How would you measure the success of Harvey Workflows?

Introduction

Harvey Workflows are pre-built and custom multi-step processes for legal work. Instead of writing a prompt from scratch, a lawyer picks a workflow such as "Draft a client alert," "Review a lease against a checklist" or "Prepare a deposition outline," uploads the inputs, and Harvey runs a series of steps designed by legal experts. Firms and legal departments can also build their own workflows with Harvey's Workflow Builder, packaging their know-how so any lawyer can use it. Harvey has reported that customers have built tens of thousands of custom workflows and agents on its platform.

The prompt asks how we would measure the success of Harvey Workflows. Workflows sit between chat (Assistant) and fully autonomous agents, so success must capture both how often lawyers use them and whether they produce work lawyers trust.

Step 1

Product Context (3 minutes)

Key stakeholders:

• Lawyers (associates, in-house counsel): want reliable outputs faster than doing the work by hand.
• Knowledge management and innovation teams: build custom workflows that capture firm expertise.
• Firm leadership and general counsels: want a return on Harvey licences and consistent quality.
• Harvey: wants daily usage, seat expansion and differentiation from general AI assistants.

User flow:

1. Lawyer opens Harvey and browses or searches the workflow library.
2. Selects a workflow and provides inputs (documents, instructions).
3. Harvey runs the steps, sometimes asking for confirmation between steps.
4. Lawyer reviews the output, edits it, and exports to Word or saves to Vault.
5. Knowledge teams monitor which workflows are used and refine them.

Product lifecycle stage: Growth. Workflows are well established, and custom workflow creation is expanding quickly as customers encode their own processes.

Competitive context: Thomson Reuters CoCounsel skills, Legora and other legal AI platforms, plus generic tools such as Microsoft Copilot. Harvey's advantages are legal-specific workflows, a builder that firms use to encode their own methods, and integration with Vault and Knowledge sources.