Phase 1
No-stack-change improvement pass
Tighten the workflow that already exists. Standardize Lightroom import metadata and AI usage, audit Pixieset proofing/download/invite settings, and decide whether Studio Manager should absorb more admin work.
Expected result: Fastest low-risk improvement path.
Phase 2
One real culling trial on one real event
Test Narrative if the pain is mostly culling. Test Aftershoot if the pain is culling plus repetitive editing. Measure hours saved, not vendor vibes.
Expected result: You learn whether AI truly saves labor in your actual lane.
Phase 3
One AI editing layer only if Phase 2 helped
If the first trial succeeds, decide whether Aftershoot or Imagen earns a place. The bar should be speed plus consistency, not speed alone.
Expected result: Scalable editing without sacrificing the brand look.
Phase 4
Agentic content and ops layer around delivery
This is where a custom assistant can really compound value, by turning finished galleries into emails, follow-ups, recaps, social posts, and internal task checklists.
Expected result: Every delivered gallery becomes both client work and marketing fuel.