Written by: Wiktor Stefański, Head of People & Operations, Digital Colliers
Cooley's partnership with OpenAI puts ChatGPT inside the workflow for complex IPO work. The firm reported faster turnaround on disclosure documents and prospectus drafts. But the interesting signal is not the speed gain. It's what happened next. Clients taking companies public started asking which sections of their S-1 filings touched an LLM and which came entirely from human associates. That question is not going away.
Why Pre-IPO Clients Want AI Disclosure Logs
A company preparing to go public builds an audit trail for everything. The SEC can request work papers, email threads, and meeting notes years after the IPO closes. If your counsel drafted parts of a registration statement using an AI tool, that fact becomes part of the record. The client wants to know before the regulator asks.
There's also the billing question. ABA Formal Opinion 512 states lawyers cannot bill hours that AI actually saved. If your firm used ChatGPT to cut a 12-hour drafting task down to 8 hours, you owe the client transparency about that compression. Pre-IPO companies are sophisticated buyers of legal services. They expect disclosure.
The third reason is reputational risk. Court cases involving AI-fabricated citations rose from 87 to over 1,300 in eleven months during 2024. That's not a small sample. That's a pattern. A company about to hit public markets does not want to discover post-IPO that its registration statement contains an AI hallucination some associate missed in review.
What Good Looks Like: Per-Matter Tagging and Audit Trails
The winning pattern I keep seeing from firms building this capability has three parts. First, matter-level tagging at the point of use. When a lawyer invokes an AI tool on a document, the system logs which matter code that document belongs to. No retroactive reconstruction. Tag it when it happens.
Second, a disclosure report the firm can generate on demand. The client's CFO or general counsel should be able to request a log showing which deliverables in the engagement touched AI and which did not. The log does not need to be paragraph-by-paragraph. Document-level granularity is enough for most audit purposes.
Third, human review documentation. If an associate used ChatGPT to draft a section, who reviewed the output and when? The audit trail should show that a senior lawyer or partner signed off on the AI-assisted work before it went to the client. That review step is what separates a defensible workflow from a liability waiting to happen.
How This Maps to Regulatory Guidance
The UK Solicitors Regulation Authority issued its AI guidance in November 2023. The core principle is supervision. You remain accountable for AI output the same way you remain accountable for work a paralegal produces. If you would not send a junior associate's draft to a client without review, the same rule applies to LLM output.
The ABA took a similar stance. Its guidance makes clear that using AI does not change your ethical duties around competence, confidentiality, and billing. If you cannot explain which parts of a deliverable came from an AI tool, you probably should not be using that tool on client work yet.
For firms working across borders, the EU AI Act adds a compliance layer. Transparency obligations under Article 50 apply from August 2026. High-risk violations carry fines up to €15M or 3% of global turnover. An IPO represents exactly the kind of high-stakes use case regulators care about. Your client disclosure workflow is not just good practice. It's increasingly a regulatory requirement.
The operations lift here is straightforward. You need a system that tags AI use at the matter level and generates audit reports. You need a partner-level process that verifies human review happened before delivery. And you need intake conversations with pre-IPO clients that set expectations about what the disclosure log will and will not show. The firms building this now are not reacting to a crisis. They are giving themselves runway before the crisis arrives.