AI Agents Shift MSP Accountability: Howard Cohen on Liability Beyond IT Infrastructure

The episode highlights a structural shift from MSPs managing infrastructure to supplying, designing, and maintaining AI-driven agents, raising new questions of accountability and operational risk. As AI agents evolve from assistive chatbots to supervised and potentially autonomous systems, the channel faces liability transfer, governance gaps, and an increased need for systems architecture competence. Companies referenced include Klarna, which serves as a cautionary tale for poor AI design, and vendors such as OpenAI, Anthropic, and Microsoft, all of whom are engaged in moving the market toward agent-based operations.

The most consequential development detailed is the shifting liability for AI-driven outcomes: agent builders and MSPs become responsible for unintended actions, errors, or hallucinations produced by deployed agents. Clarifying accountability is necessary as incidents—such as email mishandling or unauthorized decisions by AI agents—do not absolve the MSP of responsibility. Recent discussions indicate few cases where foundational technology vendors are held liable; usually, the burden falls on those who deploy and support AI agents for clients. The episode cites Klarna’s experience as a failure of design thinking, emphasizing that the design of agents—beginning with the end in mind—is key to mitigating risk.

Supporting developments include the segmentation of AI solutions across SMB, mid-market, and enterprise clients, with complexities scaling as MSPs attempt to transition from simple assistive AI to supervised and fully autonomous agents. The episode notes that fewer than 5% of deployed agents are fully automated, and security vendors are increasingly involved in AI governance, risk, and compliance (GRC) due to the importance of data governance in AI projects. Regulatory coverage and insurance gaps are recognized, with advice for MSPs to re-examine their E&O policies and move toward frameworks for AI trust and transparency.

Operational implications for MSPs and IT service providers are concrete: providers must reconsider contract exposure, review insurance coverage, and invest in AI governance mechanisms such as agent oversight and auditing. Price-to-value methods are recommended over simplistic per-agent or per-hour billing, requiring sophisticated project scoping and market analysis. The episode underscores that MSPs cannot rely solely on vendor solutions for risk mitigation—service providers are ultimately accountable for AI outcomes delivered to clients, necessitating operational safeguards and human-in-the-loop design wherever possible.

Supported by:
ScalePad
Zero Networks