Exclusive: Altorney Launches MARC — A GenAI System That Automates First‑Pass Document Review to Cut e‑Discovery Costs
Altorney has launched MARC, an AI‑powered first‑pass document review engine aimed at drastically reducing time, cost, and risk associated with legal e‑discovery workflows.
The product — now generally available after a pilot with corporate legal teams — is designed to make decisions about document relevance and privilege before material enters traditional review platforms such as Relativity or Everlaw.
Addressing a Core e‑Discovery Inefficiency
In typical discovery workflows, organizations load very large datasets into document review platforms, and then cull most of them as irrelevant. The inefficiency of this approach is both financial and operational:
- Review platforms often charge per document and per user, meaning large non‑responsive sets inflate platforms costs.
- Any document stored outside the corporate firewall also increases security and compliance risk.
Altorney co‑founder Shimmy Messing explains that existing workflows force legal teams to “import millions of documents only to remove most of them manually,” increasing both billables and risk exposure. MARC aims to flip that model by delivering first‑pass filtering inside a company’s own secure environment.
How MARC Works
MARC functions as an intermediate analytics layer placed between data collection and the traditional review platform. Key architectural points:
- It uses generative AI text analysis to evaluate documents for responsiveness, privilege, confidentiality, and other criteria before upload.
- The system is large‑language‑model (LLM) agnostic: firms can deploy MARC with Altorney’s pre‑configured Llama model locally, or switch to enterprise models hosted inside Azure, OpenAI, or other approved environments.
- All processing can occur behind the corporate firewall, reducing reliance on cloud AI uploads and mitigating external data exposure.
This local‑first design addresses one of the biggest concerns in legal AI adoption: confidentiality. As industry commentators note, lawyers are often reluctant to send sensitive case material to cloud services because of confidentiality and privilege risks.
Protocol‑Driven Analysis — No Prompt Engineering Needed
A central differentiator for MARC is its “protocol analysis” workflow rather than requiring users to craft AI prompts:
- Users upload background materials for a matter (e.g., complaints, subpoenas, pleadings).
- MARC generates a protocol document in Word format that captures parties, date ranges, key issues, and themes.
- Attorneys can edit the protocol directly in Word and re‑submit it to refine how the AI will assess documents.
- The finalized protocol guides all subsequent analysis of document sets.
This feature keeps attorneys working in familiar tools rather than forcing them to adapt to AI prompt syntax — a point legal tech analysts have repeatedly emphasized as essential for adoption.
Processing, Validation, and Iteration
Once the protocol is finalized, MARC can ingest data from multiple sources:
- Local file systems
- Microsoft Purview exports from Microsoft 365
- Live connections to saved searches in Relativity without copying data into MARC
MARC then analyzes sample documents to determine statistical accuracy before scaling results. If discrepancies arise between the AI’s classifications and attorney review, MARC can iterate the protocol, refining how it interprets relevance without disturbing already‑confirmed decisions.
Deep Analysis Beyond Relevance
MARC’s first‑pass review is not limited to responsiveness. The system can simultaneously handle multiple types of analysis for the same document at a single per‑document cost, including:
- Privilege Review: Identifying attorney‑client and work‑product protected documents, even commenting on potential waiver scenarios.
- Issue Coding: Tagging for case‑specific issues defined in the protocol.
- Confidentiality Classification: Detecting and categorizing trade secrets and other sensitive information.
- PII and PHI Detection: Flagging personally identifiable and protected health information with granular controls.
- Foreign Language Translation & Summarization: Allowing documents in other languages to be reviewed under the same protocol.
- Hot Document Identification: Highlighting priority documents for immediate review.
This breadth of analysis reflects broader trends in legal AI: courts and legal operations teams increasingly expect tools that can handle multiple legal criteria rather than simple keyword or predictive coding triggers.
Output Transparency and Reasoning
Crucially, MARC doesn’t just label documents — it provides explanations for its decisions, including textual snippets and confidence metrics. This “explainable AI” approach helps legal reviewers understand why a document was categorized a certain way, addressing a common criticism of opaque black‑box AI systems in e‑discovery.
Cost Savings and Predictability
During early pilot programs, Altorney says organizations saw:
- Up to 62% reduction in review platform costs
- Around 78% reduction in hosting costs
- An 80% decrease in the volume of documents transferred into hosted systems
- More than 85% faster cycle times compared to traditional review approaches
In one proof‑of‑concept with 30,000 documents, Altorney estimated a review budget of $2,500 — and delivered nearly exactly that figure — demonstrating a new level of predictability in AI‑powered legal pricing.
These levels of efficiency are in line with broader legal tech surveys showing that AI adoption generates measurable time savings and revenue growth for firms and legal departments.
Humans Still Essential, But Better Positioned
Altorney emphasizes that MARC is not designed to replace human legal judgment. Rather, it shifts attorneys’ focus from repetitive first‑pass review to quality control, strategy, and nuanced analysis. As former practice support leaders often point out, human oversight remains critical even with advanced AI workflows.
Market Expansion and Availability
Originally released to corporate legal departments only, MARC’s deployment expanded after litigation service providers (LSPs) requested installations within their environments. This demand, in turn, facilitated uptake by law firms and external legal teams.
Pricing Model
MARC’s pricing is based on document volume:
- A nominal per‑document fee for initial relevance assessment
- Additional per‑document fees for advanced analyses (privilege, PII/PHI tagging, etc.)
- The ability to rerun analyses without extra charge if case requirements change
This pricing aligns with industry shifts toward value‑based and predictable billing in legal technology.
Origins and Team
Founded by brothers Shimmy and Rachi Messing, Altorney began as a legal talent marketplace before pivoting to legal tech innovation.
MARC was developed in collaboration with Stephen Goldstein, a former global director of practice support at Squire Patton Boggs, and carries a name honoring the founders’ late father, Marc Messing — an attorney and educator known for his pursuit of truth.
What This Means for Legal AI and e‑Discovery
Altorney’s MARC enters a competitive field alongside other generative AI tools like Relativity’s aiR for Case Strategy, which focuses on fact extraction and narrative generation earlier in litigation workflows.
Although skepticism remains in some corners about AI reliability and legal risk, professionals increasingly recognize that securely deployed, explainable AI tools are becoming integral to efficient discovery. Embracing this trend — with strong human oversight — appears to be how many legal departments are bridging the gap between manual review and scalable technology.
