Disputes risk if service contracts fail to specify AI-related customer dependencies

In a new report that explores the implications for financial services firms of suppliers’ use of AI tools to deliver critical services, Pinsent Masons warned that broad and imprecise wording around customer dependencies in an AI context – particularly around data – might enable suppliers to claim that the firms, not them, are to blame for AI-related deficiencies in contractual performance.

The report highlighted how it is common for a services contract to detail the inputs, resources, decisions and access rights that the customer must provide for the supplier to perform the services. It is critical to get those provisions right, since failures in customer-provided information or cooperation can materially impact delivery, service levels, timelines and cost, as well as potential disputes.

Matthew Godfrey-Faussett and Mhairi Mival of Pinsent Masons said suppliers’ preference is for broad customer obligations, which can cover matters such as timely decision-making; provision of accurate data; access to systems or premises; availability of customer personnel; approvals; integrations with third party systems; hardware or software prerequisites; and compliance with security procedures. In contrast, customers will generally seek to resist such broad dependencies, since the broader the dependency list, the more likely that the supplier will blame the customer for performance issues, triggering potential disputes, they said.

As a result, customer dependencies are often heavily negotiated, with the use of AI by suppliers in service delivery only serving to increase the importance of getting the terms right at the outset, according to Godfrey-Faussett and Mival.

Godfrey-Faussett said: “The introduction of AI into the supplier’s delivery model can have an impact on how customer dependencies are drafted, monitored and enforced. AI performance depends heavily on data quality, system context and human oversight, meaning that the customer’s inputs can take on greater significance compared to traditional services arrangements.”

“One significant implication is the heightened importance of data-related dependencies. AI-enabled services will often rely on customer data either for processing and/or training purposes, meaning it is important to define clear dependencies around data quality and completeness. This is an area where customers are likely to resist broad dependencies – for example, to provide ‘all necessary data’ – in favour of more specific obligations. Suppliers are likely to place the onus on the customer via dependencies to ensure that the customer’s privacy policies allow data to be shared with the supplier so that AI can be used as part of service delivery. However, AI systems should be designed to handle imperfect real-world data. A supplier that requires perfectly clean data as a precondition to performance has effectively shifted the risk of AI underperformance onto the customer,” Godfrey-Faussett added.

Mival said: “Well defined customer dependencies are essential to ensure accountability, manage risk and maintain service quality throughout the life of a contract. AI-enabled delivery models will increase the importance of data quality, meaning it is important to have a clear, proportionate framework for allocating responsibility between the customer and the supplier, and mechanisms to address failure to meet agreed dependencies.”

Access the Pinsent Masons report on AI-enabled service delivery