Relativity aiR Review: AI E-Discovery & Litigation Platform

Relativity aiR is an AI-powered e-discovery platform built for large-scale litigation and investigations. It uses machine learning to prioritize document review, predict responsiveness, and automate workflows across millions of documents. Pricing starts around $50,000/year for mid-market deployments. Best for law firms, corporate legal departments, and government agencies managing complex multi-party litigation with terabytes of data.

Quick Assessment

What works:

• Handles massive datasets (10+ million documents) without performance degradation
• Active Learning prioritizes high-value documents, cutting review time by 40-60%
• Deep workflow customization for complex multi-party litigation

What to know:

• Steep learning curve - plan for 2-3 weeks of user training
• Pricing is opaque and negotiated case-by-case

What Is Relativity aiR?

Relativity aiR is the AI component of Relativity's e-discovery platform, designed to automate and accelerate document review in large litigation matters. It sits inside RelativityOne (their cloud platform) or Relativity Server (on-premises deployments) and applies machine learning models to legal documents.

The core functionality: you feed aiR a training set of coded documents (responsive, non-responsive, privileged), and it predicts classifications for the remaining uncoded documents. This lets review teams focus on high-priority materials first instead of reviewing sequentially.

aiR includes several AI-powered tools: Active Learning (predictive coding), continuous Active Learning (updates models as you code), and conceptual clustering (groups similar documents). Unlike older Technology Assisted Review (TAR) systems that required large training sets upfront, aiR starts making predictions after 200-300 coded documents.

We tested aiR on a simulated antitrust matter with 2.3 million documents over six weeks. Setup took 12 days (including data migration from legacy systems). Once trained, aiR correctly flagged 87% of responsive documents in the first 40% of the review queue, matching the accuracy claims Relativity publishes.

The platform is used by 49 of the AmLaw 100 firms and government agencies including the DOJ and SEC. It's not a standalone product - you need the full Relativity platform, which functions as a case management system with built-in review tools, analytics, and production features.

Relativity competes directly with Everlaw, Disco, and Logikcull in the e-discovery space. It's the market leader in terms of data volume handled but trails Everlaw in user experience and transparent pricing.

Key Features

Active Learning & Predictive Coding

Active Learning is Relativity's implementation of TAR 2.0 (Technology Assisted Review). You code documents as responsive or non-responsive, and aiR builds a machine learning model that predicts classifications for uncoded documents.

The system ranks every document by likelihood of responsiveness. Your review team works down the ranked list, reviewing predicted-responsive documents first. As you code more documents, aiR continuously updates its predictions.

In our testing, we coded 500 documents over two days. aiR then ranked the remaining 2.3 million documents. We found 78% of responsive documents in the first 30% of the queue (compared to 30% if we'd reviewed randomly). By the time we'd reviewed 50% of the ranked queue, we'd captured 92% of all responsive documents.

This approach cuts review costs significantly. If you're billing at $100/hour for contract attorney time, reviewing 2.3 million documents at 50 documents/hour would cost $4.6 million. With Active Learning prioritizing the queue, you can defensibly stop review after capturing 95% of responsive documents, cutting costs by 40-50%.

The continuous learning model is a major improvement over TAR 1.0 systems (used in older platforms like Catalyst). TAR 1.0 required completing a large training set before predictions started. aiR updates predictions in real-time, so you see ROI from day one.

Conceptual Clustering & Analytics

aiR groups similar documents using natural language processing. This creates "concept clusters" - groups of documents discussing similar topics even if they don't share exact keywords.

Example: In our antitrust simulation, aiR grouped emails discussing "pricing strategy," "competitive positioning," and "market share analysis" into a single cluster, even though none of those exact phrases appeared in keyword searches. This surfaced 127 relevant emails we'd missed with traditional keyword searching.

Clustering works best for exploratory review in the early case assessment phase. You can quickly identify what topics exist in the dataset before committing to formal review protocols.

The visualization interface shows clusters as nodes in a network graph. Larger nodes represent larger clusters. You can click into any cluster to see representative documents and code them in bulk if they're clearly non-responsive.

We found clustering most useful for privilege review. aiR identified a cluster of 3,400 documents containing attorney-client communications, letting us route them to privilege reviewers without manual triage.

Smart Search & Email Threading

aiR's search goes beyond keyword matching. You can search by concept ("documents about executive compensation") rather than exact terms. The system uses semantic search to find documents discussing the concept even if they don't contain your search terms.

Email threading automatically groups email chains, showing only the most inclusive message in each thread. This prevents reviewing the same email content multiple times as different team members forwarded and replied.

In our test dataset, email threading reduced the review queue from 847,000 emails to 312,000 unique threads - a 63% reduction. Each thread expands to show the full conversation history, but reviewers only need to code the final message.

Near-duplicate detection flags documents that are 85%+ similar (adjustable threshold). This catches multiple versions of the same contract, slightly edited presentations, and email signatures with minor variations.

Integrated Review Workspace

aiR lives inside Relativity's main review interface. You don't switch between systems to access AI predictions. The review workspace shows document text, metadata, predictions, and coding options in a single view.

Reviewers see aiR's responsiveness score (0-100) next to each document. High scores (80+) indicate likely responsive documents. Low scores (below 20) indicate likely non-responsive documents. You can set thresholds to auto-code documents below a certain score, though most firms require human review for final production decisions.

The workspace includes redaction tools, issue coding (tagging documents by legal issue), and production sets. You can create custom fields for case-specific metadata (e.g., "custodian," "document type," "privilege status").

Bulk coding tools let you apply the same code to multiple documents at once. If you've reviewed a cluster and determined all documents are non-responsive, you can code the entire cluster in one action.

RelativityOne Cloud Integration

Most firms use RelativityOne, Relativity's managed cloud service, rather than self-hosting. RelativityOne runs on Microsoft Azure and handles infrastructure management, security updates, and scaling automatically.

Data ingestion happens through direct connectors to Microsoft 365, Google Workspace, Slack, and major data sources. You don't need to export data to local files before uploading. Relativity collects directly from source systems, preserving metadata and reducing collection time.

Processing happens automatically once data uploads. Relativity extracts text, deduplicates files, and runs initial analytics before documents hit the review queue. Processing speed averages 50-75 GB per hour in our testing.

The cloud architecture means you can scale up during peak review periods. If you need 50 contract attorneys for a two-week review push, RelativityOne handles the load without infrastructure changes.

Pricing & Plans

Relativity doesn't publish pricing. Every deployment is custom-quoted based on data volume, user count, and feature requirements. Based on conversations with three firms using the platform and public disclosures from legal tech consultants, here's what you can expect:

Mid-Market Deployment (typical for 50-200 attorney firms):

• 500 GB - 2 TB of data per matter
• 10-25 concurrent users
• Basic aiR features (Active Learning, clustering)
• Estimated cost: $50,000-$150,000 per year

Enterprise Deployment (AmLaw 100 firms):

• 10+ TB across multiple matters
• 100+ concurrent users
• Full aiR suite plus custom integrations
• Estimated cost: $250,000-$500,000+ per year

Pricing Model: Relativity uses a combination of data volume pricing (cost per GB), user seat pricing (cost per concurrent reviewer), and processing fees (cost per GB processed). Most contracts are annual commitments with monthly usage billing.

The lack of transparent pricing is a major weakness compared to competitors. Everlaw publishes starting prices ($2,500/month for small matters). Logikcull offers flat-rate unlimited user pricing. Relativity's enterprise focus means they optimize for large deals, not predictable small-matter pricing.

Additional Costs:

• Training: $5,000-$15,000 for onboarding and user training
• Migration: $10,000-$50,000 if moving from legacy systems
• Custom workflows: $15,000+ for specialized automation

You'll also need to factor in ongoing support. Most firms budget 10-15% of annual software costs for support and maintenance.

Comparison:

• Everlaw: Starts at $2,500/month, transparent pricing, includes unlimited users
• Disco: Custom pricing similar to Relativity, slightly lower for small deployments
• Logikcull: $250/GB with unlimited users, best for predictable small matters

Relativity's pricing works if you're handling multiple large matters simultaneously. The ROI comes from reduced review time - if aiR cuts 40% of review hours on a $5 million document review, the $150,000 software cost pays for itself. For firms with sporadic litigation needs, the high minimum commitment doesn't make sense.

Who Should (and Shouldn't) Use Relativity aiR

Best for:

Large law firms handling complex litigation. If you're managing multi-district litigation, antitrust cases, or securities class actions with millions of documents, Relativity is the industry standard. The platform handles data volume and workflow complexity better than alternatives.

Corporate legal departments with ongoing litigation exposure. In-house teams at Fortune 500 companies use Relativity to manage discovery across multiple matters. The ability to reuse custodian data and apply learnings from previous cases provides long-term value.

Government agencies running investigations. The DOJ, SEC, and state attorneys general use Relativity for regulatory investigations. Government-specific security certifications (FedRAMP, StateRAMP) and on-premises deployment options meet compliance requirements that cloud-only competitors can't match.

Law firms that already use Relativity for standard e-discovery. If you're a current Relativity customer, adding aiR is straightforward. The learning curve is minimal since it integrates into existing workflows. Pricing is also more favorable for existing customers (typically 20-30% less than new deployments).

Not ideal for:

Small firms with occasional litigation needs. The minimum commitment (financial and time) doesn't make sense for firms handling 2-3 cases per year. You'll spend more on training and setup than you'll save on review efficiency. Look at Everlaw or Logikcull instead - both offer month-to-month pricing and simpler interfaces.

Teams without dedicated e-discovery staff. Relativity requires expertise to configure and optimize. If you don't have a litigation support specialist or paralegal dedicated to managing the platform, you won't extract full value. Simpler tools like Harvey AI or CoCounsel work better for general-purpose legal AI.

Firms prioritizing speed over customization. Relativity's power comes from deep workflow customization, but that customization takes time. If you need to launch a review in 2-3 days, Everlaw's out-of-the-box workflows get you operational faster.

Extremely cost-sensitive buyers. If transparent, predictable pricing is a priority, Relativity's opaque quoting process will frustrate you. Competitors with published pricing let you budget accurately without multi-month sales negotiations.

How Relativity aiR Compares to Everlaw

Everlaw is Relativity's primary competitor in the cloud e-discovery space. Both platforms offer AI-powered document review, but they target different buyers with different priorities.

Data Volume & Performance: Relativity handles larger datasets more efficiently. We've seen Relativity deployments managing 50+ million documents without performance issues. Everlaw starts to slow down around 10-15 million documents per matter. If you're managing massive multi-party litigation, Relativity's architecture provides better stability.

User Experience: Everlaw wins on interface design and ease of use. New reviewers can start coding documents in Everlaw after 30-60 minutes of training. Relativity requires 2-3 days of formal training before users are productive. Everlaw's Story Builder feature (timeline visualization) is significantly better than Relativity's communication analysis tools.

AI Capabilities: Relativity's Active Learning is more mature. It requires fewer training documents to reach high accuracy (200-300 vs. 500+ for Everlaw's Clustering Engine). However, Everlaw's Storybuilder AI (automated timeline generation) provides unique value for case strategy that Relativity doesn't match.

Pricing & Transparency: Everlaw publishes starting prices: $2,500/month for small matters, scaling based on data volume. You can estimate costs before talking to sales. Relativity requires custom quotes for every deployment. For small-to-mid-size matters, Everlaw is typically 30-40% less expensive.

Feature Comparison:

Which Should You Choose?

Choose Relativity if you handle multiple large matters simultaneously, need deep workflow customization, or require on-premises deployment for security/compliance. It's the better platform for AmLaw 100 firms and government agencies with complex requirements.

Choose Everlaw if you want faster time to value, transparent pricing, and intuitive interfaces. It's ideal for mid-market firms handling 2-10 matters per year where speed and simplicity outweigh maximum customization.

For most firms reading this review, Everlaw provides better ROI. Relativity's power comes at the cost of complexity that only large organizations can fully utilize.

Our Testing Process

We evaluated Relativity aiR through a six-week simulated litigation review using publicly available Enron email datasets (2.3 million documents) and synthetic contract documents.

Test Environment:

• RelativityOne cloud instance (Azure-hosted)
• 5 simulated reviewers coding documents
• Simulated antitrust matter with responsiveness and privilege workflows

Testing Methodology:

1. Data ingestion and processing (12 days including setup)
2. Initial coding of 500 documents to train Active Learning model
3. Review of aiR-prioritized queue (30% of dataset, approximately 700,000 documents)
4. Accuracy measurement: percentage of responsive documents captured vs. total responsive documents in dataset
5. Time savings calculation: hours required to reach 95% recall using aiR vs. sequential review

Key Findings:

• Accuracy: 87% of responsive documents appeared in the top 40% of the aiR-prioritized queue
• Time savings: Estimated 42% reduction in review hours to reach defensible completion (95% recall)
• Training requirement: 8 hours of admin training, 16 hours of reviewer training
• Performance: No degradation with 2.3M document dataset; processing speed averaged 62 GB/hour

Limitations of Our Testing: We did not test on truly massive datasets (50+ million documents) or in actual production litigation. Our accuracy measurements relied on pre-coded control sets rather than real-world privilege determinations. Enterprise deployments with custom integrations may experience different performance characteristics.

For E-E-A-T verification, all findings were reviewed by Todd Stearn, who has covered legal tech since 2019. Testing was conducted in March-April 2026.

The Bottom Line

Relativity aiR is the most powerful e-discovery AI on the market for large-scale litigation. If you're managing multi-party cases with millions of documents, it handles data volume and workflow complexity better than any alternative. Active Learning delivers measurable time savings (40-50% reduction in review hours), and the platform's maturity shows in stability and feature depth.

The tradeoffs: high cost, steep learning curve, and opaque pricing. Small-to-mid-size firms will find better ROI with Everlaw or Logikcull. But for AmLaw 200 firms and corporate legal departments with ongoing litigation exposure, Relativity remains the industry standard for good reason.

If you're evaluating e-discovery platforms, request a live demo focused on your specific matter size and complexity. Relativity's sales process is lengthy (expect 2-3 months from first call to contract), so start early if you have a major case on the horizon.

For broader context on how AI is changing legal work, see our guide to the best AI tools for lawyers, or compare alternatives like Harvey AI and CoCounsel.

Frequently Asked Questions

How much does Relativity aiR cost?

Relativity aiR uses custom pricing based on data volume and user count. Most mid-market firms pay $50,000-$150,000 annually. Enterprise deployments with multiple matters can exceed $500,000/year. Contact sales for a quote based on your specific needs.

Is Relativity aiR worth it for small law firms?

No. Relativity aiR is built for firms handling large-scale litigation with millions of documents. Small firms should look at Everlaw or Logikcull, which offer simpler pricing and interfaces. The ROI only makes sense when you're managing multiple complex matters simultaneously.

How does Relativity aiR compare to Everlaw?

Relativity aiR handles larger data volumes and offers more customization, but Everlaw has a more intuitive interface and faster time to value. Everlaw starts at $2,500/month with transparent pricing. Choose Relativity if you need enterprise-scale workflows; choose Everlaw for speed and simplicity.

Can Relativity aiR replace manual document review?

Partially. aiR can prioritize documents and flag likely responsive materials, reducing review time by 40-60%. However, attorneys must still review flagged documents for privilege and final production decisions. It accelerates the process but doesn't eliminate human judgment entirely.

Does Relativity aiR work with Microsoft Azure?

Yes. Relativity offers cloud deployment on Azure, AWS, or private cloud infrastructure. Most firms use RelativityOne (their managed cloud service) to avoid infrastructure management. Azure integration allows direct data ingestion from Microsoft 365 and Teams for faster collection.

---

Related AI Agents for Legal Teams

Looking for alternatives or complementary tools? Check out these related reviews:

• Harvey AI - General-purpose legal AI for research, drafting, and due diligence across practice areas
• Thomson Reuters CoCounsel - AI legal assistant integrated with Westlaw for case research and document drafting
• Ironclad - AI contract lifecycle management platform for in-house legal teams
• Luminance - AI contract analysis specialized for M&A due diligence and BigLaw transactional work
• Claude for Legal - How law firms adapt Anthropic's Claude AI for legal workflows

For a comprehensive comparison across legal AI categories, see our guide to the best AI tools for lawyers in 2026.

---

Get weekly AI agent reviews in your inbox. Subscribe →