MSPs will need to revisit client contracts in the agentic AI era

Agentic AI introduces new contract considerations

MSPs, for example, may be required to disclose the amount of training an AI agent has received and the level of human oversight provided, including how much time might pass before an issue is escalated to a human administrator.

Additionally, clients are likely to want more visibility into, and approval over, the AI models, orchestration tools, monitoring platforms, system integrations, and agent configurations being used.

In addition to defining expected outcomes, MSPs should address oversight requirements along with the guardrails and controls applied to AI agents, advises Mayer Brown.

Defining liability when AI agents make mistakes

At the moment, AI agents generally do not fall under the legal boilerplate in a services contract. Theoretically, a client that allows an MSP to use AI agents is accepting the associated risks. However, both parties will need to revisit indemnities and the treatment of first-party losses in light of the unique characteristics of agentic AI, advises the law firm. These issues include privacy or confidentiality breaches, infringement, regulatory violations, and other harms that may require indemnity provisions.

AI agent errors can also cause significant direct operational losses, including erroneous payments, refunds, transactions, communications, or changes to systems of record. The allocation of high-value losses under applicable liability caps, enhanced caps, exclusions, and other risk-allocation provisions will depend, among other factors, on the degree of control the provider exercises, noted Mayer Brown.

Planning for changing AI workflows and providers

Ideally, contracts should define a process for monitoring and evaluating changes to agentic workflows to ensure compatibility as models, tools, and configurations evolve. That challenge is especially relevant in the AI era, as providers continuously leapfrog one another in terms of capabilities. Potentially even more disruptive, it may not always be clear which AI tools or platforms will remain available following a merger or acquisition.

Any effort to terminate a contract may also need to address how service delivery workflows transfer to another organization, the law firm advises. That might not be an MSP’s first concern until, of course, it acquires a new client looking to migrate away from another MSP.

The terms and conditions of a contract are not always the first thing that comes to mind when considering the implications of adding AI agents to the IT operations behind a managed service. They may, however, become one of the biggest challenges MSPs face in the AI era. There’s no better time than the present to apply an ounce of prevention and avoid a whole lot of AI pain later.