A legal risk-management and contracting playbook for GCs and founders building and buying AI, recently updated to cover AI agents, agentic AI, and the Model Context Protocol (MCP).
This playbook comes from two of our webinar speakers, Yelena Amartsumian and Anastasia Vener Here’s the message from them about this playbook:
When we first published this playbook, "AI in your company" meant a chatbot or a content generator—software that produced an output for a human to review. That world has not disappeared. But a new one has arrived on top of it: AI agents that interpret open-ended instructions and act across your tools, and agentic systems that plan and pursue their own goals. And the assumptions baked into a standard SaaS contract—that the software runs and a human acts—no longer hold.
So ask one question of every AI agreement: who is acting? The further the answer moves from your team and toward the machine, the less your SaaS template protects you. This playbook walks the full spectrum—SaaS → generative AI → AI agents → agentic AI—and shows you what changes at each step.
What's inside:
The spectrum: who is acting—and why each step breaks a different contract assumption.
A practical method to identify your AI use cases and map them by risk.
Common use cases and their associated risks, from CX chatbots to autonomous agents that browse, e-mail, and transact.
Contracting for AI agents: limitation of liability, indemnification, and audit/logging—with model language.
Governing agentic AI: human-over-the-loop, coded boundaries, and audit rights that survive an incident.
Contracting around MCP and connected tools: scope, supply chain, and injection risk.
Free, updated templates: an AI acceptable-use policy and a vendor-agreement playbook.
Who it's for:
GCs and founders at AI-native and AI-enabled B2B SaaS and consumer-tech companies—with application across industries.
Subscribe to Stay in the Loop
Want to hear about practical training content as it drops? Our weekly newsletter is the easiest way to keep up. Subscribe now to get everything in one place.
