Contracting for Agentic AI Solutions: Shifting the Model from SaaS to Services
The landscape of AI contracting is evolving as agentic AI products transition from passive tools to autonomous actors. This shift necessitates moving beyond conventional Software-as-a-Service (SaaS) contracting models toward a hybrid approach that incorporates BPO-style clauses. These clauses cover service definitions, warranties, outcome-based Service Level Agreements (SLAs), broader indemnification, governance rights, and data ownership.
The Rise of Agentic AI
For years, the contracting framework for generative AI (GenAI) products has predominantly followed a familiar SaaS model. In this model, the provider supplies the GenAI product on its platform while the customer is responsible for its usage. This framework works well when the AI product functions merely as a passive tool—acting as a co-pilot that suggests but does not execute.
However, agentic AI diverges from this model. Defined as systems capable of autonomously planning and executing multi-step tasks, agentic AI does not merely suggest actions; it undertakes them on behalf of the company.
Understanding the Spectrum of Agentic AI
Agentic AI encompasses a range of products. On one end, there are general-purpose tools that allow companies to develop and build their own AI agents, requiring substantial training and direction from the company team. On the other end are agentic AI solutions crafted by providers to fulfill specific functions, such as managing payment inquiries or assisting employees in accessing benefits. Between these extremes lies a variety of control levels and company involvement.
As agentic AI solutions increasingly act autonomously, the provider-company relationship transitions from licensing a tool to delivering a service. This transformation prompts a shift in the contracting model from a SaaS framework—which often includes limited performance guarantees—to a more service-oriented approach.
Adapting BPO Concepts to AI
The business process outsourcing (BPO) industry has established market terms to address complex challenges when hiring service providers. The challenge is to effectively adapt these concepts to AI service delivery.
Key Contract Clauses for Agentic AI Solutions
This study identifies six critical clauses where the standard SaaS contracting framework falls short for agentic AI solutions and proposes BPO-style updates as a more balanced starting point for negotiations.
1. Definitions & Scope of Service
The SaaS Clause: A typical SaaS agreement defines the “Service” as a hosted software platform, granting the company a non-exclusive right to access and use it. The provider is responsible for the platform, while the user handles all actions taken with it.
The BPO-Style Solution: The “Service” would encompass the tasks and responsibilities the provider agrees to execute using AI agents, clearly delineating the provider’s delegation of authority and policy guardrails.
2. Service Warranties
The SaaS Clause: Many SaaS agreements contain disclaimers stating “THE SERVICE IS PROVIDED AS-IS.” Providers often argue against offering warranties due to the probabilistic nature of AI.
The BPO-Style Solution: This approach would incorporate performance warranties, ensuring services are performed diligently and in accordance with industry standards, applying to both the work of people and AI agents.
3. Service Level Agreements
The SaaS Clause: SLAs typically measure platform availability, offering little reassurance if the agent is operational but making errors.
The BPO-Style Solution: SLAs would focus on operational outcomes rather than mere uptime. Metrics could include accuracy, timeliness, and customer satisfaction rates.
4. Indemnification
The SaaS Clause: Indemnities in SaaS agreements can be narrow, often limited to defending against third-party IP infringement claims.
The BPO-Style Solution: Broader indemnities should cover risks from the AI agent’s autonomous actions, provided these actions remain within the agreed-upon scope.
5. Governance & Audit Rights
The SaaS Clause: Audit rights may be restricted to standard reports, lacking comprehensive transparency.
The BPO-Style Solution: A more extensive governance approach would allow companies the right to transparency, including the maintenance of decision logs for AI actions.
6. Data, IP Rights & Model Training
The SaaS Clause: SaaS terms often grant broad licenses to providers for data generated through the platform, which can lead to training models on confidential data.
The BPO-Style Solution: Contracts should clarify that the company retains ownership of all data and outputs generated, and prohibit providers from using company data for training without consent.
Conclusion: Forging the Hybrid Contract
The future of procuring agentic AI solutions likely involves developing a hybrid contracting model. This model can combine the scalable, subscription-based framework inherent in SaaS contracts with performance and governance commitments from BPO agreements. Such an approach enables both companies and providers to unlock new value in contracts for services delivered through agentic AI.
