Source: Bain & Company
Individual
Enterprise
Out of the box
Tailored
Personal productivity AI
Centers on unauthorized data use and unvetted tools
Amplified intelligence AI
Tools interact with confidential data and influence user actions in nuanced ways
Embedded assistant AI
AI is embedded in core business systems but typically limited to well-bounded tasks
Digital worker AI
End-to-end automation across systems demands careful design, oversight, and resiliency planning
Governance focus:
Establish acceptable use policies and reinforce boundaries on uploading confidential or regulated data
Promote experimentation through safe sandboxes and internal communities of practice
Train employees on prompt engineering, bias awareness, and hallucination risk
Governance focus:
Integrate knowledge assistants with enterprise identity and access management and permission layers
Require source citation, data lineage tracking, and rejection logic for low-confidence answers
Build user trust through transparency, accuracy thresholds, and feedback mechanisms
Governance focus:
Require AI risk reviews during procurement to evaluate model transparency, data use, and update frequency
Define fallback mechanisms and user overrides for automated suggestions or actions
Coordinate enterprise-wide configuration to avoid fragmented deployments
Governance focus:
Form an AI steering committee to approve use cases, allocate resources, and oversee strategic alignment
Require business cases, risk models, and exception handling design before deployment
Mandate traceability, auditability, and human-in-the-loop controls for critical workflows
Individual
Personal productivity AI
(Chat-based AI/copilots)
Personal productivity AI
(Chat-based AI/copilots)
Embedded assistant AI
(AI embedded within enterprise software)
Risk posture: Low
Risk posture: High
Risk posture: High
Risk posture: Medium
Risk posture: Low
Risk posture: High
Risk posture: High
Risk posture: Low
Risk posture: Low
Risk posture: High