Key takeaways
- In Smootables the timetable comes from a constraint solver. The AI is a planner assistant that answers questions about the plan and makes the changes you describe in plain language.
- The assistant acts through the same commands as the buttons in the app, with the same validation and the same school scope, so it has no shortcut past your rules.
- Full names of teachers, students and groups are replaced with codes before a request reaches the language model, and names are filled back in on your screen.
- The assistant is available to every college, and each college decides whether to use it. A college can choose not to use AI: planning, generation and editing all work without it.
What should AI do in a vocational college timetable?
The public evidence is clear on one point: a language model working alone builds poor timetables. The one documented attempt to build a real school timetable with ChatGPT failed on teacher double-booking, and planning benchmarks show the same models succeeding on a small share of tasks when they plan alone. Results improve when a model proposes and a solver or verifier checks. None of those studies used a vocational timetable, with its workshops, equipment and placement days, which only adds rules for a model to break.
Smootables follows that split. The constraint solver places lessons and keeps every hard rule. The planner assistant handles the work around it: reading a large plan for you, making many edits from one sentence, and starting runs. A planner at a college with dozens of qualifications spends far more time changing the plan than waiting for a solve, and that is the time the assistant saves.
What can you ask the planner assistant to do?
The assistant works on the plan and timetable of your college, in the terms you already use for groups, units and periods. These are the kinds of requests it is built to handle.
Change delivery in bulk
Mark a set of placements as workplace learning or remote teaching, for example every placement of a second-year group in the period it spends with employers.
Set schedule preferences
Ask for a group's theory units in the morning or on particular weekdays, across many placements at once, instead of editing them one by one.
Staff and fill placements
Assign a teacher to every placement of a unit, auto-assign the least busy eligible teacher, add individual learners to a placement, or copy a period's plan into another period.
Ask about the plan
Which teachers are over their weekly hours, which placements still have no teacher, what a group has in a given week. The answer comes from your data, read through the same tools the app uses.
How is the planner assistant different from a chatbot timetable?
Both take plain-language requests. What differs is where the timetable comes from, what happens to your data, and what the AI is allowed to touch.
| Question | General-purpose chatbot | Smootables planner assistant |
|---|---|---|
| Who places the lessons | The language model writes the grid itself | The constraint solver; the assistant can start a run |
| Double-booked teachers | The documented failure in public attempts | A hard rule the solver keeps |
| Names of staff and learners | Pasted into the chat | Sent as codes; names restored on your screen |
| Changes to the plan | Copied back by hand | Applied through the app's own commands and checks |
| Record of what changed | A chat history | Every AI-originated change is logged |
| Who can use it | Anyone with an account | Signed-in planners at a college that has chosen to use it |
How does a session with the planner assistant run?
When your college uses the assistant, a typical session sits between planning and generation, when the structure is agreed and the plan needs many small changes.
- Open the Planner assistant panel while you work on the plan or the timetable.
- Describe the change in plain language, naming the groups, units and periods as your college calls them.
- The assistant picks the tools that fit the request. A request it cannot read with confidence is answered with read-only tools, so it looks before it changes anything.
- The change runs through the same command a button would use, with the same validation and limits on how much one bulk change can touch.
- Check the result in the plan, then generate the timetable when the period is ready.
Is the timetable itself generated by AI?
No. Smootables generates timetables with a constraint solver (Google's CP-SAT), which first finds a timetable that keeps every hard rule and then optimizes your weighted preferences. The same plan always meets the same rules. The assistant can start a run and help you read the result; it cannot place a lesson the solver would reject. For how the solver works, see automatic school timetabling.
What happens to staff and learner data?
The assistant sees your data in two ways. The plan it reads is sent with teachers, students and groups as codes such as T014 or G203, and rooms as codes too. In what a planner types, Smootables replaces a teacher's, student's or group's full name with its code before the message leaves; a partial name, a first name on its own, or a name shared by two people is passed through as typed. Unit names from the curriculum are sent as they are, because they are not personal data. When the answer comes back, Smootables fills the names back in on your screen.
Smootables runs on Microsoft Azure in Sweden Central, and its infrastructure rule is that no school data is processed outside the EU. AI-originated changes are logged, use is rate-limited per planner, and each college decides whether its planners use the assistant at all. The planner assistant product page covers the feature in more detail.
Questions about AI in vocational college timetabling
Does AI generate our timetable?
No. The constraint solver generates it from your year plan. The assistant helps you prepare and change that plan, starts runs, and answers questions about the result, but it does not place lessons or override hard rules.
Can the assistant change our plan on its own?
It acts only on a request from a signed-in planner, inside that college's scope, through the same commands as the app's buttons. It has no write access the buttons do not have, and its changes are logged.
Can our college choose not to use AI?
Yes. The assistant is available to every college, and using it is each college's choice. Year planning, generation, editing, import and export all work without it.
What does the language model see?
The plan as codes for teachers, students, groups and rooms, plus unit names from the curriculum. In a typed request, full names of teachers, students and groups are replaced with codes; a partial name or a first name on its own is sent as typed, so use full names or codes when you write about people.
Can the assistant help with importing our old data?
It can walk you through an Excel or CSV import. The import itself uses the same steps as the import page: column mapping, a preview, and matching against teachers, rooms and groups you already have.