Freshfields

Freshfields Bruckhaus Deringer, a global Magic Circle law firm with 5,700 staff, faced mounting pressure to remain competitive as AI began reshaping legal services. Associates and partners were spending disproportionate hours on high-volume, repeatable tasks — sifting through case law, reviewing transaction documents, conducting due diligence across large contract sets, and managing e-discovery workflows. In a profession where billing rates and regulatory scrutiny are both high, the inability to systematically accelerate these tasks represented a real constraint on capacity and margin. The firm needed a firm-wide approach, not ad-hoc tool adoption.

Freshfields partnered with Google Cloud to deploy Google Workspace with Gemini across all 5,700 staff, making generative AI accessible at the desktop level from day one. The rollout was complemented by a roadmap to build bespoke AI agents using Vertex AI, targeting repetitive internal legal and business processes. NotebookLM was selected for knowledge extraction from existing firm materials, while Gemini's capabilities were embedded into proprietary tooling designed to accelerate document review and due diligence contract analysis. The deployment was structured to maintain human oversight at every stage — a deliberate architectural choice driven by the firm's professional obligations and regulatory environment across multiple jurisdictions.

The deployment has driven measurable progress across several high-value workflows:

• Legal research: Gemini reduces time spent locating and synthesising relevant case law and precedents
• Document drafting: AI-assisted drafting cuts time on first-pass document preparation
• E-discovery: AI tooling accelerates document categorisation and relevance review
• Due diligence: Semi-automated contract review compresses timelines on transaction work

Adoption is active across all 5,700 staff covered by the rollout. A cross-functional innovation board and a dedicated AI oversight committee have been established to govern deployment, track outcomes, and ensure compliance with professional conduct rules.

• Governance must be built before scale: Freshfields created both an innovation board and an AI oversight committee before broad rollout — not after issues emerged.
• Blanket access works better than selective pilots when the goal is cultural adoption across a large professional workforce.
• Human-in-the-loop is non-negotiable in legal AI — regulatory obligations and liability exposure make autonomous outputs unacceptable without attorney review.
• Cloud vendor co-development matters: active feedback to Google allowed tools to be shaped for regulated-industry requirements rather than generic enterprise use.
• Purpose-built agents outperform generic tools for high-frequency internal processes — the Vertex AI agent roadmap reflects this prioritisation.