The ProSearch Pulse Volume 22

Legal departments are moving on AI, but many are still in trial mode. In a JD Supra analysis, Scott Bien cites the latest Blickstein Group Law Department Operations survey: 52% of legal teams are still running generative AI pilots, while 23% say genAI is fully deployed and operational. Agentic AI is moving from concept to workflow.

Key insights

• Adoption remains uneven. 52% of departments are still in genAI pilots versus 23% fully deployed, widening the execution gap across legal operations.
• Agentic use cases are taking shape. 20% are actively deploying AI agents for contract negotiation and another 40% are exploring the technology.
• The biggest brakes are operational and risk-driven. 44% of legal professionals cite security and accuracy concerns, and many teams are dealing with tool fragmentation that limits interoperability.
• Bien’s rollout approach emphasizes discipline. Define success metrics and a go or no-go timeline before issuing logins, prioritize ROI-tied use cases, and assume tools may be replaced within two to three years.

The practical takeaway for in-house leaders is that progress depends on treating AI as a workforce change program anchored in data readiness. Bien sets a clear target for agentic automation: agents complete the first 80% of work while humans supervise and finish the final 20%. His forecast is direct: by the end of 2026, staying in pilots becomes a competitive disadvantage.

When we work with partners on legal data strategy, this maps to a familiar pattern. Early planning and clear success criteria determine whether genAI becomes operational or stays in experimentation. The teams that move fastest translate legal work into measurable workflows, validate assumptions with controlled pilots, then scale.

A new Bloomberg Law report puts a clear number on what many legal teams are feeling: legal AI is moving into the mainstream, but most Big Law technology budgets stay tight. Many firms allocate only a modest slice to “new technology,” while subscriptions and other recurring costs take up more of the technology spend.

What Bloomberg Law reported

• Roughly 67% of large firms said they spent under 2% of their 2025 budgets on building or buying new technology, a sign of cautious near-term investment
• Among firms with fewer than 500 attorneys, nearly 30% spent more than 5% of budget on new technology last year, suggesting smaller firms may be investing proportionally more
• Recurring costs are a major line item: one in three firms under 500 attorneys said they will spend more than 8% of 2025 budgets on technology subscriptions and recurring costs
• At the top end, firms are testing the economics of in-house AI. Kirkland & Ellis said it plans to invest $500 million to build a proprietary AI model

What this spending mix suggests

Legal technology planning now looks like managing a fleet, not buying a single vehicle. Subscriptions keep teams moving quickly, but they also turn innovation into an operating expense that compounds quarter after quarter. Building can create leverage when scale and proprietary data matter, but it demands patience, governance, and sustained investment.

Practical implications for legal leaders

• Subscription-first strategies speed up adoption, but they require tighter vendor management, clearer usage controls, and disciplined renewal cycles
• Selective build decisions make sense when a firm has the data, scale, and differentiation to justify ownership economics
• A hybrid roadmap is the near-term default: buy for speed, build where proprietary advantage is real, and standardize guardrails so both paths stay governable

Where this is heading

Industry AI spending looks set to accelerate in 2026 and beyond. Many firms will revisit what they want to own versus rent. The firms that plan early, validate assumptions through pilots, and keep a clear line from technology choices to measurable practice outcomes will adapt fastest.

At ProSearch, we see the same tension in eDiscovery and legal data programs. Speed matters. Defensibility matters. The teams getting ahead of it treat technology spend as a roadmap decision, then use an AI + HI model to turn legal data into strategic insight.

Relativity has introduced aiR Assist inside its RelativityOne eDiscovery platform. It’s a natural-language Q&A feature that produces cited answers directly within legal datasets. The pitch is faster, defensible answers with less need to export sensitive information or switch to external tools.

Key insights

• In-workflow, source-tied Q&A
• aiR Assist is embedded in RelativityOne workflows and returns answers with citations.
• Review stays anchored to the underlying documents, supporting traceability and audit expectations.

• Built for legal and investigative work
• Relativity targets practical tasks like early case assessment, relationship mapping among key actors, and thematic analysis across large document collections.
• The point is faster time to insight with the evidence trail intact.

• Deposition and theory validation as a headline use case
• Another advertised workflow is validating legal theories ahead of depositions.
• In that context, showing how an answer was derived matters as much as getting to it quickly.

• The expectation is improved retention, expanded usage within existing accounts, and support for premium pricing or higher seat utilization.

There's a partner-led push to put the feature to work inside enterprise legal teams. Embedded AI is becoming standard across legal and compliance platforms. Relativity is betting that cited, in-workflow Q&A strengthens platform stickiness and increases recurring revenue.