Agentic AI Reshapes Agency Contracts and Enterprise Operations
The rise of autonomous “agentic AI” is shifting agency contracts from software-as-a-service (SaaS) models toward service-based agreements that focus on quality and optimization, according to legal analysis from Mayer Brown. Major enterprises are already scaling these operations, with Cognizant using Google Cloud for its agentic AI initiatives and Unilever partnering with the same provider. Experts note that as these systems move into production, purpose-built infrastructure and robust safety guardrails are becoming essential.
- Agentic AI systems differ from traditional software by autonomously planning and executing multi-step tasks to achieve a goal, rather than simply suggesting actions like a "copilot." This capability allows them to replace entire workflows, such as processing invoices or managing customer inquiries, fundamentally changing the value proposition from user-driven tasks to automated outcomes. - The legal firm Mayer Brown suggests that as AI shifts from a passive tool to an autonomous agent, the contracting model is moving from a Software-as-a-Service (SaaS) framework to one more aligned with Business Process Outsourcing (BPO). This new hybrid model requires clearer definitions of services, outcome-based performance metrics, broader audit rights, and stricter governance to manage the risks of autonomous actions. - Unilever's five-year partnership with Google Cloud aims to create an "AI-first digital backbone" using tools like Vertex AI and Gemini. The collaboration focuses on developing "agentic commerce," where intelligent systems guide consumer discovery and shopping experiences for brands like Dove and Hellmann's. - Cognizant is operationalizing its Google Cloud partnership by deploying Gemini Enterprise internally to enhance productivity before offering commercial services. The company established a dedicated Gemini Enterprise Center of Excellence and uses an "Agent Development Lifecycle" to create repeatable, scalable AI solutions for clients in areas like contact centers and order management. - A key challenge in deploying agentic AI is integration with legacy enterprise systems like ERP and CRM, which can lead to data silos and compatibility issues. Other major hurdles include ensuring data quality, addressing security vulnerabilities from autonomous actions, and navigating the ethical "black box" problem where AI decision-making is not easily explainable. - The rise of agentic AI is forcing a shift in software pricing models away from traditional per-seat licenses. Because value is created by autonomous workflows rather than human interaction with an interface, new pricing must account for variable, consumption-based costs tied to model inference and data processing. - In 2024, Unilever initiated a broad AI strategy, training over 23,000 employees on generative AI and launching more than 500 AI projects globally. These projects have been applied to improve social media engagement and reduce waste in manufacturing facilities. - Widespread enterprise adoption of agentic AI faces several barriers, including the complexity of managing interactions between multiple agents, the risk of "agent sprawl" leading to operational chaos, and the potential for AI to amplify biases present in training data. A late-2024 study found that 53% of technology leaders view security as the primary challenge.
