Trust, transparency and the right level of control
This distinction becomes even more important with implementation of AI components (for example, agents) in these solutions. Organisations should absolutely expect transparency around how agents operate, what data they use, what actions they can trigger and how access controls are enforced. We must consider the levels of governance required – for instance, governance for a conversational agent (low risk) would be different from one that creates transactions (high risk, more controls required). But transparency does not mean producing a detailed design document for every AI agent on a cloud platform.
Similarly, failover and failback requirements, penetration testing obligations and performance testing expectations need to be interpreted through the lens of SaaS. The appropriate assurance and testing approach will depend on the contracted service, customer configuration, integrations and applicable obligations.
These are in many cases governed by the ‘cloud service agreement’, platform certifications/accreditation and service provider commitments, rather than by customer-led testing activity. The enterprise IT responsibility shifts from proving the platform at an infrastructure level to making sure the platform is used, configured and governed appropriately.