll send a calendar hold, the join link, and a short pre-read so we can start at the question, not the orientation slide. Next cohort dates are announced by email as soon as the current cohort fills.