Using an AI app

What to ask an AI app connected to FilmScorecard, how to ask so the answers are good, and where it stops, with example questions for each part of your work.

This guide is for after you have connected one. To connect it, see the page for your app: Claude, ChatGPT, Claude Code, Cursor, or another AI app.

How it works, in one minute

You ask in plain words; the app picks the tool. FilmScorecard gives a connected app 34 tools, one for each thing your screens show. You never name them. The app chooses, calls, and tells you what it found. The whole list is in The 34 tools.
It reads what is there now. Every answer comes from your festivals as they stand at that moment, the same records your screens show, not a copy taken earlier.
It works as you. It reaches exactly the festivals and screens you can open, and a section your role does not show you is left out of its answers too. The What a connected app can do card on your account's Connections tab shows what that means for your role. See Connections.
It changes only three things: a film's status between In Consideration, Selected and Not Selected, your own flag, and Priority. Everything else it only reads, and reading changes nothing: a conversation it reads stays unread. See Reading changes nothing.

Getting good answers

Name the festival and the season. "In Spring Shorts 2026, …" saves the app a guess. When you belong to more than one organization, name that too. The app starts every conversation by finding your festivals, so it can match a name you give it.

Say what you want back. A count, a short list, a table with the columns you care about, or a paragraph you can paste into an email. The app shapes the answer; FilmScorecard gives it the facts.

Narrow in the question, not afterwards. A season's submissions can be narrowed by status, category, your flag, Priority, format and a search on the title or the submitter's name, all in one go: "the Selected documentaries I have flagged" is one request, not a list of every film to sort through.

Big seasons take many calls. Every list comes 25 rows at a time, or up to 50, and the app pages through the rest. A question about every film in a 600-film season is many calls, and each person can make 120 a minute across every app they have connected. Narrow the question, or ask for counts, which the season overview already has. See Lists come a page at a time and 120 calls a minute.

Check before you decide. An AI app can misread or summarise too eagerly. For anything you will act on, such as a selection, a payout or a reply to a judge, open the record on your own screen before you commit to it.

Read the preview before a status move. When you ask it to move films, FilmScorecard first tells the app exactly which films would move and from what, and nothing changes until the app confirms. FilmScorecard asks the app to show you that preview, but it cannot make the app wait for you, so read what the app says it is about to do before you let it go ahead. See Moving films between statuses.

It cannot see files or pictures. Screeners, posters, receipts, import files and exports never reach it. It gets the facts behind them instead: whether a screener link is on file, what an export's columns hold. See What the app is never given.

What to ask it

Each example below is something a connected app can answer with your access. Some need a permission your role may not have; the What a connected app can do card on your Connections tab shows which apply to you.

Reviewing submissions

"List the documentaries in Spring Shorts 2026 that are still In Consideration."
"Show me every film I have flagged this season, with its category."
"Open the film's title and summarise it for me."
"How many submissions does this season have in each status?"
"Move these five films to Selected." It previews first; read the preview before you let it confirm.
"Mark these films Priority so judges score them first."

Judging

"Which judges on the roster haven't filed a scorecard yet?"
"Has any judge asked to leave the panel?"
"List the scorecards filed this week, with their scores."
"Summarise the problems judges have reported this season."
"Which judge invitations are still waiting for an answer?" Needs Manage judges and assignments.
"Summarise the open judge requests and the rates they asked for." Needs Manage judges and assignments.

Awards and announcements

"Rank the entrants for Best Short Documentary by score."
"Which awards have released their ranks?"
"Who hasn't read the last announcement?" Needs Manage judges and assignments.

Judge pay and sales

These need View revenue and payouts on the season.

"How much judge pay is still owed, judge by judge?"
"List the judge payments logged this month."
"What did scorecard sales come to this season?"

Across your organization

"Find judges in our directory who list documentary experience."
"Summarise my latest Inbox conversations." Reading them does not mark them read.
"What's in our scorecard library?"
"List this month's orders." Owner or Org admin only.
"Which promo codes are close to their limit?" Owner or Org admin only.
"What changed in festival access this week?" Owner or Org admin only; it reads the audit log.

What it cannot do

It cannot send anything, pay anyone, or decide anything for you. It cannot email or message a judge, send an announcement, approve or return a scorecard, settle a reported problem, decide an award, log a payment, import films or change a setting. The three changes above are all it makes.

It can write for you, though. Ask it to draft an announcement, a reply to a judge, or a note for your team from what it has read, then send it yourself from the screen that sends it.

The full list is in What an app cannot do.

Giving it Help to read

A connected app reads your festivals, not these guides. When you want it to answer "how do I …" questions about FilmScorecard itself, hand it the guide:

Copy for AI, at the top of every Help guide, copies the whole guide as plain text your AI app reads well. Paste it into the conversation, then ask. It turns into Copied once your clipboard has it.
View as Markdown, beside it, opens the same text in a new tab. Its address is the guide's own with .md on the end, such as https://www.filmscorecard.com/help/using-an-ai-app.md, which an app that can read web pages can fetch for itself.
https://www.filmscorecard.com/llms.txt lists every guide, with its one-line summary and its .md address, for an app that wants to find the right one on its own.

The text is the guide as this page shows it, and nothing more.

Keeping it safe

Only press Allow for an app you just added. If a connection page appears that you did not start, press Deny.
You can switch it off at any time, on your account's Connections tab. Its next request is refused.
Every call is recorded: which app, which tool, and when. See Every call is recorded.
A change it makes is made as you, the same change the Submissions screen makes, and your colleagues see it when their list next refreshes.

Connecting an AI app: the 34 tools, what each needs, what the app is never given, and every sentence it is told when it is refused.
Your account settings: the Connections tab and its What a connected app can do card.
Festival access: the roles and permissions that decide what an app can reach for you.