The correct option is Stand up a dedicated cross-functional delivery team with product, engineering, data, security and risk and give it a funded charter and an executive sponsor.

This approach moves the organization from prototyping to accountable delivery with clear ownership and measurable outcomes. A cross functional team aligns customer goals with technical feasibility and compliance needs, which is essential for a regulated financial services pilot. A funded charter provides the budget and scope to build production foundations such as data pipelines, CI and CD for ML, monitoring, guardrails and human oversight. Executive sponsorship removes blockers and aligns stakeholders so the pilot can integrate with existing systems and meet risk and security standards.

Hand the entire build to an external systems integrator so you can avoid the complexity of forming an internal team is not appropriate because it sacrifices internal capability building and product learning. External partners can augment the team, yet handing everything over reduces ownership of risk decisions and slows iteration with business stakeholders.

Roll out enterprise AI policies and controls with Cloud DLP and Access Context Manager across all business units before creating a delivery team front loads governance in a way that delays delivery. Enterprise controls are important, however a pilot should establish right sized controls with a product team, prove value and then scale the policies. Tools do not replace accountable product ownership and coordinated engineering.

Immediately create a large centralized AI Center of Excellence to direct every future AI effort company-wide adds heavy coordination before there is a proven delivery path. A centralized function can emerge after early pilots succeed, while the immediate need is a focused product team that can deliver and inform future standards.

Full AWS Practitioner Certification Question