An instructor runs a twelve-week course for a cohort of forty. The reading list is nine PDFs. Lectures are recorded because a third of the cohort is part-time. Office hours are dominated by three questions that the week-four reading answers on page eleven, and by students who missed a lecture and want to know what was in it.
A study assistant built from the same material does not replace any of that. It answers the page-eleven questions at eleven at night, which is when students actually have them. Here is how the setup goes.
The syllabus becomes a collection
Individual documents are the wrong unit for a course. A student does not think "which of the nine PDFs" — they think "week six". Collections exist for exactly this.
Upload the readings first, then in the dashboard's Collections tab choose Create Collection. You set a Collection name, a URL slug (lowercase letters, numbers and hyphens), a cover image, and a Description with basic formatting and a live preview. The Generate button will draft that description from the keywords Docutrain extracted from your selected documents during training, which is a reasonable starting point to edit rather than a finished paragraph.
The Documents section is where the syllabus takes shape. Tick what belongs; selected documents move to Selected order on the right, grouped by category. Each category has A–Z and 0–9 quick-sort buttons, and you can drag any document's handle to place it exactly — dragging stays within its category. That order is precisely what students see in the collection's sidebar, so categories named for weeks or units, ordered deliberately, do the navigational work for you.
Under Chat & navigation, Allow chat across all documents decides whether the cohort can query the whole reading list at once. With it on, the collection page's Chat Mode menu offers Chat across all, and students can also choose Select documents to tick a subset — week five and week six, say, before a seminar. A small ? beside the menu opens a built-in explainer contrasting the two, which saves you explaining it. Citations in a multi-document answer name which document each source came from, so a student can always trace a claim back to a reading and a page. Collections has the rest.
Recorded lectures
Audio is a first-class upload type. On the Audio tab of the Train New Assistant window, Docutrain accepts MP3, WAV, M4A, OGG, FLAC and AAC. You give it an Assistant name — suggested from the filename — and click Upload Audio. The recording is transcribed automatically and then trained exactly like a text document. Longer recordings take longer to transcribe, so a ninety-minute lecture is not a thirty-second job; allow extra processing time and let it run in the background.
Two practical notes. Voice training requires the Business plan or higher; on lower plans the tab is locked and shows "Voice training requires Business or higher" with a View plans button. And because a transcript has no page numbers, source page references behave differently for audio-derived content than for a PDF — the citations still work, they just cannot point at a page that never existed.
A lecture assistant is genuinely useful in a way a reading assistant is not: it answers "did she say anything about the exam format?" from material that otherwise only exists as ninety minutes of nobody's time.
Tuning the chat for studying
Three settings change how the assistant feels to a student.
Default chat tone, on the document editor's Behavior tab, sets the starting response style. Chatty gives full-length answers with headings and context — the default, and the right choice for readings a student is trying to understand. Succinct gives brief, often bulleted replies with only the essentials, better for quick lookups and phone screens. Students can switch during a conversation regardless: each finished answer carries a badge naming the style that produced it, and tapping it offers Redo in Succinct or Redo in Chatty, which regenerates that same answer in the other style in place. A student's Succinct choice always wins, even over a document's own custom prompt asking for detail.
Memory, also on Behavior, is off by default and worth turning on for a course. It lets the assistant use a signed-in reader's own earlier conversations with that same document as background context. In practice that means a student who explained their situation in week three does not have to explain it again in week nine, and can ask "what did we discuss last time?" and get a dated summary. Two boundaries keep it honest: facts still come only from the document — memory adds continuity, never new subject matter — and memory spans only that reader's own conversations with that one document. It never crosses documents and is never shared with other readers. It also has no effect for anonymous visitors, which is one reason to have your cohort signed in. Deleting a conversation from Prior Chats clears what the assistant remembers from it.
Key Topics costs nothing and does a lot. It is a ranked list of the most important topics in a document, extracted automatically during processing and shown under the hint "Ranked by importance · click to ask". A student clicks a topic and the question is placed into the chat for them. For a reading nobody has opened yet, it is the fastest way in. It appears when keywords are enabled — the Keywords cloud toggle under UI Options → Features is on by default. More on these in tuning the chat experience.
Access for a cohort, and on phones
Course material is usually neither public nor tied to accounts you control, which makes a passcode the pragmatic choice. Set it at the collection level: in the collection editor's Access section, choose Passcode protected. One passcode unlocks every document in the collection, including documents whose own access level is stricter — the editor warns you about that before you save, and for a course it is the behaviour you want. After a correct entry the passcode is remembered on that device for a long period, so students are not re-prompted every week. Signed-in members of your own organization skip the gate entirely.
If you would rather hand access to named groups and be able to revoke it — a second cohort, an external examiner — Token based access gives you multiple named token links with optional expiry dates and use limits, each individually deactivatable.
Students will read on phones, and there is a tool for that. On an iPhone or iPad in Safari, the chat's tools list ends with Add to Home Screen (shown as Install in the compact chip strip, described as "Keep this one tap away"). It opens a short guide walking through Safari's own steps: tap Share, choose "Add to Home Screen", tap "Add". The installed shortcut opens the chat like an app. Two settings on your side make it look right: your organization's Home screen icon in Owner Settings → Branding, which must be a square PNG of at least 512×512 pixels, and the Home screen name field on the document's Embed & Share → Home Screen sub-tab. Home screens display only about twelve characters, so "Ethics W1–12" survives where the full document title will not. The tool only appears in Safari on iOS, and disappears once the chat is already installed.
Quizzes, and what the numbers tell you afterwards
Generate a question bank from the Quiz tab in the document editor. Docutrain sizes it from the document's length — roughly one question per two sections of content, minimum 10, maximum 100 — and writes each question with an explanation and page citations. Generating requires the Pro plan or higher. Review the bank before enabling it; Edit on an expanded question lets you fix the wording, adjust the options (between two and six) and rewrite the explanation, and saving creates a new version while retiring the old one, so a student's past attempt still shows the question they were actually asked.
The Quiz toggle under UI Options → Engagement is what makes the button appear. From the student's side, selecting Quiz opens a How many questions? window offering 5, 10, 15 or 20, with 10 preselected; lengths larger than the bank are not offered. Questions are drawn at random from the bank, so two attempts rarely match. Selecting an option locks it immediately — grading happens on Docutrain's servers, so the answer key never reaches the browser before a student commits — and feedback appears right below: Correct or Incorrect, an explanation written from the document, and Page citation chips you can hover to see the source passage.
The review screen at the end is the part that makes this a study tool rather than a test. It shows the score as a fraction and a percentage with a short message (Perfect score!, Great job!, Good effort!, Keep practicing!), a Missed only checkbox that filters to the questions you got wrong, and a card per question showing your answer, the correct answer, the explanation and the same page chips. Retake Quiz draws a fresh selection. There is no attempt limit. Full detail in quizzes.
Afterwards, the numbers. Each document's Quiz Statistics panel reports total attempts split into authenticated and anonymous, average score, and highest and lowest results. Each document's Intelligence tab — unlocked once it has received at least 5 questions — reports questions, conversations, questions per conversation, an estimate of distinct askers, satisfaction from thumbs ratings, and question volume over time, with a Keyword density panel counting the exact phrases students type. The organization-wide Intelligence tab adds a By document table listing every document with its Questions, Satisfaction and Last activity, so you can see at a glance which reading the cohort actually engaged with and which one nobody opened.
Pair the two readings. A reading with heavy question volume and low satisfaction is hard; a reading with no questions at all was skipped. A quiz average that is low on material the cohort should know often points at the source text rather than the students. Generating an Intelligence report at the start and end of a term gives you two comparable snapshots of the same cohort — see document intelligence for what those reports contain.