ProSearch Pulse
Julia HasenzahlJulia Hasenzahl
Join our conversation!
ProSearch Pulse

The ProSearch Pulse

This week we flagged the following:

  • Myki Fines went from a five-question flow that took under 10 seconds to 30,000 users in day one and 60,000 in month one, then helped trigger a policy reversal when the government repealed the fines regime in month two.
  • A Top 20 pension fund using Josef reported saving 15 minutes per question across HR, compliance, and legal self-service tools, while teams try to claw back the up to 40% of task time lost to “back-and-forth” clarification.
  • Judge Jed S. Rakoff (S.D.N.Y.) held that AI-generated research documents created in the consumer version of Anthropic’s Claude were not protected by attorney-client privilege or work product (Feb. 17, 2026 opinion), pointing to Claude’s non-attorney status and terms that allow training on user data and potential disclosure to third parties, including regulators.
  • The eDiscovery software market is projected to nearly double from $5.2B to $10.1B by 2033 (an 8.3% CAGR, 2026–2033), with vendors spanning legal-tech specialists like Relativity and Everlaw alongside platform giants like Microsoft and OpenText.
  • Legal ops leaders are treating 2026 as a pivot year from tactical cost-control to strategic influence, leaning on benchmarking tools like LegalVIEW DynamicInsights to reframe outside counsel conversations around data transparency, while scaling AI into unglamorous but high-volume work like invoice review, matter summaries, and email management.

In This Issue:

  • Privilege hits a wall
  • Legal ops in 2026
  • Architecture beats chatbots
  • eDiscovery heads for $10B

Legal and compliance teams are surfacing a quieter risk with generative AI: false confidence. Hallucinations get headlines, but “pretty good” answers slip into policy, advice, and decisions without the right safeguards. That’s where the damage starts.

Sam Flynn, Co-Founder and Chief Operating Officer at Josef, puts a number on it. He pegs LLM-only chatbots at 60–70% accurate. Humans still have to clean up 25–30% of outcomes. That looks like rework, extra review overhead, and new failure modes.

Flynn’s work also shows how legal tools can drive public impact. In 2016, his Myki Fines product reportedly reached 30,000 users in a single day and helped trigger a government repeal by the second month.

Many legal teams also deal with requests arriving through 10 different channels. A “legal front door” can centralize intake and standardize triage. It also forces a clear line on what stays human-led. When the cost of being wrong includes real penalties, including jail time, automation becomes a governance design problem.

Privilege meets consumer AI

Judge Rakoff denies privilege for AI-generated legal research

Story Image

A new SDNY opinion says consumer Claude chats lacked confidentiality and human fiduciary safeguards, blocking attorney-client privilege and work-product claims.

A federal judge in New York is drawing sharper lines around legal privilege and consumer AI tools. In a February 17, 2026 written opinion, Judge Jed S. Rakoff (SDNY) explained his earlier ruling that AI-generated documents created using the consumer version of Anthropic’s Claude were not protected by attorney-client privilege or the work product doctrine.

What the court held, and why it mattered

  • Privilege failed at the threshold Rakoff treated attorney-client privilege as a product of a trusting human relationship with a licensed professional who owes fiduciary duties and faces discipline. A consumer AI platform does not qualify.
  • Confidentiality expectations drove the outcome The consumer Claude policy contemplates training on user data and potential disclosure to third parties, including regulators, in disputes or litigation, even without a subpoena. The court found that posture incompatible with a reasonable expectation of confidentiality.
  • The third-party problem The court treated the Claude interaction as sharing potential “notes for counsel” with a third party before any privileged channel existed, which undercut privilege arguments tied to later attorney communications.
  • Work product protection did not attach Rakoff found the documents were not prepared by or at the direction of counsel and did not reflect defense counsel’s mental impressions. He declined to follow Shih v. Petal Card, Inc. and emphasized the doctrine’s core purpose: protecting lawyers’ mental processes.

What legal teams should take from this

This opinion reads like a warning label for consumer AI. Privilege arguments turn on two gears that must mesh:

  • The relationship: a human professional duty framework that supports privilege
  • The data terms: whether confidentiality expectations are objectively reasonable under the platform’s policies

Rakoff also left room for future disputes, including whether enterprise contracts or a Kovel-style agent theory, used at counsel’s direction, could change the analysis.

via www.debevoise.com

Read Now →

Legal ops, retooled

Legal ops sets three priorities for 2026

Story Image

A Wolters Kluwer podcast summary maps how legal operations is shifting toward data-driven firm management, scaled AI workflows, and new team roles.

Legal operations is heading into 2026 with an expanded mandate: create measurable business value while controlling cost, and manage a growing ecosystem of outside counsel and legal technology. A Wolters Kluwer summary of the 33rd episode of Legal Leaders Exchange, recorded at Amplify 2025, flags three priorities reshaping how the function operates.

Key insights

  • Outside counsel management is shifting toward shared data. The focus is on transparency, benchmarking, and proactive planning with law firms, with less time spent on reactive rate negotiations or post-bill cleanup.
  • Spend analytics is becoming the negotiating table. Tools such as LegalVIEW® DynamicInsights are positioned as a way to make rate and billing discussions recurring, structured working sessions grounded in data.
  • AI attention is moving to workflows that scale. The highlighted use cases include streamlining invoice review, drafting matter summaries, managing email volume, and using an assistant to cut administrative workload.
  • Legal ops roles are expected to change over the next five years. As automation and specialization increase, hiring will lean further into technology and process improvement. Smaller teams will face sharper prioritization trade-offs.

For legal leaders, the throughline is practical. 2026 planning is about turning tools into repeatable operating systems, not chasing novelty. Teams that pair transparent outside counsel management with measurable workflow automation can redirect attorney time toward complex, high-stakes work.

via www.wolterskluwer.com

Read Now →

Joe Pirrotta

Joe Pirrotta

Legal ops “leveling up” feels like a New Year’s resolution we can actually keep, if we keep it practical. Three moves: (1) bring firms into shared benchmarks, (2) scale AI in repeatable admin workflows, (3) redesign roles for judgment, not busywork. It won’t be perfect; start anyway.

Feature Image

Why Legal AI Needs Architecture, Not Just Chatbots

Here’s a conversation we’ve been pointing partners to because it cuts through the legal AI noise with operational clarity. In Great Women in Compliance, Dr. Hemma R. Lomax speaks with Sam Flynn, COO and co-founder of Josef, about why AI in legal and compliance succeeds or fails based less on model selection and more on whether the system is designed for trust.

Flynn’s throughline is human-centered design as the non-negotiable. Legal and compliance operate like internal service teams. When the experience for the business is fragmented, with requests arriving through up to ten channels, long back-and-forth cycles, and policy scavenger hunts, even strong models fail to produce meaningful lift.

He grounds the point in concrete examples. Flynn recounts building Myki Fines in Australia, a five-question tool that delivered guidance in seconds, attracted 30,000 users on day one, reached 60,000 in the first month, and helped push the government to repeal the fines regime in month two. Later, he describes a lesson from work with the University of Michigan, where automating a form surfaced a better answer: eliminate the form entirely. In that framing, “substance over form” becomes a design principle you can build around.

What stood out

  • The “front door” is an AI strategy Centralizing legal and compliance intake, rather than letting work arrive through up to 10 different channels, creates visibility into demand, reduces switching costs, and clarifies what is worth automating next.

If your chatbot is only 60–70% accurate, the value isn’t the model—it’s the workflow architecture that makes it safe.

  • LLM accuracy is only one part of the risk equation Flynn points to a common chatbot accuracy range of 60 to 70 percent, which leaves a meaningful share that still needs human validation. If validation comes only at the end, scale stays constrained.
  • Architecture is the product Closed-domain content, clear escalation paths, analytics, and expert oversight turn a clever demo into something teams can run in compliance workflows. This is the AI + HI model in practice: automation for speed, humans for judgment.
  • High-risk topics require hard lines Flynn describes removing automated guidance for insider trading despite strong performance because the downside is extreme, potentially jail time, and the stakeholders are senior. The system routes those questions directly to a lawyer.

What it raises

  • If trust unlocks value, what are the minimum governance signals users need before they rely on an internal AI tool?
  • Who should own the legal architecture layer: legal ops, compliance, IT, or a cross-functional group with shared accountability?
  • Intake design serves two masters: reducing friction for employees and producing defensible records for audit and accountability. Teams do better when they design for both outcomes on purpose.
  • Clear thresholds for self-service versus mandatory escalation matter, and those thresholds should be revisited as policy changes, risk tolerance evolves, and usage data reveals real-world edge cases.

Credit to Dr. Hemma R. Lomax for steering the conversation toward how these systems work inside real organizations, and to Sam Flynn for making the case that legal and compliance AI ROI is primarily a workflow and trust problem. The right architecture supports it. Model shopping does not.

via www.youtube.com

Listen Now →

Legal tech meets AI

eDiscovery software market seen hitting $10.1B by 2033

Story Image

A PR Newswire release cites Market Research Intellect projections, pointing to litigation growth, compliance pressure, and analytics-driven review tools.

Electronic discovery software keeps finding steady footing as digital evidence piles up across email, chat, cloud repositories, and multimedia. A paid PR Newswire release carried by Yahoo Finance cites Market Research Intellect estimates that the market could grow from $5.2B to $10.1B by 2033.

Key insights from the release

  • The release projects 8.3% CAGR from 2026–2033, keeping eDiscovery positioned as a durable spend category tied to litigation, investigations, and regulatory audits.
  • Expanding volumes of electronically stored information (ESI), including collaboration chats and social media data, keep pushing organizations away from manual evidence handling and toward automated, defensible workflows.
  • The growth narrative centers on advanced analytics, AI-powered document review, and cloud-based deployments to improve speed, accuracy, and scalability during collection and review.
  • The competitive landscape blends legal-tech specialists with large enterprise vendors, including Relativity, OpenText, Microsoft, Exterro, Nuix, and Everlaw.

What this signals for legal and discovery teams

For most buyers, the headline number matters less than what it reflects on the ground. More digital litigation and tighter compliance regimes raise expectations for defensible processes, faster response times, and tooling that can handle cross-platform data at scale.

From what we see with our partners, differentiation in an expanding market concentrates around three practical levers:

  • AI review quality, including how teams validate results and keep human judgment in the loop
  • Integrations with enterprise systems, especially where data lives across Microsoft, collaboration platforms, and cloud content stores
  • Cloud governance readiness, so scale does not compromise security, retention, or auditability

 

via finance.yahoo.com

Read Now →