Common Questions.
Everything you need to know about deploying a Resident Legal Agent within your existing Microsoft infrastructure.
Security & Data Privacy
Does our data leave our Microsoft 365 tenant?
No. This is the Arivu core principle. Unlike SaaS AI providers who require you to upload files to their servers, Arivu is In-Tenant. We use your private instance of Azure OpenAI. Your documents, metadata, and queries stay within your firm’s security boundary.
Is our data used to train public AI models?
Absolutely not. Because we utilize Azure OpenAI Enterprise, Microsoft guarantees that no customer data is used to train the underlying foundation models (like GPT-4). Your institutional knowledge remains your private competitive advantage.
Accuracy & Hallucinations
How do you prevent AI “hallucinations”?
Arivu uses a technique called RAG (Retrieval-Augmented Generation). The AI is not allowed to answer from general knowledge. It is “grounded” in your documents. Every summary or answer provided by the Resident Agent includes direct hyperlinks to the source files in SharePoint/Outlook so you can verify the facts instantly.
Implementation & ROI
We have 20 years of “messy” data. Do we need to clean it up?
No. Arivu thrives on “messy” data. Our semantic indexing understands the meaning of documents regardless of their file name or folder structure. While a cleaner environment helps, the Resident Agent is designed to find the “needle in the haystack” across legacy folders and fragmented email chains.
What is the typical ROI for a 50-lawyer firm?
On average, firms see a reduction of 4-6 hours per week per associate on non-billable research and drafting tasks. For a 50-lawyer firm, this translates to reclaiming thousands of hours of high-value capacity annually, significantly improving realization rates and partner profitability.
Does this replace Microsoft Copilot?
No, it supercharges it. Copilot is a generalist; Arivu is a Legal Specialist. While Copilot helps you write an email, Arivu’s Resident Agent uses your Institutional Memory and CourtListener to ensure that email is legally sound and consistent with your firm’s 20-year history of precedents.
What about “Zero-Tagging”—how does the AI find things if we haven’t organized them?
Arivu uses Semantic Indexing. Traditional search looks for keywords. Arivu looks for concepts. If you search for “Pandemic Protection,” Arivu will find “Force Majeure” clauses even if the word “pandemic” never appears in the document. You don’t need to tag your data because Arivu “reads” and understands it for you.
Platform & Competitive Landscape
How is this different from Harvey or other AI “Vaults”?
Most “Vault” models require you to duplicate your data into their cloud. This creates a “sync lag” and a massive security risk. Arivu is a Resident Agent. We don’t have a vault; we have an engine that lives inside your vault (Microsoft 365). If you delete a file in SharePoint, it’s gone from Arivu instantly. No lag, no data residency issues.
Legora focuses on tabular views; can Arivu do that?
We go beyond the table. While tabular views are great for simple extraction, Arivu provides Matter Synthesis. We don’t just put your data in a grid; we reason across the documents to find inconsistencies, risks, and strategic opportunities that a spreadsheet can’t capture. We turn “Data” into “Advice.”
Arivu.Legal vs. The Field
| Critical Factor | Legacy AI Vaults (Harvey) | Tabular Tools (Legora) | Arivu Resident Agent |
| Data Residency | External Cloud. Data is “synced” or copied to their servers. | Third-Party Portal. Files must be uploaded to a dashboard. | Sovereign (In-Tenant). Data never leaves your M365 boundary. |
| User Experience | New Destination. Requires a separate web tab/app. | Extraction Tool. Focused on building spreadsheets. | Native Flow. Lives inside Word, Outlook, and Teams. |
| Knowledge Base | General Legal. Trained on broad law firm datasets. | Manual Tags. Requires you to “prep” or tag your files first. | Institutional Memory. Uses your private SharePoint context. |
| Implementation | Heavy. Weeks/months of “ingestion” and training. | Project-Based. Manual setup for every new deal/folder. | Agentic (14 Days). Zero-tagging automated discovery. |
| Search Logic | Keyword/Chat. “Tell me about this document.” | Data Extraction. “Put these clauses into a table.” | Matter Synthesis. “Reason across this entire matter folder.” |
| Security/IT | High Friction. Requires new data-export audits. | Medium Friction. Creates a secondary data silo. | Zero Friction. Inherits your existing M365/Entra ID. |
Security, Competitive Landscape
Why does “In-Tenant” matter if everyone uses encryption?
Encryption only protects data in transit. Once your data is in a “Vault” (like Harvey), you lose control over who at that company can access it, how it’s audited, and where it’s stored. With Arivu’s Resident Agent, your data is governed by your own Microsoft Purview and Entra ID policies. You are the only one with the keys.
Can Arivu do the tabular work that Legora is known for?
Yes, but we view a table as a “starting point,” not the finish line. While Legora extracts data, Arivu synthesizes it. We don’t just tell you what the clause says; we tell you how it conflicts with other documents in the matter and suggest a redline based on your firm’s historical preference.
Is Arivu just a “wrapper” for ChatGPT?
No. Arivu is a specialized Orchestration Layer. We use a stateless instance of Azure OpenAI for “brain power,” but the Resident Agent is what provides the legal logic, the CourtListener integration, and the secure retrieval of your private SharePoint precedents.
Still have questions?
Our lead architects are happy to jump on a technical call with your IT or Security team.