Escape Rooms Become the Trojan Horse for AI Fluency Professional Development in Schools

PBL Future Labs’ Phillip Alcock offers this commentary on how escape rooms can become the trojan horse for AI fluency professional development in schools. This article originally appeared in Insight Jam, an enterprise IT community that enables human conversation on AI.

AI fluency has become the most urgent conversation in schools, and the training built to deliver it keeps failing in contact with so many classrooms. A teacher sits through a session, nods at the slide about the future of work, and walks out with no idea what to do differently on Monday. A small number of schools are trying something else. They are getting teachers to re-design learning experiences, to reconsider what the lesson plan is, and.. to build.. escape rooms. 

AI Training or Lack Thereof

By autumn 2025, half of US teachers had received at least one professional development session on AI, up from 42 percent the year before and just 13 percent in 2023.  That is progress on paper. The catch is what RAND found when it tracked AI training across school districts: most of it is a single session, not anything ongoing. AI for Education’s Amanda Bickerstaff has argued that without sustained funding and support at the school and district level, AI in classrooms will keep depending on whichever IT staffer or curious teacher happens to push it forward, rather than becoming the kind of change that survives a budget cycle.

Picture where most AI integration efforts sit right now: a crumbling passageway out of an Indiana Jones film. The flagstones might be pressure plates. Nobody has mapped the room. aiEDU’s Alex Kotran has pointed out that nobody yet knows how to describe the jobs of the future, which makes it close to impossible to write a workforce strategy for them. Teachers are being asked to decode a structured, constraint-based system anyway, on a timeline set by people who openly admit they cannot describe what they are preparing students for. 

The keys to a working AI fluency program are not in that slide deck. They are hidden in the structure of the escape room itself, where every locked door demands a deliberate, sequenced solution. This article is the first challenge. 

Why Building Beats Receiving

I have run a small number of these workshops, and the pattern is consistent enough that I trust it. Teachers who build their own escape room leave able to write an AI instruction that survives contact with a real student. Teachers handed a slide deck about the future of AI mostly do not. 

Seymour Papert spent his career arguing that people build knowledge most effectively when they are actively constructing something in the world, not receiving it. He called it constructionism, and education researchers have since traced a direct line from his work to project-based learning.[5] The Learning Policy Institute’s review of 35 rigorous studies on what actually changes teacher practice found the same thing from the opposite direction. Active learning, where teachers do the same hands-on work they will later ask of students, is one of seven characteristics that separate professional development that changes practice from professional development that evaporates by Monday. A single AI training session, however well delivered, fails most of those seven criteria before it starts. I live in the strange place that sits in active learning and AI, and the more I dig into PBL, the more I see the connection. 

One Lock, One Session

The workshop itself is the challenge, because escape rooms are a different learning design approach to almost anything teachers have built before. 

The first rule is non-negotiable: one learning outcome behind every lock, not three unrelated puzzles wearing the same theme. 

The second rule is that the teacher has to build the room, not receive one. A pre-made escape room teaches a teacher nothing about AI. Writing the instruction, testing whether a student could solve the puzzle by guessing instead of understanding, and rewriting it until they could not, is the actual training. 

Scott Nicholson has spent more than a decade studying what separates a working escape room puzzle from a guessing game, and the failure mode he keeps coming back to is consistency between a puzzle and its environment. A puzzle a player can brute-force without understanding why the lock opened has taught them nothing, no matter how satisfying the click. That is the exact test I put teachers through when they write an AI instruction. Can a student get the right answer for the wrong reason? If yes, the lock is broken, and they rebuild it. 

What the Build Proves

I do not have a study behind the specific claim that building an escape room teaches a teacher AI fluency. What I have is a consistent pattern across a small number of workshops, and a body of adjacent research that makes the pattern plausible rather than wishful. 

A 2023 meta-analysis of educational escape rooms across subjects and age groups found an effect size of around 1.4, among the largest reported in education research for any single classroom intervention. That study measured escape rooms as a learning activity for students, not as a training method for teachers, and I want to be honest about that gap rather than paper over it. The mechanism is the same one. Working through a tightly constrained, sequenced problem produces learning gains that simply receiving information does not. 

Where This Doesn’t Replace Anything

An escape room with a vague learning goal behind the lock is not a fluency session. It is an afternoon of fun, and the gap between the two is easy to miss if you are not the one watching for it. 

This is an entry point, not the whole journey. Teachers still need follow-up. They still need someone to check their work. They still need time they rarely get, which is the same sustained duration criterion the Learning Policy Institute flags as essential and almost no district budgets for. None of that changes because one workshop worked. 

But that is the case for starting here, rather than with another session on the future of AI in education. The policy conversation happening in statehouses right now is mostly about risk mitigation. Sari Factor, chief strategy officer at Imagine Learning, has argued that most state AI guidance focuses on privacy and plagiarism rather than the harder question of how teaching and learning actually need to change. Building something that has to work, under a lock, with a real student trying to break it, changes practice in a way a guidance document cannot.