What MinuteBox actually announced
According to LawSites, MinuteBox today launched a new AI agent that turns its corporate records database into a platform for performing legal work. The feature is available immediately to existing customers as an optional, per-seat addition, priced so that paralegals and law clerks can use it alongside lawyers. It is the first of three planned AI releases over the coming months.
The agent draws on three existing components: a structured database of corporate entity information, a compliance engine that validates data against statutory and company-specific rules, and direct integrations with government business registries. Chief executive Daniel Levine described the outcome as a “closed loop”: pull a company's information from a registry, compare it with the firm's records, record a change, generate the required documents, and file them, without leaving the platform.
The interface changes too. The traditional dashboard of entities becomes a chat window — what Levine called a “single pane of glass” into the corporate record ecosystem. He predicts that within six months, 95% of users will use the agent 95% of the time.
How it differs from the previous product and from competitors
Before this release, MinuteBox was, in Levine's words, a “system of record”. The workflow began with a dashboard, and document automation used templates in a no-code editor. The agent changes the starting point and the scope: natural-language requests now trigger structured answers from the corporate record, registry searches, precedent-based drafting and record updates. Each response carries source citations, and tabular extractions carry confidence scores.
The more interesting difference is architectural. Many legal AI tools bolt a chat interface onto documents; MinuteBox is bolting one onto governed entity data and live registry integrations, with an approval step before incurring registry fees. The demonstration showed the agent searching the Ontario registry for Pizza Pizza Limited, listing similar names, identifying the active corporation and pausing for approval before ordering a report. Levine noted that most registries cannot be searched by business name alone, so the agent “sprawls out” across different registries to locate the entity.
The following table summarises the shift described in the source. No independent benchmark figures were provided, so none are reproduced here.
| Dimension | Previous MinuteBox | MinuteBox AI Agent |
|---|---|---|
| Starting point | Entity dashboard | Chat window |
| Data retrieval | Structured records | Plain-language queries with citations |
| Registry work | Separate registry services process | Prompt-driven search and filing, with approval before charges |
| Drafting | Templates in a no-code editor | Firm Word precedents matched to entity data |
| Sharing | Standard document sharing | Each conversation becomes a shareable “AI room” |
| Pricing model | Existing subscription | Optional, per seat |
Three concrete uses, with realistic examples
1. Corporate due diligence for a firm. A law clerk receives a name that may refer to several entities across jurisdictions. Instead of a chain passing from partner to junior lawyer to search clerk, the clerk asks the agent, which searches registries, lists matches and pauses before charging the filing account. The clerk selects the Ontario entity, orders the profile report and creates the record. Bernstein contrasted this with a manual workflow that might involve three or four people; the agent completes it almost instantly. For teams exploring where agents fit, our guide to AI agents for lawyers sets out the governance questions to settle first.
2. Routine entity maintenance for an in-house team. A company secretary needs to appoint a director and issue shares. The agent flags two issues in the demonstration: more than one person shares the name given, and the company has no share class called “common”. These are exactly the ambiguities that cause defective filings. The agent then drafts director consents from the firm's own Word precedent, files them in the conversation and produces a table of facts used.
3. Client-facing deliverables for a small firm. From a share purchase agreement, the agent builds a branded “visual dossier” showing purchase price, closing date, escrow holdback and other key terms, customised with the firm's logo, fonts and colours. The conversation can be shared as an “AI room” with a PIN, expiry date, verified email and watermarks, with downloads off while recipients can still upload documents. That is a credible client experience improvement, though it should be tested against your engagement terms and confidentiality obligations.
Who and what this affects
The immediate impact falls on corporate legal teams, paralegals and law clerks in firms that already run entity management. If the agent works as demonstrated, the labour arbitrage is not in drafting alone but in the coordination overhead: the handoffs between people who search registries, retrieve records and draft routine documents. That is a structural change to how corporate work is staffed, not merely a faster tool.
Second, it affects competitive positioning. Entity management vendors have generally competed on data coverage and registry access. MinuteBox is now competing on workflow, and the tabular review and dossier features encroach on document-review and contract-analysis products. The source describes a review of six commercial contracts across 11 columns with each value linked to its source and tagged with a confidence percentage — capabilities that overlap with established legal AI categories.
Third, it affects buyers' evaluation criteria. When a system of record becomes a system of work, the questions change from “how complete is the database?” to “how auditable is each action, and who approves it?” That has implications for professional indemnity, client consent and internal supervision. Our analysis of the legal client journey is relevant here: trust is won at the points where the client can see what was done and why.
What to watch: confidentiality, accuracy, IP and limits
Confidentiality and privilege. The “AI room” model is thoughtful — watermarks, expiry dates, PIN access, downloads disabled — but sharing a conversation with a client is still a disclosure decision. Firms should confirm how the room interacts with their retention policies and whether chat transcripts form part of the client file.
Accuracy and supervision. Citations and confidence scores are the right ingredients, but they are not validation. The agent's own interruption about a likely duplicate entity is encouraging because it surfaces ambiguity rather than resolving it silently. Still, a confidence percentage is a prompt for review, not a substitute for it. The same discipline applies as with any retrieval system; our note on verifying AI-generated citations remains a sensible default.
Intellectual property. The agent drafts from a firm's own Word precedents and firms can add plain-language instructions to those precedents. Who owns and maintains those instruction-laden templates, and what happens if a firm leaves the platform, are contract questions worth asking before adoption.
Limits of professional use. Registry filing is jurisdiction-specific and error-prone. The approval pause before charges is good design, but approval is not legal review. The source gives no indication of which registries are covered beyond the Ontario demonstration, nor any accuracy or performance benchmarks, so claims should be tested on your own entities before relying on them.
How and when to try it
The agent is available now to MinuteBox customers as an optional feature, priced per seat. Existing customers can enable it and begin with low-risk tasks: entity queries with @ mentions, exporting director and officer tables, and tabular review of a small contract set. Integrated platforms named in the source include iManage, Clio Manage, OneDrive, Google Drive, Dropbox and Outlook, so documents can be brought in by drag-and-drop, email or an existing integration.
A sensible pilot lasts two to four weeks and covers three tests: a registry search with a deliberate near-duplicate name, a precedent-based drafting task on a real entity, and a client-facing dossier shared through a room with downloads disabled. Track where the agent pauses for approval and where it flags ambiguity, because those are the moments that define whether the workflow is safe. Two further AI releases are planned over the coming months, so it is worth keeping the pilot lightweight enough to re-evaluate.
Sources
Frequently asked questions
Is the MinuteBox AI Agent included in existing subscriptions?
No. According to LawSites, the agent is available to MinuteBox customers as an optional feature priced per seat. The company says it is designed and priced to be used by paralegals and law clerks as well as lawyers, which suggests a broader rollout than a partner-only tool. Existing customers should confirm seat pricing directly with MinuteBox.
Does the agent file documents with government registries automatically?
It can file, but not without approval. In the demonstration, the agent paused before charging the firm's filing account for an Ontario corporate profile report. The broader design connects registry searches and filings to the firm's records, so the user approves cost-incurring actions while the agent handles the search, comparison and document generation.
What happens to client data shared in an AI room?
The source describes rooms with watermarks, expiration dates, PIN access, verified email and download controls, and recipients can upload documents while downloads remain off. Those controls reduce exposure but do not settle retention, privilege or engagement-term questions. Firms should confirm how transcripts are stored and whether the room forms part of the client file before sharing.
Want to apply this to your own work or your team? Tell me where you stand, and I will reply personally.
Get in touch
