80% of Your Clients Are Using AI. Here's the Workflow Your Firm Needs Right Now.

80% of Your Clients Are Using AI. Here's the Workflow Your Firm Needs Right Now.

A new study should stop you mid-coffee-sip.

Cordell & Cordell, one of the country's largest domestic litigation firms, surveyed 133 of their practicing family law attorneys over 20 days in May 2026. The finding: 80% of attorneys are already receiving AI-generated correspondence from clients. Fifty-nine percent are receiving documents from opposing counsel they suspect were AI-generated.

This is not a future trend. It is the current state of your intake queue.

Most coverage of AI in law has focused on one direction: attorneys using AI to draft documents, file briefs, and conduct research — and the sanctions that follow when they do it carelessly. That problem is real. But the Cordell study confirms a different problem is already larger: clients sending AI-generated documents to your firm, and the workflows most small firms have for handling that are nonexistent.

Here is what you need to put in place.

---

What the Cordell & Cordell Study Actually Found

The study focused on family law — a practice area where clients frequently prepare their own correspondence, financial disclosures, asset schedules, and factual narratives. These are exactly the kinds of documents where consumer AI tools like ChatGPT get used, and where AI-generated errors are hardest to catch because the source material is the client's own life, not public record.

The 80% figure is attorney self-report over a 20-day sample. That means attorneys recognized AI-generated content in documents coming from clients — or strongly suspected it. The actual percentage of AI-assisted client documents may be higher, because detection is imperfect.

The second number matters more for your risk profile: 59% of attorneys in the study reported receiving documents from opposing counsel they believe were AI-generated. That means AI isn't just in your client relationship — it is in the adversarial record. Documents drafted by the other side's clients, reviewed by opposing counsel, and submitted to the court.

The Cordell study covers family law, but the dynamic applies anywhere clients are preparing their own supporting materials: small business owners in commercial disputes, employees in employment matters, landlords and tenants in real estate cases, anyone responding to intake questionnaires with AI-drafted narrative.

If your clients can type into ChatGPT, they are probably already doing it.

---

The Three Client-AI Scenarios Your Firm Is Facing

Scenario 1: Client-submitted correspondence and intake materials

This is the most common. A client uses ChatGPT or Claude to help them draft a detailed email explaining their situation, respond to your intake questions, or summarize a financial dispute. The output is polished, organized, and may contain AI-invented specifics — dollar amounts that seem plausible but aren't quite right, dates that approximate the truth, descriptions of conversations the client told the AI happened rather than verified facts.

Risk: You relay those facts into a demand letter, a complaint, or a filing. The client approved it. The facts are wrong. Who verified them?

Scenario 2: Client-prepared self-help documents

A client preparing for a separation submits a property schedule, a custody proposal, or a financial affidavit — documents they drafted with AI assistance because they wanted to come prepared. The document looks professional. The numbers and descriptions may contain errors the client didn't notice because the AI confidently filled gaps.

Risk: Your firm incorporates their schedule into discovery responses or settlement negotiations. Opposing counsel finds discrepancies. The origin is client AI, but the representation is yours.

Scenario 3: Opposing counsel AI-generated filings and discovery responses

Your opposing counsel may not be carefully verifying what their AI-assisted client submitted. Discovery responses, factual assertions in settlement letters, and supporting declarations can all carry AI artifacts — generic phrasing, wrong dates, hallucinated specifics.

Risk: Lower than Scenarios 1 and 2, but not zero. If you rely on opposing counsel's factual representations without verification and they are AI-fabricated, you may have built a case on unstable ground.

---

Why 84% Confidence in AI Detection May Not Be Enough

The Cordell study also found that 84% of attorneys say they are confident they can detect AI-generated content. This sounds reassuring until you understand what that number actually measures.

It is self-reported confidence, not validated accuracy.

AI detection is hard. It is getting harder as AI writing quality improves. Detection tools (Copyleaks, Originality.ai) are imprecise and context-dependent. They struggle with short documents, mixed-origin content (a human writing with AI assistance), and non-English text. More importantly, a family law attorney who has seen ten AI-generated client letters may feel confident detecting the eleventh — but that confidence is based on pattern recognition in a rapidly shifting landscape.

The actual risk is not whether you can tell a document looks AI-generated. The risk is that AI-generated documents may contain specific factual errors that look completely plausible — incorrect amounts, wrong dates, invented details — that no detection tool flags because the prose style is fine. The problem is the content, not the style.

Confidence in detection does not solve the verification problem. A consistent review protocol does.

---

The Attorney-Client Privilege Trap in Client AI Documents

This is the legal exposure most small firms have not yet mapped.

In February 2026, a federal court ruled that documents created using publicly available consumer AI tools are not protected by attorney-client privilege. The court's reasoning: consumer AI platforms process user inputs through commercial third-party infrastructure. A client using ChatGPT.com or Claude.ai (the public app) to draft a document is not creating a confidential communication — they are sending information through a commercial service provider. The confidentiality element of privilege is broken at the point of AI use.

What this means in practice:

• A client who uses ChatGPT to draft a narrative they send to you may have a discoverable document.
• If that document is in your file, opposing counsel can seek it in discovery.
• Your failure to flag this risk to the client before litigation may create a malpractice exposure.

The ruling applies specifically to consumer-grade AI — public apps used without enterprise data controls. Enterprise AI deployments inside law firms, with appropriate data processing agreements, are treated differently. But your clients are not using enterprise AI. They are using the free or consumer-paid versions of whatever tool they discovered.

The action: Add language to your engagement letter that prompts clients to disclose AI use and explains, briefly, why it matters. See the template below.

---

What Your Engagement Letter Needs to Say in 2026

This is not optional. AI disclosure requirements in engagement letters are becoming a professional responsibility baseline — and the privilege exposure above makes it a risk management requirement now, not when regulations catch up.

Here is a working template your firm can adapt:

"If you use any AI-assisted tool (including ChatGPT, Claude, Gemini, or similar platforms) to prepare documents, correspondence, factual summaries, or financial schedules submitted to this firm, you agree to disclose that use upon our request. This firm reserves the right to independently verify factual claims in AI-assisted materials before relying on them in any legal proceeding or filing. Additionally, you should be aware that documents prepared using publicly available AI tools may not be protected by attorney-client privilege and could be subject to discovery. We recommend discussing any AI-prepared materials with your attorney before submission."

Have your malpractice carrier review the specific language. Run it by your state bar's ethics hotline. Then add it to your standard engagement letter — not as a separate document, not as a checkbox clients skip, but as a paragraph in the body of the agreement they sign before the representation begins.

---

The 30-Day Workflow for AI-Era Client Documents

You do not need to build a detection lab. You need four process changes, each of which takes less time to implement than it will take you to clean up the first AI-hallucination problem you don't catch.

Week 1: Update your client intake form

Add one question: "Did you use any AI writing tool (such as ChatGPT or similar) to prepare any materials you are submitting today?"

This surfaces the issue at intake, sets the expectation that you are paying attention, and creates a record. That record matters if a question about the origin of documents arises later.

Week 1–2: Update your engagement letter

Add the AI disclosure clause above (adapted to your state). This is a one-time update that protects every matter going forward.

Week 2–3: Establish a one-attorney verification protocol

For any document flagged as potentially AI-generated — through client disclosure, your own review, or detection tool output — assign one attorney to run a verification step before the content is incorporated into firm work product.

Verification does not mean checking every sentence. It means:

• Ask the client to confirm the three or four most specific facts in the document (dates, amounts, names)
• Cross-reference any figures against supporting documents the client can provide
• Flag and resolve any inconsistencies before using the document

Document the verification in your matter notes.

Week 3–4: Track flagged documents

Keep a simple log (a spreadsheet column in your matter management system is enough): which matters have produced AI-flagged client documents, and what practice area. After 90 days, review the log. You will know whether this is a low-frequency edge case for your firm or a systematic intake pattern that requires a more structured response.

This is the AI disclosure tracking workflow that law firms building stronger AI governance policies for 2026 are implementing now. Start simple. Improve it based on what you actually see.

---

The Market Pressure Your Firm Is Already Feeling

The Cordell study is family law, but the client AI pressure pattern documented across practice areas is consistent: clients are arriving more prepared, more opinionated, and more likely to have used AI to draft their own version of what they want before they ever sit down with an attorney.

That changes the intake dynamic. It does not make your firm's expertise less valuable — if anything, the gaps and errors in what clients produce with AI are part of what you are being hired to identify and fix. But it does require that you stop assuming documents clients give you have been carefully prepared by the clients themselves.

Eighty percent means it is not the exception. It is the norm.

The workflow above takes a managing partner about 30 minutes to implement in policy form. The actual implementation — updating intake software, revising the engagement letter template, briefing the team — takes an afternoon. The cost of not doing it is one verification failure that becomes a malpractice exposure.

Build the workflow this week.

---

The Crossing Report tracks what's changing in professional services AI — not the hype, the specific operational decisions that affect firms like yours. Subscribe at crossing.one to get the weekly intelligence briefing every Monday morning.