Before you can compose an app, you need the building blocks: SharePoint lists hold the data, views shape it, and skills act on it.
Takeaway 1
Lists are the database tables of SharePoint. The AI can create them, fill them, and read them.
Takeaway 2
Views are pre-saved filters and groupings — they turn one list into many "screens" without copying data.
Takeaway 3
Skills act on lists. Their trigger phrases turn natural-language utterances into structured operations.
The SharePoint List as Data Store
A SharePoint list is a structured table — columns with types, rows with values — that the AI can both create and query. In the time-off demo the author tells the agent "create two lists, one for time-off requests and one for the team calendar," and the agent builds both. No schema designer, no SQL.
Lists are the database tier of any SharePoint AI app. Define them in plain English, let the agent build the columns, and you're ready to start populating data.
Once a list exists, every other piece — calendars, views, skills, dashboards — references the same rows. The list is the single source of truth. Changes propagate everywhere automatically.
A SharePoint list is not "just a spreadsheet." It is a typed, permissioned, versioned data store with formal views, triggers, and a query surface the AI knows how to address.
Views — Slicing One List Into Many Screens
A view is a saved configuration over a list: which columns appear, how rows are filtered and grouped, and how they render (grid, calendar, board). The same list can have dozens of views — "by status," "upcoming," "by department" — without duplicating data.
In the demo, the time-off request list ends up with a "by status" view and an "upcoming" view. The team availability list gets a calendar view and a "by department" view. None of these are separate lists. They are filters and renderers over the original data.
Build the list once. Add views as the use cases appear. Each view is cheap, instantly editable, and stays in sync because it reads from the same rows.
Sample Data — Why You Ask for It Upfront
An empty list is hard to demo and hard for the AI to reason about. Asking the agent to generate sample data — invented people, plausible dates, a mix of statuses — makes the next steps work immediately. You can see views populate, skills produce realistic answers, and edge cases like conflicts surface.
Sample data is a first-class build instruction. Include "and create sample data so I can see it working" in your initial prompt — the agent will invent realistic rows that exercise every view and skill.
"Real users will give me real data, I'll skip the samples." You'll skip the feedback loop too. Without rows, you can't test views, can't validate skills, can't catch missing columns until production.
Skill Triggers and Utterances
A skill trigger is the phrasing pattern that calls a skill. Once a skill exists, any matching utterance — "request time off for March 17th," "show me pending requests," "who is out next week" — invokes it instead of a free-form chat answer.
In the demo, after the Ralph loop runs, the agent reports back with the prompts that trigger each skill. They are written in user-facing English, not technical syntax. The triggers are the skill's API surface.
Triggers are conversational. Design them like you'd write a help-desk macro — phrases your users will naturally say when they want this work done.
"The user has to memorize the exact trigger phrase." No — the model matches intent, not regex. As long as the utterance is close in meaning, the right skill fires.
Flashcards — Foundations
01 · Foundations
SharePoint List
tap to reveal →
A typed, permissioned, versioned data table inside a SharePoint site. The AI can create, populate, query, and modify it from plain-English instructions.
← tap to flip back
01 · Foundations
List View
tap to reveal →
A saved filter + grouping + render mode over a list. The same list backs many views — by status, by department, calendar, upcoming — with no duplicated data.
← tap to flip back
01 · Foundations
Sample Data
tap to reveal →
Synthetic rows the AI generates so views and skills have realistic content from minute one. Ask for it in your build prompt — don't wait for real users.
← tap to flip back
01 · Foundations
Skill Trigger
tap to reveal →
The phrasing pattern that invokes a skill. The model matches intent, not exact wording — users can say "request time off" or "I want to take March 20th off" and hit the same skill.