Stay Connected: Client Communication That Works

AI for Client Communications

Client communication is where legal work becomes visible. It's where trust is built or lost, where clarity matters as much as accuracy, and where attorneys routinely spend 30–40% of their billable time. AI tools. ChatGPT Plus, Claude Pro, Microsoft Copilot, and others, can draft, refine, translate, and personalize client-facing communications in minutes. This lesson is your reference guide: what AI can do, what it cannot do, and exactly how to use it without compromising your professional obligations.

7 Things to Know Before You Start

1. AI drafts, you decide. Every AI-generated communication requires attorney review before it reaches a client. Period.
2. Confidentiality rules still apply. Pasting a client's full name, case number, and sensitive facts into a public AI tool (like the free version of ChatGPT) is a potential ethics violation in most jurisdictions.
3. AI is better at tone than law. These tools excel at plain-language rewrites, empathy adjustments, and structure, not at providing accurate legal analyzis in client letters.
4. Your firm's policy matters. As of 2024, roughly 40% of Am Law 200 firms have formal AI use policies. Know yours before you use any tool for client-facing work.
5. Different tools for different tasks. Claude Pro handles long documents and nuanced tone better than most. Copilot integrates directly into Outlook and Word. ChatGPT Plus is the most versatile for one-off drafting tasks.
6. Templates beat blank-page prompting. Giving AI a sample letter from your practice area produces far better output than starting from scratch.
7. Plain language is a professional asset, not a compromise. Studies show clients who receive plain-language legal communications are more likely to follow advice, pay invoices on time, and refer new clients.

What AI Actually Does With Your Client Emails

When you give an AI tool a client email to improve, it's doing several things simultaneously: adjusting vocabulary complexity, restructuring sentences for clarity, shifting the emotional register, and flagging potential ambiguities. Think of it as having a very fast editor who has read ten thousand professional letters but has never practiced law. It knows how communications should sound and feel. It does not know whether the legal content is correct. That distinction is the most important thing to internalize before using any AI tool for client work.

In practice, this means AI is most valuable for the structural and linguistic layer of client communication, not the substantive layer. A partner who spends 20 minutes crafting the right tone for a difficult billing dispute email can now get a strong first draft in 90 seconds. A junior associate who struggles to translate a dense procedural update into plain English for an anxious client can use Claude or ChatGPT to bridge that gap. The legal judgment stays with you. The sentence-level work moves to the tool.

• Rewrites dense legal prose into plain English without losing meaning
• Adjusts formality level, from formal engagement letter to accessible status update
• Restructures long emails into clear sections with headers and bullet points
• Softens difficult messages: bad news delivery, billing disputes, missed deadlines
• Translates communications into other languages (with professional review required)
• Generates multiple draft versions with different tones for you to choose from
• Summarizes long email threads into a short brief before you respond
• Flags phrases that may read as legally risky (e.g., admissions, guarantees)

Click to enlarge

AI Tool Comparison for Client Communications

Click to enlarge

The Confidentiality Problem, and How to Work Around It

This is the issue most legal professionals hit first. Consumer AI tools, the free or standard paid tiers of ChatGPT and Claude, may use your inputs to improve their models. Pasting a real client's name, matter details, and sensitive facts into these tools raises legitimate concerns under ABA Model Rule 1.6 (confidentiality) and analogous state rules. The California State Bar, New York City Bar, and Florida Bar have all issued guidance on this. The risk is real, and it's not theoretical.

The practical solution is a technique called anonymization before prompting. Before you paste anything into a consumer AI tool, strip out all identifying information. Replace the client's name with 'Client A,' the opposing party with 'Party B,' the specific property address with 'the subject property,' and the case number with a placeholder. The AI doesn't need real names to help you draft a better email, it needs context, tone, and structure. Once the draft comes back, you reinsert the real details. This takes about 60 seconds and keeps you on the right side of your ethical obligations.

1. Replace all client names with neutral placeholders (Client A, Client B)
2. Remove case numbers, matter IDs, and docket references
3. Swap specific addresses, company names, and financial figures with generic descriptors
4. Replace opposing counsel's name and firm with 'opposing counsel'
5. Remove any details that could identify the jurisdiction or specific court
6. Paste the anonymized version into the AI tool
7. Review the output, then reinsert the real details manually before sending

Click to enlarge

Prompt Example: Rewriting a Status Update Email

Tone Calibration: The Most Underused AI Skill in Legal Practice

Legal professionals are trained to be precise. That training sometimes produces communications that are technically accurate but emotionally tone-deaf. A client facing a custody dispute, a business owner dealing with a contract breach, or an employee navigating a wrongful termination claim is not just receiving information, they're in a high-stress situation. AI tools can help attorneys calibrate tone in a way that feels unnatural to draft from scratch. You can explicitly tell the tool: 'This client is anxious and has emailed three times this week. The tone should be reassuring but realiztic, not falsely optimiztic.'

Tone calibration goes beyond empathy. It includes managing expectations without creating false hope, delivering bad news without triggering panic, and maintaining professionalism while still sounding like a human being rather than a form letter. AI tools are remarkably good at this when you give them explicit tone instructions. The key is specificity: 'professional and warm' is vague. 'Reassuring, honest about uncertainty, and focused on the client's next action' gives the tool something to work with. The more precisely you describe the emotional register you need, the better the first draft.

Click to enlarge

Practice Task: Rewrite a Client Email Using AI

Part 1 Cheat Sheet

• AI handles tone, structure, and plain language, not legal substance. Review everything before it reaches a client.
• Anonymize before you paste: replace names, case numbers, addresses, and identifying details with placeholders.
• Use the Role + Context + Output formula for faster, better first drafts.
• Consumer AI tiers (ChatGPT Plus, Claude Pro) may process your inputs. Enterprise plans (M365 Copilot, ChatGPT Enterprise) offer stronger data controls.
• Tone calibration requires specific instructions, 'professional and warm' is not enough. Describe the emotional register precisely.
• AI has no memory of your client relationship. You must edit the draft to reflect what you know about the person.
• Best tools by use case: Copilot for Outlook/Word integration, Claude Pro for nuanced tone and long documents, ChatGPT Plus for versatile one-off drafting, Harvey AI for legal-specific outputs.
• Plain-language client communications improve follow-through, reduce confusion callbacks, and support timely payment.
• Check your firm's AI use policy before using any tool for client-facing work.

Key Takeaways From Part 1

• AI tools are editing partners, not legal advisors. The value is in communication quality, not legal analyzis.
• Confidentiality risk is real but manageable. Anonymization is your first line of defense on consumer tools.
• Tone is a skill AI can augment, but only if you give it precise emotional instructions, not vague adjectives.
• Different tools have meaningfully different strengths. Match the tool to the task rather than defaulting to one option.
• The professional standard doesn't change because AI wrote the first draft. Your signature, your responsibility.

Once you understand what AI can do for client communications, the real skill is knowing exactly how to direct it. The difference between generic output and genuinely useful drafts comes down to specificity, how you frame your request, what context you provide, and which guardrails you set. This section covers the practical mechanics: tone calibration, communication type selection, and the quality checks that keep AI-assisted correspondence professional and ethically sound.

7 Things Every Legal Professional Must Know About AI Client Communications

1. AI drafts require attorney review before sending, every single time, no exceptions for 'routine' messages.
2. Jurisdiction matters: AI tools have no automatic awareness of your state bar rules or local court requirements.
3. Confidentiality warnings apply even to free-tier tools, client names and case facts can be training data.
4. Tone is controllable: you can instruct AI to write formally, conversationally, empathetically, or assertively.
5. AI cannot assess what a client actually understood, it only drafts what you instruct it to say.
6. Response summaries generated by AI from client emails can miss emotional subtext and urgency cues.
7. Saved prompt templates dramatically reduce drafting time, build a library specific to your practice area.

controlling Tone and Reading Level in Client Drafts

Most client complaints about lawyers center on communication, not competence. Clients say they felt confused, ignored, or talked down to. AI gives you a practical lever to fix this. When you instruct a tool like ChatGPT or Claude to write at a specific reading level or in a particular emotional register, it adjusts vocabulary, sentence length, and structure accordingly. A demand letter and a settlement update require completely different tones. Writing both from scratch under time pressure invites inconsistency. AI handles the tonal shift on command.

The most effective approach is to specify the client's situation alongside the tone. 'Write this for a small business owner who is anxious about litigation costs' produces a different result than 'write this for a sophisticated commercial client.' You are not guessing at readability, you are engineering it. For clients with limited English proficiency or low legal literacy, instructing AI to write at a 6th-grade reading level while avoiding all legal jargon produces dramatically clearer correspondence. Review the output, adjust any legal inaccuracies, and send with confidence.

• Formal/professional: Use for opposing counsel communications, court-related updates, retainer agreements.
• Plain English: Use for status updates, fee explanations, next-steps summaries for individual clients.
• Empathetic: Use for sensitive matters, family law, employment disputes, personal injury, estate planning.
• Direct and assertive: Use for demand letters, deadline reminders, non-responsive client follow-ups.
• Neutral and factual: Use for document request lists, deposition prep instructions, settlement breakdowns.
• Reassuring: Use when delivering difficult news, adverse rulings, settlement offers below expectation.

Tone and Use Case Reference Table

Summarizing Lengthy Client Correspondence

Long email threads from clients are a daily reality in legal practice. A client emails seven times over a weekend, each message adding new facts, questions, and concerns. Before you can respond, you need to understand the full picture. AI excels at condensing these threads into structured summaries. Paste the full email chain into ChatGPT or Claude and ask for a bulleted summary of key facts, open questions, and any deadlines mentioned. What might take 12 minutes of careful reading takes 90 seconds with AI, and the structured output is easier to work from than the raw thread.

The same technique applies to lengthy client-provided documents: timelines they've written, their own summaries of events, or background narratives they send before intake calls. Ask AI to extract the five most legally relevant facts, identify any internal contradictions, and flag dates or names that appear more than once. This is not AI doing legal analyzis, it is AI doing document triage, which is exactly the kind of time-consuming, low-judgment work that eats into your billable day. The legal analyzis remains yours. The sorting is AI's.

1. Open ChatGPT Plus or Claude Pro in your browser.
2. Paste the full email thread or client document into the chat window.
3. Type: 'Summarize this in bullet points. Extract: key facts, open questions, any deadlines, and any emotional concerns expressed.'
4. Review the output, verify dates, names, and facts against the original.
5. Add any corrections or missing details directly in the chat: 'The date in point 3 should be March 15, not March 5.'
6. Ask a follow-up: 'Based on this summary, draft a brief acknowledgment email confirming I've received their concerns and will respond by [date].'
7. Copy the draft to your email client, review for accuracy and tone, add any legal caveats, and send.

Email Summarization and Response Workflow

Click to enlarge

Prompt Example: Drafting a Settlement Update Email

Confidentiality, Ethics, and the Lines You Cannot Cross

Every bar association in the United States requires attorneys to protect client confidential information. Model Rule 1.6 is the foundation. When you paste client names, case facts, or identifying details into a consumer AI tool, you may be exposing that information to a system that uses inputs for model training, stores conversation data, or allows human reviewers to audit outputs. The ABA's Formal Opinion 512 (2024) directly addresses this: attorneys must understand the data practices of any technology they use before inputting client information. Ignorance is not a defense to a disciplinary complaint.

The practical solution is not to avoid AI, it is to use it correctly. Anonymize inputs before pasting: replace client names with 'Client A,' swap identifying case details with generic descriptions, and never include Social Security numbers, financial account details, or medical record numbers in a prompt. Enterprise versions of tools like ChatGPT Enterprise, Microsoft Copilot for Microsoft 365, and Claude for Business contractually commit to not using your inputs for training and offer stronger data protection agreements. These are the versions firms should be deploying, not the free consumer tiers.

Practice Task: Draft a Client Status Update Using AI

Quick Reference: AI Client Communications Cheat Sheet

• Always anonymize: Replace client names and identifying details before pasting into any AI tool.
• Specify tone in every prompt: Formal, empathetic, plain English, assertive. AI needs direction.
• Include word count limits: 'Under 150 words' prevents bloated drafts that clients won't read.
• Use enterprise tools for real client data: ChatGPT Enterprise, Copilot for M365, Claude for Business.
• Fact-check every output: AI can confidently state incorrect dates, amounts, or legal standards.
• Build a prompt library: Save your best prompts as templates in a shared doc for your whole team.
• Summarize before you respond: Paste long email threads and ask for a structured summary first.
• Match tool to task: Outlook Copilot for email, Claude Pro for sensitive drafts, ChatGPT for volume.
• Review aloud before sending: Catches tone problems and awkward phrasing that reading misses.
• Document your AI use policy: Every firm using AI for client communications needs written guidelines.

Key Takeaways from Part 2

• Tone is the most controllable variable in AI-generated correspondence, specify it explicitly in every prompt.
• AI summarization of long email threads is one of the highest-ROI uses available to legal professionals right now.
• Consumer AI tools carry real confidentiality risk; enterprise versions are the appropriate choice for client data.
• ABA Formal Opinion 512 (2024) makes clear that using AI without understanding its data practices is an ethics issue, not just a tech issue.
• The quality gap between a generic AI prompt and a well-constructed one is enormous, context about the client and situation is the most important input you provide.

AI-assisted client communication is only as good as the guardrails around it. This section covers risk management, ethical obligations, and the practical systems that keep your AI-assisted communications professional, compliant, and defensible. Treat this page as your reference sheet, print it, bookmark it, return to it.

1. AI drafts require attorney review before any client-facing send, always.
2. Confidentiality obligations apply to every piece of client data you paste into an AI tool.
3. AI tools can hallucinate case citations, statutes, and deadlines, verify every legal reference independently.
4. Most bar associations now have formal or interim guidance on AI use in legal practice.
5. Tone calibration matters: AI defaults to generic professional language, not your firm's voice.
6. Client consent to AI use may be required depending on your jurisdiction and matter type.
7. Audit trails protect you, document when and how AI was used in any client communication.

Confidentiality and Data Handling

Pasting client information into a consumer AI tool, names, case facts, financial details, opposing party information, can constitute a confidentiality breach under Model Rule 1.6. The risk is real. Free versions of ChatGPT and Claude may use your inputs to train future models unless you explicitly opt out or use an enterprise plan. ChatGPT Enterprise and Claude for Enterprise both offer data-use agreements that prohibit training on your inputs. Before using any AI tool for client matters, verify the data retention policy and confirm your firm's IT or compliance team has approved the platform.

The safest immediate practice is anonymization. Before drafting, strip out identifying details, replace the client's name with 'Client A,' the opposing party with 'Party B,' and substitute generic descriptions for specific case facts. You get a perfectly usable draft and expose nothing confidential. Once the AI returns a draft, reinsert the correct details manually. This adds two minutes and eliminates the compliance exposure. Firms with approved enterprise AI licenses can skip this step, but should still maintain a record of which matters involved AI-assisted drafting.

• Free ChatGPT (GPT-3.5/4): May use inputs for training. Opt out in Settings → Data Controls.
• ChatGPT Plus / Enterprise: Enterprise plan includes zero data retention by default.
• Claude Pro: Anthropic states it does not train on Pro user conversations by default.
• Microsoft Copilot (M365): Governed by your organization's Microsoft tenant, generally safe for business data if your IT has enabled it.
• Google Gemini for Workspace: Subject to your Google Workspace agreement, check with IT.
• Notion AI: Data governed by Notion's enterprise terms, review before use on client matters.

Accuracy, Hallucinations, and Legal Verification

AI language models do not look things up, they predict plausible text based on patterns in their training data. This means they can generate citations to cases that do not exist, quote statutes with the wrong section numbers, or state a legal standard with confident authority and complete inaccuracy. The 2023 Mata v. Avianca case, where attorneys submitted AI-generated briefs containing fabricated case citations, resulted in sanctions and national news coverage. That risk is just as present in client communications that reference legal standards, timelines, or procedural requirements.

The rule is simple: never let an AI-generated legal reference reach a client without independent verification. Use AI for structure, tone, and plain-language explanation. Verify every statute number, case name, deadline, and procedural claim against a primary source. Westlaw, Lexis, the court's official website, or the relevant agency. Build this into your workflow as a non-negotiable final step, the same way you would proofread a signature block. AI drafts that contain no legal citations at all carry far lower risk and are often more appropriate for routine client updates.

1. Never include AI-generated case citations in client communications without Westlaw/Lexis verification.
2. Treat all AI-stated deadlines as unverified until confirmed against the court docket or governing rules.
3. Ask AI to explain concepts in plain language, this is low-risk and high-value.
4. Flag every legal standard or procedural claim in an AI draft with a highlight before review.
5. When in doubt, remove the legal reference entirely, the client update is still useful without it.
6. Document your verification step in your matter notes for defensibility.

Click to enlarge

Building a Sustainable AI Communication Workflow

The attorneys who get the most from AI are not the ones who use it occasionally, they are the ones who build repeatable systems. That means creating a small library of prompt templates for your most common communication types: status updates, intake confirmations, fee explanation letters, settlement summaries, and hearing preparation reminders. Store these in a shared document, a Notion page, or even a simple Word file. When a new matter requires that communication, you open the template, fill in the anonymized facts, run the draft, review, and send. Total time: under ten minutes for a polished, professional communication.

Consistency is the second payoff. When your client communications follow a reliable structure, clear subject lines, plain-language summaries, explicit next steps, and a warm but professional tone, clients feel informed and confident. That reduces inbound calls asking 'what's happening with my case?' which is one of the biggest time drains in legal practice. A well-designed AI-assisted communication workflow does not just save drafting time. It reduces client anxiety, improves perceived service quality, and frees attorney time for billable substantive work.

Click to enlarge

Quick-Reference Cheat Sheet

• Anonymize client data before pasting into any AI tool, replace names and facts with placeholders.
• Use enterprise-grade AI tools (ChatGPT Enterprise, Claude for Enterprise, Microsoft Copilot) for sensitive matters.
• Never send AI-generated legal citations to clients without verifying on Westlaw or Lexis.
• Use AI for plain-language explanations, this is its highest-value, lowest-risk use in legal comms.
• Always review AI drafts before sending, no exceptions.
• Check your state bar's current AI guidance, it is being updated frequently.
• Store your best prompts as reusable templates for consistent, efficient communication.
• Document AI use in your matter notes for audit trail purposes.
• Client consent to AI use may be required, check your jurisdiction.
• AI saves time on drafting, not on professional judgment, that part is always yours.

Key Takeaways

• Confidentiality rules apply to AI tools, anonymize data or use approved enterprise platforms.
• AI hallucinations are a documented, serious risk, verify every legal reference independently.
• Plain-language client updates are the safest and most valuable AI communication use case.
• A prompt template library turns occasional AI use into a reliable, time-saving workflow.
• Every AI-drafted communication requires attorney review before it reaches a client.
• Bar association guidance on AI is active and evolving, stay current with your jurisdiction.