MESSY INPUTSBefore
Texts, spreadsheets, and three people checking the same time window.
AI-run operations, human-approved
AI drafts your schedules, reports, and payment follow-ups every day. Hard rules verify each one. Nothing ships until you approve it.
Running today's
Interactive preview · synthetic academy · no data saved
Today's brief3 decisions ready · hard rules checked each request
Move Mia's piano lesson to Thu 4:30 PMSchedule change · Synthetic AI draft · 7 example checks passed
March progress report for the Chen familyParent report · Synthetic AI draft · 7 example checks passed
PICK A JOB. WATCH THE HANDOFF.
From scattered inputs to one decision.
Replay the flow: preparation, rule checks, and your final say.
Outcome 1 of 3: Schedule recovery
01 · Schedule recovery
Turn one messy request into a safe decision.
7hard checks visible
AI INTERPRETSPrepared brief
AI structures the time intent. Deterministic rules evaluate and sort eligible possibilities.
✓Teacher qualified
✓Room available
✓No learner conflict
Deterministic rules remove every unsafe assignment.
HUMAN GATEReady for review
PROPOSED NEXT ACTION
An operator reviews one exact teacher, room, time, and effect.
YOU
Final decision stays with youNothing changes until you approve.
01 Inputs gathered02 Assistant prepares03 Rules verify04 You decide
AI drafts. Rules verify. You approve.
- 01AI drafts
Reads today's requests and writes the exact change.
- 02Rules verify
Conflicts, capacity, and permissions checked deterministically.
- 03You approve
One decision, then it ships. Nothing moves without you.
Deterministic rules own validity. AI never bypasses them.
FOUR INTERACTIVE JOB DEMOS
Choose one job. Follow it to review.
Each demo starts with different academy data and ends with a different result. The five control stages stay consistent so you can see where AI helps, rules decide, and a person takes over.
Synthetic previewNo model callNo API requestNo data savedNo external action
CHOOSE A DEMOWhat do you want to see?
Scheduling finds a valid time. Post-class uses teacher evidence. Collections keeps ledger facts fixed. Reports locks an exact PDF version.
ACADEMYTEMPO · SYNTHETIC
TASK-SCOPED REQUESTDescribe one operational job.
Operations AI or fallbackNo provider call in tour
Fixed scenarioUse synthetic information only. Editing demonstrates invalidation and scoped handling. The prepared cards are fixed scenario examples, not outputs generated from this text, and no model, API, or storage receives it.
LOCAL SCENARIO INTERPRETATION
Current requestFind a 60-minute option next Tuesday after 4 PM. Keep Sofia if possible.
WindowNext Tuesday after 4 PM
Duration60 minutes
PreferenceKeep Sofia if safe
The workflow bundle separates inputs by authority.
Preparation context, deterministic rule inputs, and approval-only evidence stay distinct while every source keeps its provenance.
ONE TASKSchedule recoveryTenant and role scoped
EXCLUDED FROM THIS REQUEST
× Guardian contacts× Student notes× Unrelated classes× Raw recordings
Three nearby assignments survived the first pass.
AI interprets the time request. Deterministic scheduling logic evaluates and orders eligible possibilities without creating a lesson.
These pre-authored outputs demonstrate how an AI-prepared review can work. They are not generated from the textarea, and no assistant owns validity, approval, or persistent state.
Inspect the evidence required before review.
This pre-authored scenario shows what a deterministic result must expose in the real workflow.
- Teacher qualificationSofia Rivera has the required Robotics qualificationEVIDENCE 01
- Class capacity8 enrolled / 10 class seatsEVIDENCE 02
- Location capacity8 learners fit Maker LabEVIDENCE 03
- Teacher availabilitySofia Rivera is available at Tuesday, 4:30 PMEVIDENCE 04
- Student availabilityAll enrolled learners are availableEVIDENCE 05
- Teacher conflictSofia Rivera has no overlapping lessonEVIDENCE 06
- Location conflictMaker Lab has no lesson overlapEVIDENCE 07
AI cannot override a failed check.A changed source requires a fresh deterministic run and a new exact review.
One exact decision is ready for a person.
The tour can preview the effect, but it cannot approve, save, send, or execute it.
- Class
- Robotics Lab
- Teacher
- Sofia Rivera
- Location
- Maker Lab
- Time
- Tuesday, 4:30 PM
Changing any field after approval would require a new review.
GOOGLE WORKSPACE, ONE JOB AT A TIME
One academy decision. Four Google steps, kept under control.
Choose a job below to see what Calendar, Meet, Drive, and Gmail would do, what still needs approval, and what remains locked.
Interactive capability preview
PREVIEW ONLYNo Google account is connected here. Nothing is created, attached, or sent.
SELECTED ACADEMY JOB
Robotics Lab · Tuesday 4:30 PM
Operator approves one rule-checked Robotics Lab assignment.
Four follow-ups can drift apart.
- Copy the approved time
- Create or find the meeting
- Locate the right file
- Rewrite the recipient message
One review keeps the job understandable.
This preview illustrates how one approved internal lesson could source separately controlled Google operations.
Exact sourceSeparate approvalsVisible outcomes
WHAT EACH TOOL WOULD DO
Every card is a separate operation, not one automatic bundle.
GOOGLECalendar
Robotics Lab
A provider write would require the exact approved app operation.
GOOGLEMeet
Meet space for this lesson
Created only when requested after approval
GOOGLEDrive
robotics-brief.pdf
Owner or administrator selects it · app-visible metadata only
GOOGLEGmail
Robotics Lab schedule update
Calendar time, access note, and selected material are assembled for exact review.
Human control stays between preparation and execution.
NOT CONNECTEDA connection, the exact approved operation, and a recorded provider outcome are still required for every Google action.
Access stays narrowThe designed Calendar scope reads free and busy only. Meet operations request app-created spaces without recording access. Drive operations use selected or app-created files only. Gmail operations are send-only and never read the inbox.
Ready to run your academy?
Sign in to workspaceThe interactive tour above uses synthetic data, requires no signup, and saves nothing.