Your first month with ArthurAI™.
Four weeks. A practical onboarding for educators new to the platform. By the end of week four you've run an AI-assisted course, attested AI-suggested grades, and built a workflow you actually trust.
- Week 1
Account setup, the LCP, and your first AI-generated course.
Your institution provisions your account through their admin portal — you log in with single sign-on or your institutional credentials. You take the LCP yourself first; we want educators to experience the diagnostic before they hand it to learners. You generate one AI-assisted course as practice. Don't ship it to students yet — read every step. Get the rhythm of how the AI proposes and you decide.
- Week 2
First lesson with the AI tutor live in the room.
Pick one course. Generate the lessons for the week. Review every step before you publish to learners. Run the class. The AI tutor handles the long tail of student questions; you handle the close-loop instruction. At the end of the week, look at the tutor logs — not the conversations themselves (we never log conversation content), but which lesson scopes generated the most tutor questions. That's signal.
- Week 3
First round of assessments and analytics.
Quiz sessions track every attempt. The AI suggests rubric-aligned scores on short-response items with citations to the student's text. You read each one. You attest. The first time you override an AI-suggested grade, write down why — the pattern of overrides is your signal of where the AI's rubric application doesn't match your judgment. Use it to refine the rubric for next round.
- Week 4
Adjusting based on what the data tells you.
By week four you have a week's worth of cohort data: who's on pace, who's stuck, which lessons generated the most tutor questions, where your overrides clustered. Adjust the next week's lesson scope. Adjust the rubric. Adjust the pacing for the trailing students and the stretch material for the leading students. The platform gives you the signal; you make the call.
- Month 2 onward
The workflow becomes muscle memory.
By the end of month two you stop reading every AI-generated step in detail — you scan, you spot-check, you attest. The trust is calibrated to where the AI is reliable and where it isn't. You know which lesson types it nails and which still need your hand. You know when to lean on the tutor and when to teach in the room. The AI never replaces you; it gives you back the time the long tail used to take.
- "Will the AI ever publish a grade without me?" No. Every grade in the gradebook is educator-attested. AI is decision support; the educator decides.
- "What does the AI know about my students?" What the institution has loaded into the platform — name, role, enrollment, LCP results, lesson progress, assessment artifacts. Conversation content is never logged. See data handling.
- "What happens when the AI gets it wrong?" You override. The platform is built so override is the easy path. We track the pattern of overrides at the system-prompt level so administrators can refine the prompts; we never use student data to train models.
- "What do I tell parents / faculty / management?" The disclosure language is configurable per institution. Use the language your institution adopted. See disclosure surfaces.
- "What if I don't want to use AI for a particular thing?" You don't have to. The AI is decision support across the workflow; you can review-and-attest more lightly on the surfaces where you trust the AI, and override completely on surfaces where you don't.