The automation is the easy part. The ethics are not.

Why generic AI is the wrong starting point

Any developer can bolt a chatbot onto a law-firm site. The hard part is keeping it on the right side of confidentiality, unauthorized practice, and attorney-advertising rules. That is the part I own, as a California attorney who also builds these tools.

Off-the-shelf AI is built for retail, not regulated practice

Generic chatbots give legal conclusions, promise outcomes, and imply representation. On a law-firm site that is a discipline problem, not a feature.

A lawyer who also builds the systems

I am a California attorney who builds and runs these tools daily. I review the architecture and the disclaimers, then design escalation so a licensed attorney stays in the loop.

Automation type

Pick the kind of automation you are dealing with

Open the automation-type picker

Five common builds: an intake chatbox, structured legal intake and routing, a document generator, an AI implementation audit, or a website and disclaimer review. Each tab below covers what it is, what I build, and who it is for.

AI intake chatbox

Most requested

A conversational front door that answers general questions, gathers facts, and routes serious matters to a human, without crossing into legal advice.

💬 What it is

  • Visitor-facing assistant on your site
  • General information plus fact-gathering
  • Hands off to a person on real matters

🔧 What I build

  • Scope, prompt, and guardrail design
  • Disclaimers and consent text
  • Escalation and human-review routing

👤 Who it is for

  • Law firms wanting after-hours intake
  • Professional-service businesses
  • Anyone whose current bot over-promises
What I build

The components I design and review

See the eight components

Eight discrete pieces I build into a new system or fix inside an existing one. Most engagements combine three or four.

💬

Intake chatboxes

Attorney-supervised assistants scoped to inform and gather facts, never to give jurisdiction-specific advice.

📐

Matter routing

Classification and routing logic that sorts each inquiry into the right matter type and the right next step.

📝

Document-generation tools

Tools that turn structured inputs into draft documents using attorney-set authority rules.

🔎

Website & disclaimer review

A focused pass on public AI claims, chatbot disclaimers, and no-advice and consent language.

📊

Workbook-to-document workflows

Pipelines that take one structured workbook and generate batches of drafts for review.

Engagement-letter automation

Generation of engagement and disclosure letters from structured inputs, with review routing built in.

📜

AI use policy

An internal policy governing what staff may put into AI tools, what is off-limits, and who supervises.

🛡

Escalation rules

Explicit triggers that hand a matter to a licensed attorney the moment it crosses into legal judgment.

Risks I fix

Where unmanaged legal AI gets firms in trouble

Confidentiality & data handling

Client facts or documents flowing into third-party AI without a data agreement, retention control, or training opt-out.

High
How I reduce this risk
  • Map exactly what client data each tool sees and where it goes.
  • Lock down vendor terms, retention windows, and any training opt-out.
  • Write explicit rules for what the tool may and may not ingest.

Unauthorized practice of law (UPL)

An automated tool giving jurisdiction-specific legal conclusions or applying law to a user's facts without a licensed attorney.

High
How I reduce this risk
  • Scope the tool to general information and fact-gathering only.
  • Build escalation that routes anything resembling advice to a licensed attorney.
  • Strip jurisdiction-specific legal conclusions from automated replies.

Overpromising & outcome guarantees

Marketing or chatbot copy that predicts results or implies a guarantee, which collides with attorney-advertising rules.

Medium
How I reduce this risk
  • Rewrite public claims to be accurate and non-promissory.
  • Strip outcome and guarantee language from automated replies.
  • Align the copy with attorney-advertising rules before it goes live.

Client confusion & attorney-client formation

Users believing a chatbot created an attorney-client relationship, or that its output is vetted legal advice.

Medium
How I reduce this risk
  • Add clear no-advice and no-relationship notices where users actually see them.
  • Put consent gates in front of anything that could imply representation.
  • Route serious matters to a human so expectations match reality.
Deep diveHow I score these risks, and the authorities behind them

The four failure modes above are the ones I look for first. Each gauge shows typical exposure for an unmanaged off-the-shelf deployment; expand a gauge for how I reduce it. The ratings are calibrated to the common case, and your facts can move the needle either way.

These ratings are a starting framework, not a substitute for a review of your actual setup. Confidentiality and UPL sit at the top because they map to the duties most likely to draw a bar inquiry or a malpractice claim: the duty of confidentiality and the prohibition on the unauthorized practice of law. Advertising and client-confusion risks are graded medium because they are usually fixable with disclosure and copy changes, though they can escalate quickly if a tool predicts outcomes or implies representation.

In a paid review I cite the operative authorities for your jurisdiction. For California work that includes the Rules of Professional Conduct on competence, confidentiality, fees, supervision of non-lawyer assistance, and communication, plus Business and Professions Code provisions on the unauthorized practice of law and attorney advertising. I confirm the current operative text of any rule before relying on it; I do not quote from memory.

I am a California attorney. For matters governed by another state's rules, I will tell you plainly where you need local counsel rather than guess at another jurisdiction's bar rules.

Case study (anonymized)

One workbook, large batches of engagement letters

Workbook to engagement letters

Structured input -> authority rules -> generated drafts -> attorney-review gate

Open case studyProfessional-service firm: workbook-driven engagement-letter generation

A worked example of the architecture I design. Details are generalized and the client is not identified.

How the system is structured

A professional-service firm needed to produce large batches of engagement letters from a single structured workbook, without a person retyping each one. I designed the rules and the architecture; the tool generates drafts from structured inputs; where legal review is required, the workflow routes documents for attorney review.

STEP 1

Structured workbook

One spreadsheet of validated inputs is the single source of truth.

STEP 2

Attorney-set rules

I encode the authority rules and clause logic that govern each draft.

STEP 3

Drafts generated

The tool, with an attorney-supervised AI helper, produces draft letters at volume.

STEP 4

Routed for review

Where legal judgment is required, documents route to attorney review before use.

Important: the tool generates drafts from structured inputs. The drafts are not automatically attorney-vetted by being produced. Documents that require legal review are routed for attorney review in the workflow before they are relied on or sent.

What I owned: the rule design, the clause logic, the architecture of the workflow, the attorney-review routing, and the disclaimers and use constraints around the AI helper.

What the firm got: a repeatable pipeline that turns one structured workbook into batches of consistent drafts, with a clear line marking where a licensed attorney reviews before anything is used. The throughput gain comes from removing manual retyping, not from removing the lawyer.

What I do not claim: I do not claim the machine output is finished legal work the instant it is generated, and I do not design these systems to imply that. The value is a disciplined pipeline with the review step built in, not an unsupervised document machine.

This site is the live demo

You are standing inside one of the systems I build

See the six live examples on this site

The chatbox, calculators, generators, intake flows, and cross-linking on this site are working examples of what I build. Try the AI Legal Analyst in the corner; the rest are below.

Live

AI Legal Analyst

The attorney-supervised chatbox on this page, scoped to inform, not advise.

Live

Calculators

Tools that compute a result, then explain it in plain language.

Live

Document generators

Structured-input tools that produce draft documents from your answers.

Live

Intake flows

Guided question sets that capture the right facts and route a clean summary to a human.

Live

Related-resource routing

Cross-linking that sends each visitor to the next most relevant page.

Live

Disclaimers in context

Every tool here carries a no-advice, no-relationship notice I write and review.

Packages & pricing

Start with a review, or scope a full build

How the pricing works

Flat-fee reviews start instantly. Build, configuration, and audit engagements start by intake so I can size the work and quote it accurately before anyone pays.

Best starting point
AI website / chatbot risk review
$575

A written findings memo on your public AI claims, chatbot scope, disclaimers, and UPL exposure, with specific fixes ranked by risk.

  • Public AI claims & advertising language
  • Chatbot scope, disclaimers & consent
  • Confidentiality & data-handling claims
  • Attorney-client-formation risk
  • Prioritized fix list with sample copy
Request this review

Flat fee, written deliverable. Email intake to start.

Legal automation build & configuration
$1,500–$2,500+

Design and configuration of a new intake chatbox, document-generation workflow, or routing system, scoped to your practice.

  • Scope, prompt & guardrail design
  • Disclaimers & escalation rules
  • Matter routing & intake question sets
  • Configuration with your vendor stack
Start intake

Scoped & quoted by intake, not instant checkout.

AI implementation audit
$2,500+

A full review of how AI touches your practice: data flows, client-facing claims, advice and formation risk, supervision, with a ranked findings memo.

  • Inventory of every AI touchpoint
  • Confidentiality & data-flow map
  • Ethics & advertising-rule check
  • Use policy & supervision rules
Start intake

Scoped & quoted by intake, not instant checkout.

Tuning & automation-counsel retainer
Monthly, optional

Ongoing tuning, prompt and disclaimer updates, and standing automation counsel as your tools and the rules evolve.

  • Periodic tuning & copy updates
  • Disclaimer & policy refresh
  • Standing automation-counsel access
  • Scoped to your volume
Discuss a retainer

Scoped to your volume. Email intake to start.

Prefer a written second opinion first?

Prefer a written second opinion before committing to a build? Start with a $240 Written Attorney Consultation or a $400 1-Hour Zoom Strategy Session.

What I need from you

Send these and I can start

Open the intake checklist

You do not need to share privileged client data to begin. Send the public-facing material and a description of the workflow; I will tell you if a deeper review needs anything redacted first.

Website URL
The pages where AI features appear.
Chatbot link
A link or screenshots of the live tool.
Practice areas
What kinds of matters the tool touches.
Jurisdictions
Which states or countries you serve.
Intake questions
The questions your flow asks today.
Disclaimers
Your current no-advice and consent text.
Tool / vendor stack
The platforms and AI models behind it.
Sample leads
Redacted examples of what comes through.
Conversion goals
What a successful outcome looks like.

If any of this lives in a link I cannot open from your message (a shared drive, a portal, a login-gated page), tell me and I will sort out how to get it, rather than guessing at the contents.

FAQ

Common questions

Will an AI intake chatbox commit unauthorized practice of law?

It can, if it gives jurisdiction-specific legal conclusions, predicts outcomes, or forms an attorney-client relationship without supervision. I scope the tool to gather facts, route, and explain general information, and I write the disclaimers and escalation rules so a licensed attorney stays in the loop on anything that crosses into advice.

Do I have to name a client to show you my current setup?

No. Send me the public URL, the chatbot link, and a description of the workflow. I do not need privileged client data to review the system architecture and the disclaimers. If a review needs sample transcripts, I tell you what to redact first.

Is the AI Legal Analyst on this site legal advice?

No. It is an attorney-supervised AI Legal Analyst that provides general information, not legal advice, and it does not form an attorney-client relationship. It is also a live demonstration of the kind of intake system I build for clients.

What does the $575 AI website and chatbot risk review cover?

It covers your public AI claims, chatbot scope and disclaimers, confidentiality and data-handling representations, attorney-client-formation risk, and unauthorized-practice exposure, delivered as a written findings memo with specific fixes prioritized by risk. It is a flat fee with a written deliverable.

Can you build the automation, or only review it?

Both. I review existing tools at a flat fee, and I design and configure new intake chatboxes, document-generation workflows, and escalation rules on a scoped build engagement. Build and audit engagements start by intake so I can size the work before quoting, which is why those tiers use an intake request rather than an instant payment button.

Do you guarantee the AI will be error-free?

No, and you should be skeptical of anyone who does. AI systems produce errors. My job is to design the guardrails, disclaimers, human-review routing, and use policy so that errors are contained and a licensed attorney reviews anything that requires legal judgment.

I am outside California. Can you still help?

For system architecture, disclaimers, and AI-governance work, often yes. For state-specific bar rules outside California, I will tell you plainly where you need local counsel rather than opine on another jurisdiction's professional-conduct rules. I am licensed in California (Bar No. 279869).

Get started

Let me review your AI setup, or build it right the first time

Send the public URL, the chatbot link, and what you want reviewed or built. You work directly with me, a California attorney, with no intake team in between.

Informational, not legal advice. This page and the AI Legal Analyst provide general information only and do not constitute legal advice. Using this site, the chatbox, or any tool on it does not create an attorney-client relationship. Any AI-generated content is attorney-supervised general information, not vetted legal advice for your situation, and it is not a substitute for review by a licensed attorney. Sergei Tokmakov is licensed in California (State Bar No. 279869); for matters governed by another jurisdiction's rules you may need local counsel. Flat-fee amounts are starting points; build, audit, and retainer engagements are scoped and quoted by intake. Contact me at owner@terms.law.

© 2025 Terms.Law · Sergei Tokmakov, Esq., California Bar No. 279869 · Outside General Counsel