In short

Launching a new training course now takes 6 hours instead of 2 days: about 60% of the time saved. At Semisto Academy, Nova, an AI teammate built into the organisation's internal tool, sets up the project: tasks, day plan and calendar. The activities coordinator keeps what needs a human: prices, teaching choices, review.

The context

Semisto Academy runs forest-garden training: one-day introductions, 5-day courses, a 30-day programme. Each new course is a small project of its own, managed in My Semisto, the organisation's internal tool (projects, tasks, documents, calendar).

To launch an activity, the Semisto Academy activities coordinator has to think of everything: communication, registrations, logistics, follow-up after the day. He writes the day plan, breaks the work into dated tasks and sets the milestones. Without help, that was about 2 days of work per course.

The solution: an AI teammate inside the team's tool

Nova is not a separate chatbot: she is a team member in My Semisto, with her own account. You write to her in the chat or mention her in a message, and she acts with the same rights as a colleague: she creates tasks, writes documents and adds events to the calendar.

  • She reads the context of the project and the organisation before acting.
  • She does not make things up when a human decision is needed: with no price given, she creates a "Set the price" task instead of inventing one.
  • She flags where she departs from the request, and asks before any action outside the tool (for example syncing with Google Calendar).
  • The code belongs to the client: it is hosted on GitHub and another developer can take it over.

The demonstration, step by step

Three messages, about 5 minutes: Nova created 23 tasks, a complete day plan and 5 events. Reconstruction carried out on 21 October 2026 in a test project, for a fictional one-day course: "Gentle pruning of fruit trees", 12 places, at Les 4 Sources.

  1. The tasks (1 minute). A one-sentence request. Nova creates four lists (communication, registrations, logistics, follow-up after the course) and 23 dated, assigned tasks. She explains why three tasks fall after the course.
  2. The day plan (1 minute). A document from 9:30 to 17:00: theory in the morning, practice in the orchard in the afternoon, equipment, bad-weather plan and open decisions.
  3. The calendar (1 minute). The course and four milestones (registrations open, reminder, close, review), aligned with the task deadlines.

Result and what we learned

Before: about 2 days. With Nova: 6 hours, review and decisions included. The gain does not come from Nova's speed alone: the coordinator starts from a complete structure instead of a blank page, and spends his time where his expertise is needed.

  • One task at a time. Asking for everything at once went over Nova's step limit; split into three requests, everything worked first time.
  • The human decides. Price, lunch option, choice of trees: Nova lists them as decisions to make, without inventing anything.
  • An agent inside the existing tool. No new application to adopt: the team talks to Nova where it already works.

And for you? The AI Assessment finds your equivalent of those "2 days": the recurring task an agent can prepare for your team.