People and responsibility · proposed working roles
Shared context. Explicit responsibility.
AI assistance can support the work. People remain responsible for business intent, evidence, decisions, authority and the service that runs afterwards.
Agree responsibilities before introducing new titles
The existing CAE roles article treats roles as an editorial recommendation, not a validated taxonomy. Organisations can assign these responsibilities to existing jobs, and one person may hold several roles. This page proposes an engagement responsibility model; it does not redefine the book or establish a professional standard.
Outcome Engineer, Nexus Outcome Engineer and Contextual Outcome Engineer are naming options requiring the product owner’s decision. Outcome Engineer also has existing uses outside CAE. No canonical CAE definition of Nexus Outcome Engineer was established in the inspected records. The descriptions below are options for discussion, not approved title definitions.
| Role or proposed responsibility | Problem and decisions | Inputs; appropriate AI collaboration | Client ownership and escalation |
|---|---|---|---|
| Outcome Engineer — proposed engagement responsibility | Connect the requested change to an observable outcome; maintain the baseline and acceptance criteria. Recommend which hypotheses to test. | Outcome brief, frontline evidence, quality and cost measures. AI can draft measures and compare evidence; the engineer validates definitions. | Sponsor owns the outcome and spending; domain owner accepts the workflow. Escalate disputed targets, missing baselines and uncontrolled consequence. |
| Nexus Outcome Engineer — unresolved naming option | If retained, coordinate dependencies and hand-offs across domains rather than creating a new decision authority. | Dependency map and domain contracts. AI may identify inconsistent definitions for owners to resolve. | Domain and service owners retain authority. Escalate cross-domain conflicts to the sponsor and architecture governance. Confirm naming and remit before use. |
| Contextual Outcome Engineer — proposed naming option | If retained, connect enterprise context, evidence and bounded execution to outcome design and verification. | Context map, policy versions, system contracts and evaluations. AI drafts alternatives; a human verifies context and feasibility. | Does not inherit the client’s authority to deploy or approve business effects. Escalate conflicting sources, unclear delegation and unsafe execution paths. |
| Enterprise architect | Keep the target coherent with the wider enterprise and existing investment; recommend options and record architectural trade-offs. | System inventory, standards, dependencies, constraints. AI helps compare alternatives with cited assumptions. | Client architecture governance accepts exceptions. Escalate boundary, portability, tenant isolation and recovery gaps. |
| Domain and process owner | Establish how work actually happens; decide rules, exceptions and accepted service outcomes within delegated authority. | Walkthroughs, policy and operational cases. AI helps draft a process map; staff validate it. | Client owns business rules and allowed effects. Escalate contested rules, safety concerns and changes beyond delegation. |
| Product owner | Maintain priorities, user needs, increment scope and acceptance evidence. | User research, backlog, outcome brief. AI organises feedback and drafts acceptance cases. | Sponsor controls funding; service owner accepts operations. Escalate scope changes, unmet accessibility needs and conflicting priorities. |
| AI and software engineer | Implement and test permitted behaviour, integrations and deterministic controls; report technical limits. | Approved architecture, API contracts, test data and threat model. AI can draft code and tests under review. | Client approves access and releases. Escalate untested provider behaviour, secret exposure, bypass paths and failed acceptance criteria. |
| Security, privacy and risk specialists | Assess threats, permitted information handling and risk treatment; recommend conditions for release. | Data inventory, access model, applicable obligations and test evidence. AI helps organise issues, never declares compliance. | Accountable client officers accept residual risk and legal positions. Escalate unknown processing basis, excessive access or unresolved high-consequence exposure. |
| Operations and service owner | Own support, monitoring, capacity, recovery, incident handling and ongoing costs. | Service objectives, runbooks, recovery tests and cost attribution. AI summarises incidents with provenance. | Client names the service owner and funds support. Escalate missing ownership, unbounded demand or failed recovery. |
| Executive sponsor | Set intended value, resolve cross-domain priorities and authorise the agreed engagement and subsequent investments. | Outcome evidence, options, risk and cost ranges. AI can prepare decision briefs; the sponsor decides. | Client retains commercial and business accountability. Escalate matters beyond the sponsor’s mandate to the appropriate governing authority. |
Assistance, approval and action are different responsibilities
- AI assistsOrganises permitted evidence, drafts summaries and proposes options. Outputs remain unverified until reviewed.
- People review and decideValidate sources, resolve uncertainty and approve a specific scope, design or action within their authority.
- Controlled systems executeA separately implemented execution path checks identity, permissions, current policy and the approved action before an effect occurs.
- Owners examine the evidenceReview outcomes, incidents, exceptions, cost and recovery. Revise or stop the capability when evidence requires it.
The diagram is a proposed responsibility boundary. This informational release does not implement the execution system or an AI workforce. A generated recommendation cannot approve itself, and a reviewer’s approval must not be reused for a changed action.
Reuse the published work, preserve the manuscript
The CAE article library provides the public reading route. Related roles and book work exist in the canonical project; their source status and actual live availability must be checked independently. No unpublished chapter prose, companion implementation or confidential customer example is reproduced here.
YOUR ENTERPRISE. YOUR NEXT CHAPTER.
Turn complexity into
a clear next move.
Bring a business question. We can discuss a discovery scope and the decisions it should support.
Start a discovery conversation