Artificial intelligence (AI) is poised to transform the infrastructure of health care. AI can now interpret clinical conversations and automate back-office operations, and will soon be able to deliver clinician-grade care under the direction of a clinician. This model holds particular promise for primary care, where workforce shortages and rising chronic disease burden demand scalable, integrated solutions. A key barrier to adoption is that U.S. reimbursement is not designed for clinical AI agents. Time-based billing structures penalize physicians for using AI tools that enhance productivity. Traditional transaction-based payment models risk misalignment with care delivery. And without guardrails, added AI workforce capacity can inflate utilization and cost. Current payment models risk bypassing physician oversight of AI services, fragmenting care, and undermining integration with value-based systems. The authors propose a payment framework that aligns incentives around clinical AI agents by reimbursing for care delivered through validated workflows rather than per software license or time spent. Payers would reimburse physicians for outputs of care, enabling them to invest in AI tools and, over time, build the foundation for linking payment to measurable health outcomes. This payment architecture keeps AI-delivered care anchored in physician responsibility, preserving accountability while enabling innovation. When combined with the traceability of digitized AI workflows, this approach lays the groundwork for a system that scales care while preventing fraud and misuse.