DocuSign's CEO on AI Dangers: Why Legal Tech Stays Cautious
When DocuSign's CEO talks about AI, he doesn't sound like most tech executives. Allan Thygesen warns about risks others minimize.
The reason: legal documents carry real liability.
The Intelligent Agreement Shift
DocuSign started as e-signatures. Now it's evolving into "Intelligent Agreement Management."
The vision: AI that handles the entire contract lifecycle—from drafting to signing to enforcement.
But the approach is deliberately cautious.
Why Legal AI Must Be Conservative
Legal documents create binding obligations. Errors carry financial and legal consequences.
This changes everything about how AI should be deployed:
Hallucinations are unacceptable. A contract AI that makes things up exposes clients to liability.
Source verification matters. Every AI-generated provision should trace to real legal authority.
Human oversight is non-negotiable. Even good AI outputs need lawyer review.
DocuSign reorganized around these principles. More engineering investment. More emphasis on accuracy over capability.
The Competitive Advantage
Counterintuitively, caution creates competitive advantage.
DocuSign's large contract library contains millions of real agreements. This data enables AI that learns from actual legal language—without making things up.
Competitors without this data face a choice: build risky AI or build trustworthy AI. Building trustworthy takes longer but creates defensible moats.
The Product-Led Evolution
DocuSign's shift mirrors broader enterprise trends:
Product-led growth. AI features that demonstrate value directly to users.
Engineer-heavy teams. Investment in technical talent over sales force.
Benchmarking rigor. Extensive testing before releasing AI features.
This approach accepts slower launches in exchange for reliability.
What Other Industries Can Learn
Legal tech's caution offers lessons for other high-stakes AI deployments:
Start with human-in-the-loop. Pure automation looks impressive but carries unacceptable risk in consequential applications.
Invest in verification. Source-linking and citation matter when accuracy has legal implications.
Accept slower deployment. Speed matters less than reliability when mistakes cost real money.
The Takeaway
DocuSign's conservative approach might seem backward in an era of aggressive AI. But in legal contexts, trustworthiness beats capability.
The lesson: not every AI application should move fast and break things. Some domains require the slow, careful approach that Silicon Valley typically rejects.
Stay ahead of AI trends. tldl summarizes podcasts from builders and investors in the AI space.