How Harvey drives AI transformation and token efficiency with Glean
Business Outcomes Driven
- Org-wide: Increase in employee satisfaction
- Org-wide: Decrease in repetitive questions and requests, preserving SME capacity
- Product: Faster identification and prioritization of feature requests that will move the needle for Harvey customers, driving customer satisfaction
- Onboarding: Faster onboarding decreasing time to impact per new employee
- Support: Lower cost per case by reducing manual investigation and handling time
Introduction
Harvey is an AI platform that empowers legal and professional services teams to make faster, smarter decisions. With over 3,500 customers across more than 80 countries, including global law firms and Fortune 500 enterprises, Harvey provides a secure, centralized platform for contract analysis, due diligence, compliance, litigation support, and workflow automation—enabling legal professionals to streamline high-volume workflows and strategic work.
Avant Glean
As a fast-growing company welcoming over 80 new employees every two weeks, and a leader in AI for legal teams, Harvey was intent on staying at the forefront of AI even as they faced mounting challenges around onboarding and workflow optimization to drive productivity.
With knowledge becoming hard to find across systems and experts as the company quickly scaled, it created a tax on champions and technical teams who had to answer repetitive questions. Harvey needed a solution that would reduce time-to-productivity for new hires, provide a unified knowledge layer across its entire tech stack, and keep them at the forefront of AI as they partnered with their customers to drive AI transformation across the legal sector.
That’s why they turned to Glean.
Avec Glean
Harvey’s single source of truth
Harvey employees were navigating dozens of systems with fragmented knowledge and no centralized way to find answers—forcing them to either hunt through multiple applications or repeatedly track down subject matter experts. Harvey initially rolled out Glean with just search, focusing on solving their biggest knowledge-discovery problems. Today, Harvey’s teams rely on Glean as the starting point for many of their workflows—a trusted enterprise knowledge source that works out of the tools they already use.
Agents that drive workflows
Glean enabled Harvey's teams to experiment with, build, and deploy agents that tackle their most time-consuming workflows. From IT operations to customer success, Harvey built agents that automate aspects of complex, multi-system workflows and give hours back to employees for higher-value work
Triggered through Glean MCP, these agents are embedded directly within Harvey’s Slack channels, providing precise, up-to-date answers for legal, product, customer success, and sales to help customers get answers faster. These agents resolve an average of 100 questions per month without human intervention required, and triage another 120 questions per month, saving several minutes per question and speeding up the time to resolution. The agents have also inspired other teams like Sales and Support to build similar Ask agents of their own.
Run primarily by customer support, legal, product, and customer success teams, this agent forms a comprehensive view of customer feedback and issues across internal platforms to identify trends and important product requests. This has significantly sped up the process of submitting customer requests, identifying themes across customers that should drive roadmap decisions, and in turn, improved customer satisfaction and outcomes.
This agent reviews internal support tickets, past history and resolutions, and suggests answers to customer questions based on context and internal documentation. Prior to this Agent, customer-facing teams spent significant time reviewing documents and pinpointing potential fixes or commonalities that could drive resolution with a given issue. With the Agent acting as a partner, they are driving faster resolution without first having to review hundreds of Slack channels and documents to pinpoint commonalities across issues.
This agent helps Harvey’s managers prepare for one of leadership’s most important, yet time-consuming, responsibilities: performance reviews and peer feedback. The agent surfaces key points to fold into team member feedback, including projects the manager may have forgotten about or wins they didn't remember to capture. Treated as a strong starting point rather than a final product, this agent helps managers reclaim over 50% of their time back per review - allowing them to spend more time on true coaching and making feedback conversations personalized and actionable.
Better token efficiency through auto-routing
Harvey never focused on maximizing token usage when it came to measuring AI adoption and impact—rather, they focused on the best way to leverage AI to make a difference. However, with so many new models and tools available, it became difficult to keep everyone up to speed on which models were best for which tasks in terms of token efficiency and fit.
Glean provided them with a way to tackle this by having an overarching system handle the decision-making. Glean’s auto-routing enables Harvey’s employees to query confidently without worrying about the granularity of model selection.
“Glean’s auto-routing has given Harvey teams a safe, cost-effective way to experiment with agents and produce high-quality output, without worrying that exploration will lead to unexpectedly high token costs.” said Curtis Salinas, Head of Business Technology at Harvey.
Driving transformation: The AI fluency program
Looking to push transformation initiatives across the entire company, Harvey built a structured AI fluency program organized around three pillars:
Build focuses on enabling communities of builders across the company. While not every employee is expected to create agents on their own, this initiative looks to embed capable builders in each team capable of developing efficient, purpose-built AI workflows rather than generic ones.
Measure keeps Harvey focused on how they gauge success. Looking for meaningful indicators of impact rather than just usage and token volume, Harvey tracks individual agents and platforms to see what’s actually being improved. For example, on their ITSM platform, Harvey tracks how much of the case load across teams is automated or assisted by AI, with the goal of increasing that over time.
Enable takes the form of office hours. During these, teams may build live agents, show how to take an idea and turn it into a skill or Glean agent, or demonstrate how to use prebuilt agents. These sessions also now go in-depth into specific use cases, feature guest speakers from across the company, and serve as a way for employees to share what they built, how they did it, the challenges they encountered, and how they overcame them. Employees can also sign up for 1:1 office hour slots to get help with an idea or tool they’re building. They’ve served to grow an AI-first, builder community and culture within Harvey.
Better onboarding, satisfied customers
With Harvey onboarding over 80 people every two weeks, time-to-productivity was critical. New hires entering complex environments with processes unfamiliar to them now had a clear place to start. The impact was immediate: Harvey saw a 20% increase in CSAT from onboarding employees. With a more comprehensive, accelerated onboarding process, new employees could make an impact more quickly.
