Graham's smarts (AI features)
The AI throughout SortedResearch is Graham. He works in a few specific places, and every Graham feature degrades gracefully: if he's briefly unavailable, the platform falls back to a sensible non-AI behavior and nothing is ever blocked. Graham's group-naming, quality checks, screener import, and insight reports are paid-plan features (Individual and up); the Free plan gets one view-only insight report and Graham chat.
| Graham feature | Where | What it does | If Graham is unavailable |
|---|---|---|---|
| Group-naming (participant) | End of a participant's sort | Suggests names for groups they created, which they can edit | Falls back to color names ("Blue group") to edit |
| Group-naming (researcher) | Suggested groups / 3D view | Names all your clusters in one click, each with a distinct name | Uses the most common participant label, or "Group N" |
| Category quality | On sort submit | Flags gibberish/placeholder group names as suspect/rejected | Keeps a basic profanity/gibberish check |
| Effort quality | On sort submit | Judges whether a response shows real effort or just dumping | Uses timing/behavior signals only |
| Screener import (paid) | Screener tab → Import with Graham | Reads a .docx/.pdf screener into a reviewable draft | Feature unavailable (needs Graham) |
| Insight report | Results → report | Writes the plain-English report | Falls back to a facts-only template |
Notes:
- Quality and effort flags are signals, never blockers. Participants always complete; you decide what to include in analysis.
- The insight report is built for speed and reliability; transient failures are retried automatically before falling back to the template. When it does fall back, the report says so in a banner and offers a retry, so a basic version is never passed off as the finished article.
Graham, your assistant
Graham (the lion in the corner of every page once you're signed in) is your in-app assistant. He can answer how-to questions, explain what's on your screen, and help interpret your study's results. When you're on one of your studies, he reads that study's setup and aggregated results so he can talk about your data specifically, but he never sees individual participants' personal information, and he doesn't make changes to your account. He advises, you act.
For tricky results questions he'll bring in his sharper judgment automatically. And when he can't help, use the Message the team link in the chat to send your question straight to a person, who replies to your account email. Graham doesn't appear on participant pages.
Graham also remembers your workspace: he knows your studies and picks up the thread of past conversations, so he feels like a partner who knows your work rather than starting from scratch each time. On a team workspace this memory is shared across the team, and it always stays inside your own workspace. It is never shared with other accounts, and Graham can only ever see your workspace's data.