The Problem: AI Output Can Look Finished Before It Is Reliable

Most law firms now use AI somewhere — research, drafting, intake, correspondence. Far fewer can say their AI-assisted work is consistent, checkable, and supervised. If any of the following sounds like your firm, you have a workflow problem, not a technology problem:

• The system rewrites the assignment instead of performing it
• Facts or documents you supplied are ignored
• Propositions are stated without reliable authority
• Citations are inaccurate, fabricated, or do not support the proposition
• Output changes materially when the same task is run twice
• Attorneys and staff prompt differently and get different results
• Confidential information is entered with no data-handling rule in place
• Nobody knows which portions of the output require factual, legal, citation, or supervisory review
• The same corrections are made repeatedly but never converted into better instructions
• A useful individual experiment never becomes a firm workflow

These are not prompting problems. They are workflow-design, source-control, testing, supervision, and quality-assurance problems.

The regulatory context is moving. ABA Formal Opinion 512 addresses competence, confidentiality, supervision, candor, and fees when lawyers use generative AI. New York’s Part 161 permits AI-assisted court papers while requiring independent attorney review. Appellate courts have sanctioned attorneys and firms for filing unverified AI-generated authority. This service is designed to support the professional-responsibility obligations that apply to your practice — it identifies issues requiring review by jurisdiction-specific ethics counsel where those arise.

What Gets Reviewed

An audit examines the full chain, not just the prompt: system instructions and master prompts; reusable prompts and templates; the facts, documents, and source material supplied; model or platform choice; confidentiality and data-handling restrictions; source-verification requirements; drift points; recurring corrections; review, approval, and escalation checkpoints; attorney–paralegal–staff handoffs; and the final handoff into your practice management system.

The goal is not a longer prompt. It is a controlled process that generates more usable work, exposes uncertainty, and tells the human reviewer what must still be checked.

Workflows That Can Be Audited

Legal Research & Analysis Authority verification, controlling-vs.-persuasive discipline, and research memos that expose uncertainty instead of hiding it.	Drafting & Document Preparation Correspondence, pleadings, motions, affidavits, agreements, estate-planning documents, and bankruptcy papers.	Document Review & Case Analysis Extracting parties, dates, deadlines, and governing provisions — with traceability back to the source documents.
Client Intake & Matter Opening Consistent screening, data capture, and handoff from inquiry to open matter.	Client & Internal Communications Status updates, plain-language client letters, and internal summaries that stay accurate under volume.	Practice-Specific Workflows Processes built around a particular court, agency, or matter type your firm handles at volume.

Engagement Options

Every engagement starts with the $500 Working Session. It is a deliverable in itself — you leave with a diagnosis and a prioritized action list whether or not you engage further — and the fee is credited toward any larger consulting engagement booked within 30 days. Larger engagements are scoped in writing before work begins.

Service	Fee	What You Receive
AI Workflow Working Session	$500 (payable at booking; credited toward any larger consulting engagement within 30 days)	90-minute nonlegal AI workflow consulting session: pre-session questionnaire, screen-share review of your highest-friction workflow, written summary of failure points, and a prioritized next-step list. If the session is delivered, the fee is earned and nonrefundable. You may reschedule once with at least 24 hours’ notice; cancellation with less notice or non-attendance forfeits the fee.
Focused Workflow Audit	$1,250 flat (credited toward larger consulting engagements)	One nonlegal workflow audit engagement — scope and deliverables confirmed in writing before work begins. Includes: pre-audit questionnaire, screen-share walkthrough, failure-point diagnosis, model and tool assessment, confidentiality and source-control review, revised instructions, review checklist, sample test cases, and a prioritized implementation plan.
Workflow Reliability Sprint	$3,500 flat	Full repair of one workflow — scope and deliverables confirmed in writing before work begins. Includes: redesigned instructions and prompts, source and citation controls, staged drafting and review procedures, model comparison, adversarial test cases, testing against closed or anonymized matters, written SOPs and templates, implementation training, and post-launch refinement.
Firm AI Governance & Implementation	Starting at $6,500 (quoted after scope review)	Nonlegal governance consulting, quoted after scope review. Includes: tool inventory; permitted, restricted, and prohibited-use categories; confidentiality controls; vendor-review criteria; output-verification requirements; supervision standards; court-filing and citation-review procedures; error-escalation paths; approved workflow templates; attorney and staff training; and onboarding materials.

Book a $500 Working Session

The Review Method

1. Map the existing process. What actually happens today — not what the manual says.

2. Establish the baseline. Test the current workflow against representative matters and document where it fails.

3. Diagnose the cause. Instructions, source material, model choice, context length, conflicting prompts, weak review gates — or a task AI should not be doing.

4. Rebuild and test. Against normal, difficult, and adversarial examples.

5. Document the human controls. What AI may do, what it may not do, what must be verified, who reviews it, and when to escalate.

6. Hand off a repeatable system the firm owns. Instructions, checklists, and SOPs your team runs without me.

Why a Practicing Lawyer — Not a Technology Consultant

This service is built inside an active New York practice spanning bankruptcy, tax, traffic and criminal matters, estate planning, consumer protection, business law, real estate, and civil litigation — where AI is used daily in real research, drafting, document analysis, intake, and communications. That matters because the failure modes that put firms at risk are legal failure modes, and only a lawyer reliably asks the questions that expose them:

Can every factual assertion be traced to a source document? Did the system answer the assignment actually given? Is the cited authority real, current, and supportive of the proposition? Is controlling authority distinguished from persuasive? Are deadlines, procedural posture, venue, and unresolved facts identified? Is attorney review meaningful or ceremonial? Can a paralegal run the workflow consistently? Can the firm explain, afterward, how the work was checked?

The diagnostic edge is distinguishing a prompt problem from a source problem from a model problem from a human-review problem — because the fixes are entirely different.

No Vendor Sales Pitch

This review is model-agnostic and sells no software. The recommendation may be to improve the tool you already use, restructure your inputs, split one task into stages, switch models for a single function, add a verification step, restrict AI for certain tasks — or stop using AI where it creates more risk than value. A responsible review must be willing to conclude that AI is not appropriate for a particular task.

Confidentiality & Data Handling

Audit work is performed on anonymized materials, closed matters, or sample documents — no live client files are ordinarily required, and no access to your case-management system is required. A mutual NDA is standard. A secure transfer method is established before any materials move. Your firm owns all customized deliverables: instructions, SOPs, checklists, and templates. Do not transmit confidential, privileged, or time-sensitive information through this website or booking form.

Frequently Asked Questions

Does my firm need to be technically sophisticated to benefit?

No. The review meets your firm where it is. Many engagements start with a single attorney or paralegal using an AI tool inconsistently — that is a normal starting point, not a disqualifier.

Do you need access to our active client files?

Ordinarily, no. Testing is done against anonymized materials, closed matters, or sample documents you approve in advance.

Can you guarantee the AI won’t produce incorrect output?

No — and no honest consultant can. The purpose of this work is to reduce predictable failures and make the remaining errors easier to catch through structured review. Independent professional review always remains required.

Which AI platforms do you work with?

The review is platform-agnostic. The methodology — instructions, source control, testing, review gates — applies regardless of which model or vendor your firm uses.

Do you guarantee that a particular AI tool is compliant or safe for client information?

No. Consulting may include evaluation of third-party software or AI tools, but no representation is made that any third-party tool is legally compliant, privileged, secure for legal-client information, or appropriate for any particular legal matter. Tool selection always remains subject to your firm’s own vendor review and professional judgment.

We already built our own prompts and workflows. Is that a problem?

The opposite — existing workflows are often the best starting point. Something your firm built and actually uses tells us what the work really requires; the audit tests it, finds where it fails, and strengthens it.

Do you train our staff?

Yes. Sprint and Governance engagements include implementation training for attorneys and staff, and training can be added to a Focused Audit.

Is this legal advice?

No. These consulting services are not legal services or legal advice, no attorney-client relationship is formed, and the protections of a client-lawyer relationship — including attorney-client privilege and legal-client confidentiality — do not apply to this consulting. Ronald S. Cook is admitted to practice law in New York only; New York legal services are available under a separate written legal-services engagement. Firms in other jurisdictions should consult locally admitted counsel for jurisdiction-specific legal questions.

These AI workflow consulting services are not legal services or legal advice, do not create an attorney-client relationship, and do not include attorney-client privilege or legal-client confidentiality protections. AI tools are not substitutes for attorney judgment. Lawyers remain responsible for their work product, supervision, representations to courts, cited authorities, filings, and client communications. Consulting services do not guarantee AI accuracy and do not eliminate the need for independent professional review. Ronald S. Cook, Esq. is admitted to practice law in New York only; the availability of nationwide nonlegal consulting services is not a representation of admission in any other jurisdiction. No attorney-client relationship is formed through this website.