Safeguarding — After the Bell

Build a safeguarding training product for the UK schools market — every member of staff, every school, annual renewal under KCSiE 2025 — that trains the actual disclosure conversation rather than safeguarding vocabulary. The Module 1 demo must be playable in fifteen minutes and credible enough that a Designated Safeguarding Lead would deploy it across their multi-academy trust. The mechanic is conversational, not decision-architectural. The audience is not compliance officers; it is teachers, teaching assistants, lunchtime supervisors, and pastoral staff.

Most safeguarding training trains *recognition*: signs of abuse, types of harm, the role of the Designated Safeguarding Lead, the Single Central Record. These are knowledge tests with knowable answers, and the off-the-shelf market has saturated them. The skill that fails when a real disclosure happens is not recognition — it is *language under emotional pressure*. A Year 9 girl stays behind after registration. She says one sentence that is testing whether you are safe to talk to. What you say in the next thirty seconds determines whether she discloses, qualifies, retracts, or never speaks again. The discovery question was: can you train that conversation? Not in a workshop with a role-play partner — in scalable, deployable eLearning that 24,000 UK schools can buy, assign, and renew annually. The answer the off-the-shelf market gives is "no, you can't train conversational micro-skill in eLearning, so we'll train recognition instead." This product is the answer that says yes you can — by building a 135-decision dialogue tree dense enough that the failure modes become visible to the learner as they make them.

The course is built around a single conversation, told in real time. Maya stays behind after registration. The classroom is empty. The bell has rung. The opening line is hers, and from there every line of dialogue is the learner's choice. Design decisions that broke the safeguarding-as-recognition mould: • **The decision tree IS the curriculum.** Most safeguarding eLearning structures content as topics with quizzes attached. This product structures the entire Module 1 as one branching conversation. Pedagogical content — what counts as a leading question, what undermines confidentiality, what KCSiE 2025 requires of all staff — surfaces *as the consequence* of choices, not as content delivered before the choices. The learner builds the rule from feeling the wrong answer fail. • **Conversational micro-skill, not decision-architectural skill.** The other compliance scenarios in the catalogue train decision architecture (which option is correct under what statute). This trains language. The lessons are about word choice, silence-holding, question framing, and the difference between reassurance and over-promising. The format is dictated by the skill — anything coarser than 135 decisions would not be granular enough to surface those moves. • **The audience is plain-English-first.** Teachers, TAs, and pastoral staff are not compliance professionals. The product avoids policy vocabulary except where KCSiE 2025 requires it (DSL, single agency referral, threshold). Where the law has to surface, it surfaces *because* a choice has triggered it, not as preamble. • **Modules 2–5 propagate Module 1's choices.** Documentation in Module 2 is the documentation the learner can honestly write given what they actually said in Module 1. The DSL handover in Module 3 is constrained by what Maya disclosed, which depends on how the Module 1 conversation went. Multi-agency referral thresholds in Module 4 either trigger or don't, depending on whether the disclosure crossed the threshold. The course is one disclosure with five tiers of consequence. • **Visual quietness as a deliberate choice.** The opening shot is an empty classroom — the bell has rung, everyone has left. The visual idiom across the course is quiet, intimate, school-realistic. No ops-center red lights, no boardroom scenes, no regulator's hearing. The visual gravity matches the emotional weight of the subject. Stack: native HTML/CSS/JS modules, custom Node SCORM 1.2 build pipeline, AI-assisted scene generation tuned for school-realistic interiors, multi-character TTS-driven voicework with age-appropriate voice casting (Maisie/Libby for the student, adult voices for staff), automated path-walking QA across the full 135-decision tree before ship. Reviewed and signed off by qualified UK safeguarding counsel before release.

The shipped product covers what 30-minute video-and-quiz safeguarding training cannot: • **A 135-decision dialogue tree inside one conversation.** Module 1 — "After the Bell" — is the disclosure itself. Maya, a Year 9 student, stays behind after registration. Across roughly five minutes of conversation, the learner makes 135 micro-decisions about exactly what words to say, what questions to ask, what silences to hold, and what assumptions to avoid. Each branch closes one possibility and opens another. The decision tree is the lesson. • **Active-listening as the mechanic, not a topic.** The course doesn't *describe* active listening — it *requires* it. Choosing leading questions narrows the disclosure. Choosing reassurances about confidentiality you can't legally keep collapses the conversation. Choosing to hold a silence rather than fill it changes what Maya says next. Learners feel the conversational gravity in real time. • **Audience: every member of staff, not just the DSL.** Pastoral staff, classroom teachers, TAs, lunchtime supervisors — KCSiE Part 1 obliges all-staff training. The product is designed to work for someone whose first language about safeguarding is plain English, not policy. • **5 modules, 2.5 hours total.** Module 1 is the disclosure. Modules 2–5 cover the 24-hour aftermath: documentation, the DSL handover, the multi-agency referral threshold, and the boundary between safeguarding and pastoral support. Decisions in Module 1 surface as constraints in the modules that follow — what you said in the conversation determines what you can later document and who you can later involve. • **Built for the 24,000-school recurring-revenue market.** Annual renewal under KCSiE; SCORM 1.2 packaged; deployable to school LMSes (Frog, Vocate, Smoothwall, RM Unify). Aimed at the multi-academy trust as the buyer, with per-staff licensing to scale across the trust without per-school setup. • **Free playable Module 1 demo, no login.** Fifteen minutes, the entire 135-decision conversation, with the full debrief at the end.