Deliverables Are "Works of Art", PMs Are "Curators"

Transcript

1. 2026.2.17 Deliverables Are "Works of Art", PMs Are "Curators" ~Managing Context in Alignment with Your Project Phases~
2. About Me Shinya Ema PO Lead, Business Design Division After working as an engineer and PdM at a major manufacturing company, I now support POs and lead strategy design in agile development environments. I'm also working on streamlining product development processes through AI adoption. 2
3. Agenda # Topic 1 Deliverables Are "Works of Art", PMs Are "Curators" 2 Curation Changes by Phase (Real Experience) 3 Practical Operations: Context Management with Google Drive as a Hub 4 How AI Changed the "Meaning of Organizing" + Summary 3
4. 1. Deliverables Are "Works of Art", PMs Are "Curators" The Museum Metaphor
5. Your Project Is Full of "Works of Art" Meeting notes, design documents, code, Jira tickets, Slack threads, Miro sticky notes, PowerPoint decks, voice transcriptions, screen mockups… Every business deliverable born from hands-on work and discussion. These aren't mere "documents." They are the trail of decisions and creation in a project — in other words, "works of art." 5
6. PMs Are "Curators" Curator = A person in a museum who collects, selects, and arranges works, giving them context to maximize their value. The word originates from the Latin "curare" (to care for). PMs do the same. How you gather, arrange, and contextualize your business "works of art" determines the quality of your project's decision-making. 6
7. And Now, the "Ultimate Audience" Has Arrived In the past — no matter how neatly you organized things, nobody read all of it. Honestly, it often ended up as self-satisfaction. Now — AI reads everything without missing a thing. The first line of meeting notes, the footnote on the last page — it reads it all. 100 slides, 500 lines of code — it sees it all. It understands context, connects the dots, and makes suggestions. The cleaner you organize, the better AI's output quality becomes. The meaning of organizing has never been greater. 7
8. 2. Curation Changes by Phase Real Experience: From Project Execution at KAG
9. However, "How to Organize" Changes by Phase The volume and nature of works differ between the early and late stages of a project. Just as curation differs between a museum's storage and a special exhibition, your approach to context management should also change according to the project phase. If you stick with the early-stage approach, AI's responses start to drift after two months. It revisits abandoned ideas. Old and new information get mixed up. → This is the "context contamination" problem. 9
10. Overview: 4 Phases and the Focus of Curation Phase Museum Analogy Curation Focus ① Pre-Kickoff Collecting works (gather everything) Collection ② Direction Setting Choosing the exhibition theme Collection + Refinement ③ Feature Decision Selecting works to exhibit Subtraction ④ During Development Rotating exhibits during the run Ongoing Maintenance 10
11. Phase ① Pre-Kickoff — Collect Works Relentlessly Project situation: Assigned to a project right after joining KAG. No time before kickoff. What I Did Sales history, proposal documents, Slack, internal Wiki → Collected everything I could find Fed it all into NotebookLM → Caught up by "listening on the go" via audio In this phase, don't be afraid of "putting in too much." Grasping the big picture is the top priority. In museum terms, this is the stage of "putting all the works into storage first." Selection comes later. 11
12. Phase ② Direction Setting — Deciding the Exhibition Theme Project situation: Current state analysis → Vision formulation. Deciding the broad scope of "what to build." "Put everything in" is still valid. But once the direction (theme) is decided, start consciously storing away works that don't fit the theme. 12
13. Phase ③ Feature Decision — Selecting Works to Exhibit Project situation: Feature identification and prioritization. What to do is being finalized. The Problem That Arose Context Contamination Ideas that were decided as "not doing" = works that should have been stored away were still lingering in the exhibition room (context). AI was "viewing" those too, returning unfocused interpretations. In museum terms: It's an Impressionist exhibition, but contemporary art has crept in from the storage room. 13
14. Phase ④ During Development — Rotating Exhibits During the Run Project situation: Agile development. Plans and feedback keep evolving every sprint. Museums rotate their exhibits even during a running exhibition. Context is the same. At each sprint boundary, take stock of your inputs too. Keep the exhibits fresh. 14
15. Practical Operations Context Management with Google Drive as a Hub
16. Current Context Management Architecture 16
17. 3-Layer Structure: Collection → Structuring → Injection Layer What We Do ① Collection Aggregate data from various tools into Google Drive (manual export + partial automation) ② Structuring Organize using Drive's folder structure (this has some cost) ③ Curated Injection Select information suited to the purpose and phase, then feed it to AI (this is where the biggest impact is) 17
18. Why Google Drive as the Hub? Reasons for Choosing It Almost every tool can export to it (versatility) Folder structure naturally supports classification by phase and category Native integration with NotebookLM and Gemini (drag & drop injection) In Museum Metaphor Terms Each tool = Atelier (where works are created) Google Drive = Storage room (where works are preserved and organized) AI = The audience in the exhibition room (where curated works are viewed) 18
19. 3. How AI Changed the "Meaning of Organizing"
20. Before AI: Organizing Tended to Be "Self- Satisfaction" Carefully written meeting notes → Nobody re-reads them Neatly organized Confluence → Never searched Time-consuming design documents → Skimmed during review "If nobody's going to look at it anyway, maybe good enough is good enough…" The motivation to organize was weak. 20
21. After AI: Organizing Directly Translates to Value 21
22. After AI: 3 Changes ① An "Audience That Reads Everything" Has Arrived Humans don't read all 100 pages of a document. AI does. For the first time, an entity has appeared that fully receives the benefits of organizing. ② Organizing Has Become an "Investment" The cleaner you make it, the more AI's output quality proportionally improves. The more you do, the more value it creates. ③ Mental Well-being Stabilizes "I'm providing solid information → I should get solid answers." This sense of assurance also improves the quality of a PM's decision-making. *Personally, I also find peace of mind in the state of things being well-organized itself. 22
23. Furthermore, AI's Output Is Also a "Work of Art" Receiving curated collections of works (context), the proposals, summaries, and deliverables AI produces — these are also "works of art." Gather works (deliverables) Curate (design context) Have AI view them New works (output) are born Running this cycle is the PM's job in the AI era. 23
24. 4. Summary
25. It's Not About Technique — It's About Creating the Right Environment Curation (context management) isn't just about techniques to get better answers from AI. When information is well-organized, you can make decisions with confidence. And you can trace the rationale behind those decisions. For PMs, it is the act of creating an environment where you can make decisions with confidence. 25
26. Next: What I Want to Do Going Forward Current Challenges Curation (② Structuring & ③ Injection) is person-dependent Manual information in/out doesn't scale Target State Automate collection from each tool → Google Drive Develop guidelines for curated injection tailored to phase and purpose Build a system where every team member feels that "organizing works means AI leverages them" Still a work in progress. If you'd like to think about this together, let's talk! 26