Questions
Questions people ask us.
Short answers, with links to the pages that go further. If yours is not here, write to hello@synthetic.school.
The product
What is Synthetic School?
A complete school in software, which one person runs as its leader for a term or a year. Every member of staff, every student and every family is an AI agent, and a rules engine decides what each decision causes. How it works
Who is it for?
Leadership programmes, school groups and boards use it to train aspiring and new school leaders. Companies building AI for schools can test their products inside it, and school groups or education systems can run policy questions through it. Leadership training
How much time does it take a trainee?
A sitting covers one school week, with a target of 45 to 75 minutes that the first players will test. A programme seat covers 12 played weeks, about 9 to 15 hours, and the other weeks are fast-forwarded or delegated to the senior team.
Which school systems does it cover?
Packs for a public school system and for international schools under a regulator are in build. Sources are mapped for five more systems on four continents, and each new pack is reviewed by a serving school leader from that system. School systems
Can a school group use its own policies?
Yes. A group pack layers a school group's own policies and standards over its regulator's pack, for the group's own leadership academy.
How do the AI agents work, and can they get things wrong?
Agents voice the people in the school and propose what they do. The engine resolves every proposal by rule, so an agent's mistake in narration cannot change the school's state. Narration is checked against that state, and personas are audited group by group.
Which AI models does it use?
The agents run on Anthropic's Claude models, and each moment goes to a model tier by how much it matters to the trainee. Model versions are pinned for each run, and the validation suite runs again before any switch.
Is it a game?
It is a training simulation built around decisions. The school appears through an inbox, a diary, conversations, dashboards and a 2D floor plan. In health education, high- and low-fidelity simulators differed by 1% to 2% on average.
Can we test our AI product with it?
That is what the test harness is for. Your agent joins the school in a staff role through a REST API or an MCP server, and the run report shows what it did and what followed. The harness comes in the fourth build stage, in the order the first buyers ask for. AI test harness
Can I use it on my own, without a programme?
Individual seats are planned. Anyone playing without a coach is told that the evidence favours a coach, and child protection storylines appear only after they opt in.
Safety and data
Is any real student or staff data used?
No. No import path accepts real student or staff records, and no persona, storyline or school is modelled on an identifiable person or school. The only personal data in the system belongs to the adults who play it.
Does it score or rank school leaders?
No. Nothing leaves the system as a score or a ranking. The debrief maps choices to the system's leadership standard as evidence for learning, and contracts bar any use of play records to select, appraise, promote, discipline or rank a leader.
How do you handle child protection storylines?
They come from a reviewed library and follow each system's published procedure as fixed steps. Harm reaches the school leader only as a report, without graphic detail. Facilitators choose which kinds of storyline a term may contain, trainees see that list first, and anyone can pause or leave a scene. Safety and privacy
Where is my data stored?
Play records are stored in a UK or EU region. Model calls go to Anthropic in the US by default, and a buyer who needs data kept in a region can use a regional cloud endpoint. Privacy notice
Evidence
Does it work?
A prototype of the engine reproduces five published attendance figures in two school systems, each within 0.2 points. Published research backs the design's debriefs and its rules engine. No study yet links simulation-based leadership preparation to leaders' practice or students' outcomes. The pilot trial starts with the first: leaders' practice on new scenarios. Evidence
Will it improve outcomes for students?
We make no such claim. School leaders' measured effects on test scores are small even in large datasets, so the pilot trial measures leaders' practice on new scenarios and in the job.
Can the policy lab predict what a policy will do?
Its results stay labelled illustrative until validation supports more. It compares options under stated assumptions, with a sensitivity analysis showing how far each result depends on them. Policy lab
Pilots and pricing
What does a pilot involve?
Trainees give 90 minutes a week for 12 weeks, plus the assessments, in a randomised trial against case-based learning with the same coach-led debrief. Every trainee gets a term of Synthetic School, the comparison group after the follow-up. Pilots
How much does it cost?
Pilot places are free or at cost while Synthetic School is validated. Prices for later cohorts will be set with the first design partners.
Our system has no pack yet. Can we still pilot?
Yes. A pilot in a new system starts with building its pack together, and a serving school leader from your system reviews it before anyone plays.
Still have a question?
Ask it in a pilot request, or write to hello@synthetic.school.