The lesson is one hour. The invisible second hour is everything around it: deciding homework for each student, checking whether they did it, remembering that Marieke confuses essere and stare while Tom’s problem is the plural, keeping some record of who is where. For a freelance language teacher that second hour is unpaid, unloved, and (this is the point of this article) almost entirely automatable, without automating a single minute of the actual teaching.

I write this wearing both hats. I teach Italian privately between Ede and Amsterdam, online and in person, and I built amo Italian partly because I wanted these tools for my own students. What follows is the setup I actually use, described honestly, limits included.

What should stay human, and what should not

The division of labour I defend: explanation, conversation, correction of real writing and speaking, motivation: human, always. Assigning practice, collecting completion, logging errors, scheduling review: software, always. The research case is scattered through this series: practice between lessons should be spaced and retrieval-based, which software schedules better than memory ever will; while the feedback moments that change a learner, Swain’s noticing-the-gap in live conversation (the output article), do not survive automation. An app that promises to replace the teacher is lying to one of you.

Apps for Italian Teachers: Automate Admin, Not Teaching

The class-code setup

In ti’s teacher mode the mechanics are deliberately boring. You create a class and get a class code; students enter it and are linked to you. From there:

  • Assignments by tappa: you assign specific stages of the course as homework (“before Thursday: the two tappe on the imperfetto”) and see who completed what, without chasing anyone by WhatsApp.
  • Their real errors, visible: the app’s error-driven review system logs each student’s actual mistakes, and as their teacher you see the patterns. This is the feature that changes lesson planning: Thursday’s lesson starts from what Marieke actually got wrong on Tuesday, not from page 47 because it follows page 46.
  • The portfolio as lesson material: students' guided free writing accumulates in their portfolio; bring a piece to the lesson and correct it together: real output, corrected by a human, which is exactly the split defended above.
  • Motivation, optional: there are high-fives and a leaderboard, and the leaderboard can be switched off per class. Adult one-to-one students rarely want to compete; teenagers sometimes need to. Your call, per group.

The honest limits

What this does not do: it does not correct free production: the writing tool checks requirements (length, required words), deliberately leaving grammatical judgement to you, because a wrong automatic verdict is worse than none. It does not teach pronunciation nuance; the speech recognition checks word recognition, not accent (see the shadowing article for that boundary). And it will not make a reluctant student diligent; it will only make their non-diligence visible earlier, which is worth something too.

Teacher mode in amo Italian is free. If you teach Italian and drown in the second hour, the setup takes an evening, and if you want to compare notes teacher to teacher, write to me; this workflow is genuinely how my own weeks run. For group and school settings, the sequel article covers phones in the classroom, GDPR included.

References and attributions: the input/output framing draws on the work of Paul Nation and Merrill Swain as discussed, with sources, elsewhere in this series; all ideas are paraphrased with attribution. Part of the Science of Learning Italian series.