# EM-WRK-04 provider comparison

Claude and Grok agree on the primary boundary: profile Portfolio on WM-ACT-029; keep Program as a disjoint sibling; retain component mastership externally; keep PortfolioComponent, Prioritization, InvestmentAllocation and PortfolioScenario as portfolio-owned releases; create one independently identified but identifier-unassigned Selection Criteria Set candidate; and allocate no catalogue or runtime identifier.

Both reviews separate Portfolio from Program, catalogue, Project and Initiative; membership from mastership; criteria definitions from observations and rankings; portfolio allocation intent from budget authority, commitment, accounting actual and component funding; scenario from approved baseline; and strategic alignment from causal benefit claims. Both preserve immutable historical scenarios and criteria pins through rebalance.

Grok sharpens the allocation semantics. Overlapping portfolio shares express investment intent and need not sum to 1.0, because other sponsors may exist and share basis may be capital, cost, capacity or headcount. The share basis must be explicit and cannot default to a fraction of initiative budget. Consolidated actuals, capacity, outcomes and benefits must never be inferred or double-counted from those shares.

Grok also requires explicit profile constraints disabling Program delivery, rollup, relatedness and exclusive-owner semantics; independent identity for released criteria sets; an external WM-ECO-012 oversubscription hook; a scenario-level decision-triad attachment; explicit observation authority and comparability pins; and lifecycle rules for cancelled, split or re-mastered components.

Publication remains held by the identifier-unassigned Selection Criteria Set, unresolved WM-ACT-029 containment and profile semantics, unallocated metric/benefit/capacity/funding referents, unapproved relation rows and non-canonical base drafts. The result is a reviewable allocation-plus-profile dossier with no installability or publication-readiness claim.