Docusign's CEO on the dangers of trusting AI to read, and write, your contracts

Summary of Docusign's CEO on the dangers of trusting AI to read, and write, your contracts

by The Verge

1h 5m•February 2, 2026
Listen to Episode

Overview of Decoder — Docusign's CEO on the dangers of trusting AI to read, and write, your contracts

This Decoder episode (host Neil A. Patel) is a wide-ranging interview with DocuSign’s CEO (referred to in the episode as Alan) about what DocuSign actually does, how the company is organized, and — centrally — how it’s using AI to read, summarize, and manage contracts. The conversation probes technical choices (models, data, accuracy), product strategy (moving beyond signatures to the full agreement lifecycle), and the legal/ethical risks of letting LLMs interpret or draft agreements.

Core topics covered

  • DocuSign’s fundamental product: e-signatures as the combination of identity verification + recorded consent.
  • Company footprint and structure: ~7,000 employees; ~1.8M companies as customers; >95% of Fortune 500 use DocuSign; ~$3B+ revenue; ~1,200 engineers and ~1,000 account executives.
  • Expansion from signature-only to end-to-end agreement management: drafting, mass customization, approvals, ingestion, repository, and lifecycle automation.
  • Intelligent Agreement Management (IAM): DocuSign’s AI product for extracting data from agreements, surfacing insights, and summarizing clauses for users.
  • Identity, risk and compliance: multiple verification methods (biometrics, digital IDs, notary, risk-based flows).
  • AI architecture and sourcing: using frontier foundation models (OpenAI/GPT, Gemini, Anthropic etc.), multi-model strategy, and vendor contracts (typically 3+ years).
  • Accuracy, hallucination risk, and guardrails: human-in-the-loop, risk scoring, source linking, consent for private data usage.
  • Commercial model: bundling AI features into subscription plans and charging a premium for AI-enabled capabilities.

Main takeaways

  • What “signing” means: DocuSign treats a signature primarily as identity + consent; the company records an auditable trail designed to stand up in court, rather than emulating a handwritten signature.
  • DocuSign’s moat: network trust + deep integrations (Salesforce, Workday, CRMs, Word/Google for authoring) and the repository of agreements. That repository becomes a competitive advantage for AI-driven extraction and analytics.
  • AI is a horizontal enabler, not a replacement for lawyers: IAM is positioned as an assistive tool (summaries, extraction, deviation alerts), not legal advice. DocuSign deliberately includes human checkpoints and heavy disclaimers.
  • Accuracy improved with data: initial extraction models trained on public docs did well; early private-document testing lowered accuracy ~15 percentage points; accuracy has improved as the company has built a consented private corpus (150M+ agreements, growing).
  • Cost dynamics favor adoption: foundation-model costs per token have fallen sharply, making it viable for DocuSign to include AI functionality in standard enterprise pricing for many customers.
  • Go-to-market and org shift: the CEO pivoted DocuSign from sales-led to product-led, reinvesting in engineering/product and leaning more on partners/system integrators.

Notable quotes & concrete figures

  • Signature defined as “identity and consent, commingled.”
  • “Mass customization is…advanced mail merge.” (CEO’s refreshingly down-to-earth take on common automation.)
  • Company stats cited in the episode: ~7,000 employees, 1.8M companies as customers, >95% of Fortune 500 usage, $3B+ revenue, ~1,200 engineering staff.
  • AI adoption figures: 25,000 customers live on the new AI platform within ~18 months; 150 million private consented agreements used to improve accuracy; accuracy initially fell ~15 percentage points when moving from public to private documents, then improved with more private data.
  • Contracts: many foundation-model vendor deals are multi-year (3+ years).

How DocuSign constrains AI risk (guardrails)

  • Consent and data governance: customers explicitly opt in to let DocuSign process private agreements; opt-out capability provided.
  • Human-in-the-loop: deliberate checkpoints where a human reviewer must sign off for higher-risk automations.
  • Risk scoring and benchmarking: extracted results are scored and compared to company templates/playbooks to highlight deviations.
  • Source transparency: outputs link back to source clauses and documents to “show the work.”
  • Positioning and messaging: marketed as assistive summaries and insights, not legal advice; users are warned to consult lawyers for sensitive matters.

Business and product implications

  • Competitive positioning: DocuSign leverages its repository and workflows as an advantage even while relying on third-party LLMs for inference.
  • Commercial model: DocuSign is bundling AI features into subscriptions rather than metering AI per token for most customers; customers have been willing to pay a premium for AI-enabled agreement management.
  • Vendor dependency: DocuSign uses multiple frontier models (OpenAI, Gemini, others) and designs its stack to swap or score models as needed.
  • Long-term industry dynamics: CEO expects both enterprise and consumer use cases to coexist; model providers will pursue different strategies (consumer advertising vs. enterprise deals), and large cloud/API commitments create multi-year inertia.

Practical recommendations from the episode (for listeners who manage contracts)

  • Treat AI contract summaries as helpful orientation tools, not definitive legal interpretations.
  • Keep humans in critical decision loops: use automated extraction to surface issues, then let domain experts validate.
  • Maintain a single, centralized, searchable agreement repository to enable useful analytics and AI benefits.
  • Use template/playbook benchmarking to flag deviations and prioritize review effort.
  • If using commercial AI services, ensure you have consent and clear data governance for private documents.

Short evaluation / verdict

  • DocuSign is evolving from a signature utility into an AI-enabled agreement management platform. The CEO frames the move as both a moral and business necessity: providing summarized context improves consumer outcomes and opens enterprise value.
  • Key risks remain around hallucination, over-reliance on automated advice, and vendor lock-in; DocuSign is addressing those with data consent, human review, and source-linking, but the technology isn’t a substitute for legal counsel in sensitive cases.
  • From a buyer perspective, the immediate upside is productivity (faster discovery, renewal prioritization, deviation detection); the prudent approach is to adopt incrementally with human oversight.

If you want a one-line takeaway: DocuSign is betting its future on turning its massive, consented agreement repository into a practical AI advantage — delivering useful automation and summaries while keeping humans, consent, and legal safeguards central.