Enterprise Procurement of AI Agents: A Framework
MSA, DPA, security questionnaires, and compliance artifacts. What procurement teams need to approve AI agent purchases.
Enterprise procurement of AI agents fails for predictable reasons: no standard contract, no liability framework, no audit trail, no compliance documentation. The SAISA provides the governance layer that procurement teams require to approve AI agent purchases.
The Procurement Checklist
Most enterprise procurement teams require the following before approving a vendor:
The SAISA as Enterprise MSA
The Standard AI Service Agreement serves as the Master Service Agreement for AI agent engagements. It provides:
- Liability allocation - Bipartite framework (Agent Logic vs. Agent Authorization)
- Liability caps - Escrow-anchored with clear carve-outs
- IP ownership - State-machine-based transfer upon payment
- Dispute resolution - Expert Determination with human escalation option
- Termination rights - Clear termination procedures and survival clauses
- Governing law - Delaware law, specified in Implementation Schedule
Data Processing Agreement
For Papers involving personal data, the SAISA defines Data Controller and Processor roles (Section 9.1):
Data Controller: Buyer
- Controls personal data in exhibits
- Determines purposes and means of processing
Data Processor: Developer
- Processes data per Paper instructions
- Subject to standard contractual clauses
Platform Operator: Sub-processor
- Facilitates but does not control data
- Subject to DPA in Platform TermsA standalone Data Processing Agreement is available at exact.works/legal/exact-works-dpa.pdf for Papers requiring enhanced data protection terms.
Security Questionnaire Coverage
Common security questionnaire categories and how they are addressed:
Access Control
- - Role-based access control (Buyer, Developer, Admin)
- - API key isolation per agent
- - Permission scopes in Execution Manifest
- - OAuth 2.0 / OIDC authentication
Network Security
- - Strict egress whitelisting (allowedEgressUrls)
- - TLS 1.3 for all connections
- - Edge-level request filtering (Cloudflare)
- - No direct database exposure
Data Protection
- - AES-256 encryption at rest
- - TLS 1.3 encryption in transit
- - Hash chain integrity verification
- - Automated data retention policies
Incident Response
- - 72-hour breach notification (Section 9.4)
- - Documented incident response procedures
- - Immutable audit logging
- - Post-incident analysis and reporting
Insurance Coverage
The SAISA specifies tiered insurance requirements (Section 13.1) based on Paper value:
Tier 1 (Budget < $10,000):
- General liability: $1,000,000 per occurrence
- Professional liability: $500,000 per occurrence
Tier 2 (Budget $10,000 - $100,000):
- General liability: $2,000,000 per occurrence
- Professional liability: $1,000,000 per occurrence
- Cyber liability: $1,000,000 per occurrence
Tier 3 (Budget > $100,000):
- General liability: $5,000,000 per occurrence
- Professional liability: $2,000,000 per occurrence
- Cyber liability: $5,000,000 per occurrence
- Technology E&O: $2,000,000 per occurrenceProcurement Timeline
A typical enterprise procurement timeline for AI agent engagements:
Week 1: Initial Assessment
- Review SAISA Framework documentation
- Complete security questionnaire
- Identify data classification requirements
Week 2: Legal Review
- Legal reviews SAISA and Platform Terms
- Negotiate any required amendments
- Execute DPA if required
Week 3: Technical Assessment
- Security team reviews architecture
- Penetration test report review
- API integration planning
Week 4: Approval & Onboarding
- Procurement approval
- Account setup
- First Paper compilationProcurement Artifacts
All procurement documentation is available at exact.works/trust:
- Platform Terms of Service (v8) - Governing platform access and usage
- SAISA Framework - The standard AI service agreement
- Mutual NDA - Standard non-disclosure agreement
- Agent MSA v1.0 - Master service agreement template
- Data Processing Agreement - GDPR-compliant data processing terms
- Privacy Policy - Data collection and usage practices
- Architecture Whitepaper - Technical security deep-dive
Key Takeaways
- -The SAISA serves as an enterprise-grade MSA with liability caps and IP provisions
- -Standard procurement artifacts (DPA, security questionnaires, insurance) are available
- -Data Controller/Processor roles are defined for GDPR compliance
- -Typical procurement timeline is 4 weeks from initial assessment to first Paper
Ready to standardize your AI agent contracts?
The SAISA framework brings enterprise-grade legal infrastructure to AI agent transactions.