Answers
The questions partners ask before they trust an agent with the work.
Oversight, accountability, isolation, pricing and deployment — answered plainly, without asking you to book a call first.
Platform basics
- What is Vanguard Command OS?
- Vanguard Command OS is an AI-first operating system for law firms. It deploys 57 role-specific AI agents — intake and conflicts, litigation support, senior associate, billing, client communications, IT and more — each scoped to a role, restricted to an explicit tool allowlist, and bound to an oversight tier that decides what it may do without a human.
- Can Vanguard file for me?
- No agent ever clicks submit or pays a fee. Browser agents can read public records and pre-fill court or agency forms, then stop. On Enterprise, a prepared session opens in live view so the responsible attorney reviews it and clicks submit personally, and only after the document has passed the pre-filing check. Sites whose terms prohibit automation, such as NYSCEF and eFileTexas, are never visited.
- How is this different from a legal AI chatbot?
- A chatbot answers questions. Vanguard executes firm work: it runs intake and conflicts screens, triages discovery, prepares billing, drafts client updates, and files calendar deadlines through real integrations. Every one of those actions is scoped, logged, and either approved by a human or explicitly permitted to run alone.
- Do we need to replace our practice management system?
- No. Vanguard runs alongside your existing stack and acts through integrations — email, calendar, document systems, and e-signature — rather than requiring a migration. Firms typically start with one practice group and expand once the oversight record is trusted.
Oversight and accountability
- What are the three oversight tiers?
- Tier 1 actions are low-risk and run autonomously with a logged record. Tier 2 actions execute but are surfaced for review in the approvals queue. Tier 3 actions — anything client-facing, financial, or filed — cannot complete until a named human with the right role approves them. The tier is a property of the agent and the action, not a user preference.
- Who is accountable for AI output?
- A licensed attorney. AI output is decision-support, never advice. Anything delivered externally carries an attribution footer naming the AI assistance, the date, and the reviewing attorney, and the approval itself is recorded in the audit chain.
- Can we prove after the fact who approved something?
- Yes. Every action is hash-chained into an append-only audit trail, so altering a past entry breaks the chain. On qualifying plans the chain is periodically sealed with a signed root, which lets you demonstrate to a regulator, insurer, or opposing party that an approval record has not been edited since it was written.
- What happens if an agent gets something wrong?
- Tier 3 gating means a wrong output on a consequential action never leaves the firm without a human catching it. Errors are reportable from the observability console, the agent's tool allowlist and oversight tier can be tightened immediately, and the full action record — inputs, model, rationale, reviewer — is available for the post-mortem.
Security and data
- Is our data used to train AI models?
- No. Third-party foundation models are not trained on your firm's data. Prompts carry the minimum context a task requires, and retrieval is scoped to your firm's own indices.
- How is one firm's data isolated from another's?
- Row-level security is enforced at the database layer on every table, so isolation does not depend on application code remembering to filter. Documents sit in access-controlled storage, and cross-firm retrieval is blocked at the same layer.
- Can clients see their own matters?
- Yes, through the client portal: matter status, documents, messages, and e-signature, in English or Spanish. Clients see only what the firm publishes to them, governed by the same row-level policies.
- Do you hold any certifications?
- The trust center describes the controls that are live in the product today — tenancy isolation, access control, secret handling, and audit integrity. Certification status and any subprocessor list are provided directly during diligence rather than claimed on a marketing page.
Plans and pricing
- How is Vanguard priced?
- Four plans — Solo, Boutique, Firm, and Enterprise — priced per firm rather than per query, so agent usage does not become a variable your partners have to budget around. The pricing page carries the full per-tier feature matrix.
- What actually differs between the plans?
- Capability, not seat count alone: approvals workflow, agent studio configuration, workflow automation, signed audit roots, inbound email routing, and advanced integrations unlock as you move up. Every one of those gates is enforced on the server, so a plan boundary cannot be bypassed from the browser.
- Can we try it before committing?
- Yes. The public sandbox runs a scoped agent on synthetic input so you can see the voice and the Tier-3 gate without sharing firm data, and the readiness diagnostic returns a profile plus a recommended plan in a few minutes.
Deployment and adoption
- How long does deployment take?
- A first practice group typically runs in weeks, not quarters: connect email and calendar, set the oversight tiers your partners are comfortable with, and start with intake, conflicts, and deadline control before widening the roster.
- Does the platform work in Spanish?
- Fully. The staff application, the client portal, notifications, and outbound messages all run in English or Spanish, with legal terminology pinned to a maintained glossary so translations stay consistent across matters.
- What do partners have to do differently?
- Review a queue instead of drafting from zero. The partner surface is a pulse dashboard plus a pending-decisions list; the work arrives prepared, with the agent's rationale attached, and the partner's job is to approve, amend, or reject.
AI-first, partner-accountable
Still deciding?
Run a scoped agent on synthetic input in the public sandbox, or have a partner-level specialist walk your committee through the oversight record.
- 57 specialist agents
- 3 oversight tiers
- Hash-chained audit trail
- EN / ES parity