Next-Generation CaseBot™ Turns Case Data Into Instant, Defensible Insights

HaystackID Blog | August 23, 2026
Next-Generation CaseBot™ Turns Case Data Into Instant, Defensible Insights

Editor’s Note: Expert case analysis has always carried a five-figure price tag, and much of it never gets commissioned as a result. The newest generation of CaseBot™ changes that equation. At ILTACON 2026 in Nashville, HaystackID® will introduce the next evolution of CaseBot technology from eDiscovery AI®. The release extends CaseBot beyond conversational search by introducing agentic retrieval and Specialized Skill Agents. Rather than simply answering questions, CaseBot plans, retrieves, evaluates, and synthesizes evidence through purpose-built analytical workflows, transforming targeted case files into structured, cited case intelligence outputs in minutes. For cybersecurity, data privacy, regulatory compliance, and eDiscovery professionals, the release matters because it moves generative AI (GenAI) beyond question answering and into accurate fact narratives, chronologies, deposition kits, and record checks produced where the data already lives, under existing access controls and audit trails. Case Insight™ reads the entire document population to establish what a team has; CaseBot interrogates the targeted record to establish what it means. Watch next for persistent saved sessions and client-specific custom agents in the second half of the year as the model extends toward firm-specific workflows.

Next-Generation CaseBot™ Turns Case Data Into Instant, Defensible Insights

By HaystackID Staff

Produced by hand, a single issue-organized fact narrative, cited chronology, or deposition kit can cost a legal team $5,000 to $14,000 or more, according to HaystackID’s analysis of Am Law 100 midlevel associate rates of $650 to $1,000 per hour. Much of that work is never commissioned at all because the cost puts it out of reach.

CaseBot produces those analyses in minutes, using agentic retrieval and Specialized Skill Agents, directly from the case file, with every factual statement linked to its source. HaystackID will introduce the next generation of CaseBot, evolving the platform from conversational case exploration into an agentic case analysis system built on technology from eDiscovery AI, a HaystackID company.

The release builds on a foundation already in production. CaseBot reached general availability on May 11, 2026, following a limited release with founding partners that began in January 2026. CaseBot gives legal teams standalone access to supported matter datasets and direct integration within Relativity workspaces. CaseBot “changes what legal teams can expect from their case data,” said Jim Sullivan, Founder and Chief Executive Officer of eDiscovery AI, in the general availability announcement. The new generation extends that foundation in three significant ways by introducing agentic retrieval, Specialized Skill Agents, and structured case intelligence outputs designed around common legal workflows. The new generation debuts at ILTACON 2026, Aug. 23-27 in Nashville, where HaystackID will demonstrate it at booth 241.

From Conversational Search to Agentic Case Analysis

The latest generation of CaseBot represents a shift from conversational AI to agentic case analysis. Earlier discovery assistants primarily responded to questions by retrieving a set of documents and generating an answer from that initial search. While useful for exploration, that approach relied on a single retrieval pass and placed the burden on the user to ask the right follow-up questions or manually connect related evidence.

CaseBot takes a different approach. Instead of simply responding to a prompt, it plans and executes a multi-step analytical workflow. It determines what information is needed, retrieves evidence, evaluates the results, identifies gaps or conflicting information, refines its search, and repeats the process until it has assembled sufficient evidence to support a structured analysis. The result is not simply a better answer to a question but a more comprehensive understanding of the record.

That agentic architecture also provides the foundation for the Specialized Skill Agents. Each agent applies that agentic workflow to specific legal tasks, allowing attorneys to generate consistent, evidence-backed case intelligence outputs directly from the case file. Rather than functioning as generic prompts, each agent has been designed by experts around a specific legal workflow and applies the same iterative retrieval and evidence evaluation process to produce consistent, evidence-backed outputs. Whether developing a chronology, preparing for a deposition, evaluating a claim, or testing factual assertions against the record, each agent follows a purpose-built methodology while working exclusively from the underlying case file.

Each workflow concludes by generating a professionally formatted, exportable PDF that captures the analysis in a structured, shareable format. The reports include an executive summary, organized findings, supporting citations, and documentation of the evidence underlying each conclusion, allowing attorneys to incorporate the results into existing workflows and share them with colleagues, clients, experts, or other stakeholders without requiring access to the CaseBot interface.

The result is a system that moves beyond conversational exploration to perform structured case analysis, allowing legal teams to generate consistent, reusable case intelligence directly from the data they already manage.

That export capability answers a request HaystackID and eDiscovery AI heard repeatedly from clients and their outside counsel. Esther Birnbaum, Executive Vice President of Data Intelligence at HaystackID, said in a pre-launch offering briefing that the agents grew directly out of client conversations. Firms wanted tailored work product for specific discovery workflows, she said, produced where the data already lives rather than exported to a general-purpose legal AI tool. Attorneys who run an agent for multiple deponents or issues can save the exports locally and share them with stakeholders who never log into the platform, extending the analysis beyond the review team.

Eight Specialized Skill Agents, Ready to Run

At the center of the new generation of CaseBot are eight Specialized Skill Agents, each designed to execute a distinct legal analysis workflow. Rather than functioning as general-purpose prompts, the agents apply a consistent, purpose-built methodology to retrieve evidence, evaluate the record, identify supporting and conflicting information, and produce structured, evidence-backed analysis tailored to a specific legal task.

The agents are organized into two categories: Fact Development and Case Preparation, reflecting the progression from understanding the record to acting on it.

Fact Development agents help attorneys build a comprehensive understanding of the evidence. Fact Narrative develops an issue-organized account of what the record establishes, where the evidence is concentrated, and where important gaps remain. Chronology constructs a cited timeline for a matter, topic, or individual, highlighting documented events while identifying missing or conflicting evidence. Entity Map analyzes the relationships among people and organizations, illustrating key participants, their roles, and their connections within the record. Hot Document Finder identifies the documents most likely to accelerate case understanding by ranking materials based on their analytical significance and supporting each recommendation with cited reasoning.

Case Preparation agents transform that understanding into targeted litigation and investigation support. Deposition Kit assembles witness-specific facts, supporting documents, contradictions, and suggested examination topics. Claim Analysis evaluates individual allegations by organizing evidence that supports or contradicts each claim while identifying areas where the record remains inconclusive. Request Analysis breaks discovery requests into discrete elements and identifies candidate responsive documents supported by citations. Record Check compares factual assertions, including uploaded deposition transcripts, against the documentary record and classifies each statement as supported, contradicted, mixed, or unsupported, with citations to the underlying evidence.

Although each agent addresses a different legal workflow, they share the same underlying approach. Every analysis is generated directly from the case file, every factual statement is supported by cited evidence, and every output follows a consistent structure that allows attorneys to quickly evaluate findings, understand the basis for each conclusion, and determine where additional investigation may be warranted. Each Specialized Skill Agent concludes by generating a professionally formatted, exportable PDF that captures the analysis in a consistent, shareable format, complete with an executive summary, structured findings, supporting citations, and documentation of the evidence underlying the analysis.

The Specialized Skill Agents were developed in response to real-world legal workflows rather than abstract AI capabilities. Record Check, for example, originated after a law firm told HaystackID it had evaluated eight different tools for transcript fact-checking without finding one that could handle the workflow. Birnbaum said that conversation highlighted a recurring need for an evidence-based mechanism to validate factual assertions against the documentary record, leading to the development of Record Check as a dedicated analytical workflow rather than another conversational prompt.

Agentic Retrieval Changes How CaseBot Analyzes the Record

The first generation of discovery chatbots, eDiscovery AI’s original CaseBot among them, relied on single-pass retrieval: one search against the index, one set of documents returned, one answer generated. Sullivan said in the prelaunch briefing that the answer-generation step was never the weak link; retrieval was. CaseBot replaces the single pass with an agentic loop. It works out what information it needs, runs a search, reads what comes back, learns from the results, and searches again, repeating the cycle until it has assembled the evidence to respond.

The difference is visible in the product itself. Rather than returning a block of text after one lookup, CaseBot narrates its progress as it reads documents and runs additional searches, so users watch the investigation unfold. Ask who a particular custodian is and the assistant establishes aliases and email addresses, confirms roles across multiple sources, and situates the person in the context of the matter before it answers. HaystackID describes the shift as a move from conversational AI to agentic legal intelligence.

What the Record Establishes and What It Does Not

Trust boundaries are built into every response. CaseBot works exclusively from the case file. Every factual statement includes a citation. Analyses flag supporting evidence, conflicting evidence, and evidentiary gaps, and the assistant makes no legal conclusions or credibility determinations. Those boundaries echo the design of CaseBot itself, which HaystackID positions as directional intelligence, “not a substitute for traditional document review, fact development or attorney judgment.”

The assistant operates within HaystackID’s secure infrastructure, with encryption in transit and at rest, role-based access control tied to existing Case Insight permissions, and logging that supports audit trails. Before release, CaseBot underwent functional testing, accuracy and quality review, user acceptance testing with senior developers and legal data experts, and security review with penetration testing. Teams planning deployments should also address index retention in matter scoping and governance discussions, since ongoing question-and-answer work relies on a maintained index.

Where Case Insight Ends and CaseBot Begins

HaystackID draws a clear distinction between the two products. Case Insight understands the data: it reads the entire document population, identifies what exists, what is missing, and what matters, and answers the question “What do we have?” CaseBot, acting through its Specialized Skill Agents, understands the record: it performs deep analysis of a targeted document set and builds evidence-backed work product, answering “What does it mean?”

Case Insight remains the upstream engine. Given only a document population and a brief case background, the GenAI-powered matter intelligence platform classifies and categorizes the data with no seed sets, training rounds, or iterative prompting, delivering recall rates of 90% and above for relevant materials, cutting review volumes by as much as 90% and compressing fact development from weeks to days. It presents one analysis from three perspectives: a population-level view that ranks relevance and maps the composition of the data; a narrative that synthesizes facts, issues, and evidence into a document-supported story; and structured case elements organizing key people, key documents, and a document-supported chronology.

“Case Insight changes when understanding arrives,” Birnbaum said in HaystackID’s ILTACON announcement. “Legal teams no longer wait months into review to learn the story of a matter. From the moment of collection, they have the facts, the people, the timeline, and a working narrative, with every conclusion cited to the underlying record, so strategy can start on day one.”

The two solutions are designed to work together. Teams beginning a matter can use Case Insight to understand the overall document population and identify the most relevant materials before using CaseBot to investigate specific issues, develop factual narratives, prepare witnesses, or validate assertions against the record. Teams working from a targeted production or a mature case can move directly into CaseBot to perform focused analysis without rerunning the corpus.

Together, Case Insight and CaseBot provide complementary perspectives: one develops intelligence across the document population, while the other performs agentic analysis of the evidence that matters most without reprocessing the broader corpus.

Availability and What Comes Next

CaseBot supports fact validation, timeline development, communication identification, deposition preparation, and meet-and-confer planning, with application to legal, compliance, regulatory, and cyber matters, and it operates standalone or directly within Relativity workspaces.

The roadmap points further out. The prelaunch briefing described persistent saved sessions and custom agents built for individual clients, extending the model to firm-specific workflows. Attendees can see the new generation in live demonstrations at booth 241 throughout ILTACON 2026. CaseBot is available to qualified HaystackID clients as part of the Core Intelligence AI™ portfolio, and teams can contact HaystackID to schedule a demonstration or briefing.

Expert analysis that once cost five figures per deliverable now arrives in minutes with citations attached. The question for legal teams is no longer whether they can ask their case data questions; it is which analysis they will run first. Which would yours be?

News Sources

HaystackID® solves complex data challenges related to legal, compliance, regulatory, and cyber requirements. Core offerings include Global Advisory, Cybersecurity, Core Intelligence AI™, and ReviewRight® Global Managed Review, supported by its unified CoreFlex™ service interface and eDiscovery AI® technology. Recognized globally by industry leaders, including Chambers, Gartner, IDC, and Legaltech News, HaystackID helps corporations and legal practices manage data gravity, where information demands action, and workflow gravity, where critical requirements demand coordinated expertise, delivering innovative solutions with a continual focus on security, privacy, and integrity. Learn more at HaystackID.com.

Assisted by GAI and LLM technologies.

SOURCE: HaystackID