The end of the handoff

In our ongoing series on the Future of eDiscovery, we have tracked evolving target operating models, the rise of trusted AI orchestration, the redrawing of the Electronic Discovery Reference Model (EDRM), and legal investigations that now move at the speed of a prompt. Running beneath all of it is a quieter structural shift: the handoffs that have organized eDiscovery for two decades are collapsing. Relativity's August 12 announcement of Relativity claiR²—a conversational AI experience built for lawyers, grounded in the Relativity aiR system of record is the clearest signal yet of this.

claiR is available through Relativity’s Advanced Access program, and KPMG is the first professional services advisor in that program—putting the capability in front of client legal teams while it is still being shaped. General availability is planned for early 2027.

Two decades of handoffs

Discovery has never really been a workflow. It has been a relay. Counsel frames a question. A legal data team translates it into search syntax, analytics, or a coding protocol. A review team applies judgment at scale. Someone assembles the output into a memo, a chronology, or a set of hot documents. Then counsel reads it and asks the next question, and the relay runs again.

Every handoff in that chain added three costs: latency measured in days, translation loss between the person with the question and the person with the query, and expense that scaled with volume rather than with insight. The industry has spent years optimizing each leg of the relay—better culling, better technology-assisted review, better project management—without questioning the relay itself.

What claiR actually changes

claiR removes the translation layer. A lawyer asks a question of the matter in plain language and gets an answer back with citations attached—no syntax, no ticket, no intermediary. Relativity describes the scope as the complete Relativity aiR matter record: the documents, their metadata, and the relationships among them, rather than a curated extract, with answers that are "grounded, cited, auditable."¹

The architectural choice matters as much as the experience. The prevailing pattern in legal AI has been to move a subset of data into a separate environment for the AI agent to work on. claiR inverts that: the agent operates where the evidence already sits. As Relativity President Chris Brown put it, "Nothing is exported or sampled, the answer comes from the record itself, right where the data lives."¹ Because the corpus never moves, the permissions, security posture, and audit trail of the system of record travel with every answer.

Relativity’s CEO Phil Saunders framed the shift in organizational terms: with claiR, "for the first time, the full richness of a matter's data is directly in lawyers' hands, rather than something they wait on technical legal data teams to surface."¹

The experience is the strategy

What makes claiR more than a chat window is what sits around the conversation. Based on pre-release materials and Advanced Access demonstrations, three design choices stand out.³

Agent Skills, not prompts. Instead of a blank prompt box, claiR surfaces a library of named legal tasks—drafting, summarizing, comparing documents, creating timelines, outlining proof—where the name is specific to the task. Teams can add a skill to the library or extend an existing skill with their own requirements and expertise for future matters.

Canvases, not transcripts. Rather than forcing everything into a scrolling chat, interactive views are exposed intuitively: a document grid where you can add AI-powered columns on the fly, a viewer to see outputs grounded at the document level, and a drafting surface where work product takes shape beside the record.

Shared work, not sessions. Work is saved and can be shared with the case team rather than lost at the end of a session, so a line of inquiry, a skill, a draft can be collaborated on or used to drive future work.

Individually these are features. Together they describe an operating model where questioning, analysis, and work product happen in one governed place instead of across three teams and four tools.

From eDiscovery to legal data intelligence

The deeper consequence is a reframing of the discipline. eDiscovery has been accounted for as a cost center—a defensible obligation executed as cheaply as possible, then archived. What claiR exposes is that the archive was always the asset. A mature Relativity aiR environment holds years of processed, enriched, issue-coded, privilege-reviewed data across an organization's most consequential matters. Once that record is interrogable in plain language, a closed matter becomes a research corpus, and revisiting a prior production for newly relevant facts stops being too expensive to justify on a hunch.

This is the trajectory the EDRM 2.0 model anticipated when it recast analysis as continuous connective tissue spanning the lifecycle rather than a discrete phase, and set information governance beneath the entire model.⁴ Reference model and tooling are converging on the same conclusion: discovery is not a linear project that ends. It is an intelligence capability that compounds.

Governance gets harder, not easier

A plain-language experience does not reduce the governance burden—it redistributes it. When only specialists could query the corpus, methodology was controlled by scarcity. When any authorized lawyer can, control has to be designed: permissioning set deliberately rather than by default, citation trails treated as a hard requirement, practitioner validation built into the live workflow rather than the post-mortem, and a clear-eyed view of how prompts and AI-generated drafts interact with privilege.

As Brown put it, claiR "doesn't replace legal judgment, it gives lawyers more complete access to the facts and insights that judgment depends on."¹ The judgment—and the accountability for it—stays exactly where it always was.

The technology will be table stakes. The enablement won't.

When claiR arrives, the technology stops differentiating anyone the moment users have access to it. What separates programs is enablement: which matters you point it at, who is permitted to ask, whether the skills were validated or improvised, and whether the citation trail was built into the workflow or reconstructed after a court asked.

Those are operating-model decisions, made early, with consequences across every matter that follows—which is the case for a partner rather than a toggle. KPMG has been in the room while claiR was shaped, and brings the enablement and governance that turn a conversational experience into work product that stands up. To discuss what this means for your program, explore the KPMG Forensic Technology and eDiscovery capabilities and our KPMG and Relativity alliance.

Sources

¹ Relativity, "Relativity Launches claiR, Giving Lawyers Conversational Command of the Full Depth of Their Legal Data," PR Newswire, August 12, 2026.

² Relativity, "Relativity claiR | Conversational AI for Lawyers," relativity.com, accessed August 13, 2026; "Relativity Develops claiR Conversational AI System," Artificial Lawyer, August 12, 2026.

³Relativity claiR product materials and Advanced Access demonstrations, August 2026. Pre-release capabilities are subject to change.

⁴ EDRM, "EDRM Invites Global Community to Shape EDRM 2.0 via Public Comment," June 30, 2026; Doug Austin, "EDRM 2.0 Model Released by EDRM for Public Comment," eDiscovery Today, June 30, 2026.