Firm: [Firm name] Attorney: [Name, Bar No.] Date: [Date]
I use Advisor Prep Hero, a local-first AI workspace for professional client work. Conversations and outputs are saved as plain text files in a folder I control on my own machine. Advisor Prep Hero does not store work product on its own servers.
Local model path (Ollama): When I use a local AI model such as Ollama, my prompts never leave my machine. The AI runs on local hardware. No data reaches any external server. This is the path I use for matters where the client has requested no external transmission, or where my own professional judgment warrants it.
Cloud API key path: When I use a cloud provider (Anthropic, OpenAI, or Google), my prompts go directly from my machine to that provider under my personal API key. Advisor Prep Hero's servers are not in the data path. Client data is processed under that provider's API data-processing terms, not under any consumer product terms.
Each matter has its own folder. When I open an AI session, it is scoped to that matter's folder only. The AI cannot see files from other matters. I do not use shared AI chat histories across matters. There is no cross-matter contamination.
Advisor Prep Hero maintains an append-only log of every AI action: timestamp, model, files included, and output produced. The log is per-machine and per-matter. I can produce a complete AI activity record for any matter on request. This log is retained as part of the matter file.
Local model path: Because no data leaves my machine, there is no transmission to a third party. The privilege and confidentiality analysis under the local path is straightforward: the communication never left my possession.
Cloud API key path: Prompts reach the cloud provider's API under my account. This is a third-party transmission. The current leading analysis holds that using a BYOK ("bring your own key") tool that routes directly to the provider does not waive privilege, provided the attorney exercises reasonable care and the provider's terms include appropriate data-handling commitments. The S.D.N.Y. in United States v. Heppner (Feb. 17, 2026) emphasized the importance of attorney control over the data path when evaluating privilege claims in AI-assisted work. The local path eliminates this analysis entirely.
This section is informational, not legal advice. Verify privilege and confidentiality implications with your own bar counsel.
The most commonly cited AI tool in legal practice right now is Clio Duo, the built-in AI layer for Clio practice management software. Clio Duo is a cloud SaaS tool with a contractual no-training policy, and it genuinely integrates tightly with matter management, billing, and deadlines inside the Clio platform. Clio also holds SOC 2 certification. For a carrier reviewing whether the attorney uses a credentialed cloud vendor, Clio Duo clears that bar in a way Advisor Prep Hero does not today.
The architectural difference that matters for this review: Clio Duo routes client matter content through Clio's servers. It is a third-party vendor in the data path. Under the Heppner analysis, transmission to a cloud vendor is the third-party contact that creates a privilege risk, regardless of the vendor's privacy commitments. Advisor Prep Hero with a local model has no such transmission: the AI runs on the attorney's machine and no content reaches any external server. With a cloud API key, Advisor Prep Hero routes prompts directly from the attorney's machine to the chosen provider under the attorney's own account, with no Advisor Prep Hero server in the path.
Advisor Prep Hero does not offer the practice-management integration, billing context, or SOC 2 credential that Clio Duo provides. Attorneys who need those features and whose confidentiality bar is met by a contractual no-training policy should use Clio Duo for that function. This document addresses the specific work for which the attorney chose a different architecture, a workspace where the data path is architecturally controlled, not contractually managed.
This document describes my AI data-handling architecture. The key facts for carrier review: (1) client data is never stored by Advisor Prep Hero; (2) I control the entire data path; (3) AI outputs are reviewed and approved by me before any use; (4) all AI actions are logged at the matter level; and (5) a fully local path exists for matters requiring zero external transmission. I am happy to provide technical architecture documentation or answer questions on request.