AI tools
Lurno’s AI tools help you build content faster and give learners on-demand support. They fall into four groups:
- The assistant — one conversation reachable from every page, which answers from your own data and proposes changes you confirm. This is the front door; it has its own section.
- Create — the flow that turns a description or a document into a whole course. See Create a course with AI.
- Authoring assist — the ✦ controls inside programs, lessons, and assessments, documented below.
- The tutor — help a learner can call on while working through a question.
Everything on this page is subject to the same three rules: AI drafts, you decide; every AI surface says it is AI; and nothing is written or published until you accept it.
What this does
Section titled “What this does”AI tools sit alongside the work you already do. They draft; you review and decide. Generated content lands as a reviewable draft you explicitly Add, and any change the assistant proposes is shown on a confirmation card before it runs — the model never gets a privileged path to your data. Grades are drafted for review only; releasing them is always a person’s action.
Three properties hold across every AI surface:
It says it is AI. Every AI surface carries a disclosure — “AI-generated. It can be imperfect, so check anything that matters.” Content that was generated is separately marked where people can see it: “AI-generated content. Review before relying on it.” Voice sessions say it out loud as well as on screen. This is a legal obligation under the EU AI Act’s transparency rules, not a style choice, and it isn’t switchable.
It abstains rather than guessing. If the data it read doesn’t support an answer, it says so and offers you somewhere better to look, instead of producing a plausible number. Where it could read only part of a large set, the figure is marked partial rather than presented as exact.
There is one assistant, not a menu of them. Behind the scenes different specialists handle authoring, data questions, document search and operations, and a request may involve several of them at once. None of that is exposed: you never pick an agent, there is no model switcher, and no reply is labelled with a robot’s name. If fewer specialists ran than the question needed, the answer says so.
Before you start
Section titled “Before you start”- Ask Lurno to switch AI on for your organization if it isn’t already.
- You need permission to use AI. Without it the ✦ entry points simply don’t render.
- Each person also has to have AI assistance consent in place — it’s an optional policy in account settings, and withdrawing it turns AI off for that person alone. Speaking to the assistant needs a second, separate consent to voice processing.
- Optionally connect your own model provider key so your organization’s providers are tried before Lurno’s default. See AI provider keys.
Authoring assist
Section titled “Authoring assist”The entry point is a ✦ Ask AI control, and what it generates depends on where you open it.
In a lesson — open a lesson and use Ask AI, available both from the editor header and from the content column below the blocks. A prompt panel opens: describe what the lesson should cover, and optionally open Add specifics to refine count, resource types, difficulty and instructions, and to flip Broader knowledge on so the draft may go beyond the lesson’s own material. Pick a modality — Text, Questions, Image or Audio. Generate, then review each draft and choose Add to insert it or Dismiss to throw it away. Regenerate re-runs a single draft with your feedback.
On an existing block — the ✦ action on a content block opens Adjust with AI. Describe the change, or use a preset: Make harder, Simplify, Add an example, Fix wording, More engaging. You get a Before / After comparison and choose Apply change or Keep original. Audio and image blocks swap the presets for their own configuration, with Regenerate narration / Regenerate image.
On a module — the module’s menu offers Outline with AI, which drafts the module’s child lessons.
On a program — Build modules with AI drafts the program’s top-level modules. It appears in the empty state when the program has no modules yet, and again below the module list once it does.
On an assessment — assessments have one AI surface, Ask AI, with a Generate / Ask toggle: Generate drafts questions into a section you choose, Ask is the copilot described below.
Nothing is inserted, and nothing is published, until you accept it.
Ground AI in your program’s sources
Section titled “Ground AI in your program’s sources”You can give a whole program a set of source documents so every AI generation inside it draws on your material — a syllabus, a policy handbook, a product manual, a set of readings — instead of only general knowledge.
- Upload the documents to Sources in the sidebar and wait until each one’s status reads Ready. Documents still showing Uploaded, Parsing, Parsed or Indexing can’t be attached yet. See Document library.
- Open the program and go to its Settings view. In the Source documents card, choose Add sources and pick the documents to attach, up to the limit the card enforces.
- That’s it — you attach sources once here. From then on, drafting a lesson, outlining a module, or generating other content inside that program is grounded in those documents by default. Authors never have to re-pick the same files in each dialog.
Individual generation dialogs also carry a source-grounding control, so an author can narrow or widen the sources for one generation without changing the program’s default.
Attaching sources to one program doesn’t affect any other program. Editing a program’s sources needs permission to manage the program.
The assistant
Section titled “The assistant”The floating Ask AI about this program button on a program, and Ask AI on an assessment, no longer open their own private conversations. They hand your question to the same assistant the rest of the product uses, pre-scoped to what you were looking at — so a question you asked on a program and one you asked from the search palette are one conversation with one history.
The assistant answers from your own data, read under your own permissions, and cites what it read. Ask it to make a change and you get a proposal card; the change only runs when you press Confirm, through the same gated path a manual edit would take.
Two extra safeguards apply to riskier requests:
- Mass or irreversible changes — bulk withdrawals, enrollment transfers, and adding or editing a prerequisite gate — arm a second button and ask you to Confirm again before running. For bulk enrollment work, the server first runs the change without committing it and shows you the outcome each person would actually get.
- Publishing and archiving a program are four-eyes actions: they don’t run on your confirmation. They queue as an approval request that a second person holding the matching permission must approve. If your organization has no second person who could approve it, the request auto-approves 48 hours after it was raised — it isn’t instant.
Everything the assistant changed is receipted, and most of it can be undone for 24 hours.
Full detail is in its own section:
- Ask the assistant — opening it, scope, attaching documents, what it can do
- What the assistant changed — receipts and undo
- Approvals — Waiting for you — four-eyes changes
- Voice conversations — speaking instead of typing
The learner tutor
Section titled “The learner tutor”Inside the lesson player, learners get an AI help panel on a question with three buttons: I don’t understand, I need a hint, and an adaptive answer button that reads Show me the answer until the answer has been revealed and then becomes Explain the answer. A server-side guard enforces try-first — a learner who hasn’t attempted the question gets coaching instead of the answer key. Every reply carries a disclosure that it was AI-generated and can be imperfect.
Learners also get the assistant itself, as a sheet on their own pages, scoped to their own enrolments and nothing else — their courses, their work, their results. It cannot change anything, and its suggested questions are a learner’s questions (“How am I doing in my courses?”, “What is due next?”), never staff ones.
Access to both is tied to enrollment, not to any staff permission — anyone with an active enrollment in the program can use it, and people who aren’t enrolled are refused even if they’re org members. Two other conditions apply: AI must be switched on for the organization, and the learner must not have withdrawn the AI processing consent in their privacy settings. A learner who has withdrawn it sees “AI help is off. Turn on AI assistance in your privacy settings to use the tutor.”
Using Show me the answer writes an entry to your organization’s audit trail.
Provider keys, priority & usage
Section titled “Provider keys, priority & usage”Integrations in the sidebar lists the external services you can connect, grouped by category; the AI group holds the model providers. The page itself is open to any signed-in member, but each card — including the AI group — only appears for someone holding permission to manage integrations, granted per role; without it the page simply shows nothing to connect. Connect a provider, then Save & test connection — the credential goes into Lurno’s secure vault and the connection is tested; Disconnect removes it. Your organization’s providers are tried before the platform default. Below the provider cards, an Advanced settings disclosure holds Provider priority (rank your connected providers) and AI usage (what your organization has consumed).
Spend budgets are not set here — they’re set per organization by Lurno.
How AI is gated
Section titled “How AI is gated”| Control | What it does |
|---|---|
| Organization AI enablement | Set by Lurno per organization — one switch turns AI on or off for the whole organization. Off means every AI request is refused with a reason, whatever the caller’s permissions. Sub-organizations inherit it. |
| Assistant rollout | A separate, organization-level opt-in that Lurno controls. While it’s off, the Ask button and the assistant page don’t exist for anyone in the organization. |
| Permission to use AI | Shows and allows authoring assist, Create and the assistant. Checked in the interface, again in the handler, and backed by row-level security. |
| Per-action permissions | The assistant’s list of what it can do is filtered by your permissions, resolved server-side, and re-checked when the action runs. |
| Four eyes | Publishing and archiving a program need a second, different person to approve. |
| Enrollment | Governs the learner tutor and the learner assistant. Independent of staff permissions. |
| AI assistance consent | Per person, optional, withdrawable. Withdrawing turns AI off for that person alone. |
| Voice processing consent | A separate policy from AI assistance, because speaking sends your recorded voice. You can accept one without the other. |
| Learner voice | Off by default in every organization and enabled only on request — an unknown learner is treated as possibly a minor. |
| Provider | Your connected provider keys are tried first, then Lurno’s platform default. Voice additionally needs a provider whose account includes realtime. |
| Grounding | A program’s attached Source documents ground every generation inside it. Set once on the program’s Settings view. |
| Budget | A per-organization monthly spend budget set by Lurno; requests are reserved against it and settled after use. Voice has its own allowance. |
| Rate limits | Per person, per kind of request. Exceeding one returns “Too many tries — wait a moment and try again”, not a failure. |
Learners may be renamed in your organization; see Terminology.
Troubleshooting
Section titled “Troubleshooting”AI entry points don’t appear. You’re missing permission to use AI on the active organization. Ask an administrator to grant it. Plan level is not the cause — AI is not plan-gated.
Everything fails with “AI is not enabled for this organization.” AI hasn’t been switched on for your organization. That’s a Lurno-side setting; contact Lurno rather than looking for a toggle.
The tutor isn’t available to a learner. Check that they have an active enrollment in the program, that AI is switched on for the organization, and that they haven’t withdrawn AI-processing consent in their account privacy settings. Staff roles are irrelevant here.
A proposed change didn’t apply. Publishing and archiving a program need a second approver. Look for the Waiting for a second approver state on the card, or open Waiting for you. Mass or irreversible changes need a second Confirm again click.
There’s no Ask button in the header. Your organization isn’t in the assistant rollout yet. That’s separate from AI enablement and separate from permissions — contact Lurno.
It refused to answer rather than giving a number. Deliberate. It abstains when the evidence it read doesn’t support an answer. Take the Look at the reports instead or Search the documents instead offer, or narrow the question.
“Too many tries — wait a moment and try again.” A per-person rate limit, not a failure. Retry shortly.
AI isn’t using my document. Grounding uses the program’s Source documents card on the program’s Settings view, and only documents whose status in Sources reads Ready can be attached. Confirm the document is Ready, then attach it to the program — attaching it to one program doesn’t affect others.
Requests fail or degrade after heavy use. Your organization’s AI budget may be exhausted. The tutor degrades to a safe reply rather than erroring; authoring reports the failure. Budgets are adjusted by Lurno.