AI Assistance

The forms and assessments lists each have a generate-with-AI action, and the workspace Themes page has one for custom themes. Describe what you want — topic, audience, number of questions, tone, or for themes the colors and mood — and a complete draft is created in the workspace, ready to open in the builder. For assessments you also pick the type: a standard assessment with mixed question types, or a coding assessment whose problems come with starter code, a reference solution, and runnable test cases.

The result is a normal form or assessment: every input, question, option, correct answer, and point value can be edited, reordered, or deleted like anything built by hand.

The more specific the prompt, the better the draft. Mention how many questions you want, which styles, the difficulty, and what must be covered — “a 20-question quiz on basic SQL, mixing multiple choice and fill-in-the-blank, moderate difficulty” beats “an SQL quiz”.

For assessments, the generator sets up gradable question types complete with correct answers and points; for forms, it picks input types that fit the goal — ratings for satisfaction, choices for preferences, text for open feedback.

Attach source material and the generation works from it — a lecture PDF becomes a quiz about that lecture, a product spec becomes a targeted feedback form.

Up to 5 files per generation, at most 5 MB per file and 20 MB in total.

On the grading screen of an assessment submission, Grade with AI suggests a score and a short justification for every question awaiting manual review — essays, code writing, and other open-ended answers. When a question has a grading rubric, the AI grades against it.

Suggestions only prefill the grading screen — nothing is saved until you review the scores and submit the grading yourself, and every suggested score can be adjusted before that. Each run uses one AI generation from the monthly allowance.

Every question also has its own AI review action: the AI looks at the answer and the score you assigned, then either backs it or suggests a different one with a short justification. The score itself is never changed. Each per-question review uses one AI generation.

Every plan includes a monthly allowance of AI generations — 5 on Free, 100 on Professional, and 1,000 on Enterprise. Generations that fail aren't counted against it.