Trust is built through evidence, not promises.
Every EdTech company claims to improve learning. The DfE’s EdTech Evidence Board — delivered by the Chartered College of Teaching — exists to help schools look past the marketing and ask a harder question: how has this actually been evaluated?
We think that’s exactly the right question — so this page answers it on the Board’s terms, not ours. Where the evidence is strong, we show it. Where we’re still early, we say so. Evidence is a journey, and we’d rather tell you where we are on it than pretend we’ve arrived.
The framework
The five questions a school should ask
The Evidence Board encourages schools to judge a product not on its feature list, but on five interconnected questions. Here are ours, answered straight.
Question 1
What educational challenge does it solve?
Three that schools name themselves: SPHE and wellbeing get taught inconsistently and leave little trace; safeguarding concerns surface in a corridor conversation weeks too late; and teachers lose hours to marking and to assembling evidence for SSE, DEIS and inspection by hand.
Question 2
Who is it designed for?
Whole-school. Four scoped roles — student, teacher, coordinator, admin — that each see only what they need. Post-primary and primary, across Ireland, England, Northern Ireland and the US, with the curriculum, safeguarding framing and reporting adapting to the jurisdiction.
Question 3
How has usability been considered?
Roles are scoped so nobody meets a screen that isn't theirs. Built-in walkthroughs, one-tap routine sign-off, school SSO (Microsoft 365 + Google), offline fallbacks for opened lessons, and Learning Profiles that adapt the task for autism, dyslexia, ADHD and EAL pupils.
Straight answer: Shaped by design-partner schools (e.g. MSLETB principal feedback drove the plain-English safeguarding explainer). We have not yet published a formal usability study — that sits on the roadmap below.
Question 4
What evidence demonstrates its impact?
We are early on independent efficacy — and we say so plainly. What we have is a clear logic model, a product that instruments its own outcome signals, and design-partner feedback. Independent third-party evaluation is the next stage, not a claim we make today.
Straight answer: This is the honest position. Any EdTech telling you it has proven impact from a standing start is selling a promise, not evidence.
Question 5
How does it support accessibility, safeguarding and teaching?
Accessibility: Learning Profiles and bilingual EAL mode, same curriculum adapted. Safeguarding: the AI surfaces possible signals for a trained adult to review — it never diagnoses, labels or acts on its own. Teaching: the teacher signs off every AI level, edits any feedback and decides when pupils see it. GDPR-native, zero-retention AI rider.
Our logic model
From problem to signal — the whole chain
A product only earns an impact claim if it can name the problem, the mechanism, the intended outcome, and the signal that tells you whether the mechanism is working. Here is ours, in full.
Problem — Wellbeing is taught but rarely evidenced.
Does — Every reflection is scored on the NCCA 1–5 rubric and rolled into SSE / DEIS / inspection packs.
Outcome — Schools can show a real, documented wellbeing programme.
Signal — Reflection-level trajectory per pupil, per class, over a term.
Problem — Disclosures slip through unstructured channels.
Does — Reflections are AI-scanned and possible concerns are surfaced to the care team for review.
Outcome — A concern reaches the right adult faster, on a documented pathway.
Signal — Flag action logs — how many are confirmed, actioned or stood down, and how fast.
Problem — Marking and evidence-gathering eat teacher time.
Does — AI drafts a level and formative feedback; routine sign-off is one tap.
Outcome — Time moves from paperwork back to pupils.
Signal — Teacher override rate and sign-off patterns — how well the AI matches professional judgement.
Problem — Wellbeing delivery is hard to prove at inspection.
Does — Delivered lessons, hours and WBI indicators roll up automatically.
Outcome — The evidence pack writes itself instead of being assembled by hand.
Signal — Delivered-hours tracker and the six Wellbeing Indicators, school-wide.
The evidence journey
Where we are — and where we’re going
Evidence is not gathered once and framed on a wall. It’s a cycle of building, measuring and improving. Here is our honest position on each stage.
1 · Design with schools
OngoingWe build with design-partner schools and change the product on their feedback — the safeguarding explainer, the two-coordinator model and the feedback-release embargo all came from real school pushback.
2 · Measure in-product
BuildingThe product already captures the outcome signals in the logic model — reflection trajectories, flag action logs, override rates, delivered hours. The next step is surfacing these back to each school as their own impact picture.
3 · Independent evaluation
PlannedThe step no vendor can mark its own homework on. We are pursuing an independent evaluation partner to test impact in real classrooms — the kind of third-party evidence the Evidence Board is right to hold the sector to.
What we don’t claim (yet)
- We do not claim a proven, quantified effect on exam results or wellbeing outcomes.
- We have not run a randomised controlled trial, and we won’t imply one.
- We do not publish testimonials or statistics we can’t stand over.
- Our safeguarding scan is decision-support, not detection — it never diagnoses or acts alone.
When any of these changes, this page changes with it. That’s the point of an evidence journey.
Instrumented today
The evidence base is already in the product
These are captured on every school’s live data — the raw material an independent evaluation would draw on, and the picture we’re building back for each school.
Reflection-level trajectory
Every reflection carries a 1–5 level; the trend over a term is a per-pupil growth signal.
Flag action logs
Each safeguarding flag is tracked to an outcome — a ground-truth signal on how well the scan matches real concern.
Teacher override rate
How often teachers change the AI's proposed level — a running check on whether it earns professional trust.
Delivered wellbeing hours
Lessons and hours delivered, mapped to the curriculum and the inspection target.
Six Wellbeing Indicators
Active · Resilient · Connected · Aware · Respected · Responsible — tracked class- and school-wide.
Engagement & pulse
Daily Pulse and weekly check-in participation — the earliest signal that a cohort is disengaging.
The human stays in charge
The AI proposes. A trained adult always decides.
Evidence is only trustworthy if the system it describes is trustworthy. In Skilly, the teacher signs off every AI level, edits any feedback, and controls when pupils see it. Safeguarding signals are surfaced for review, never acted on automatically and never used to label a child. Data is processed under GDPR with a zero-retention AI rider.
Judge us on the evidence. We’d rather you did.
If you’re evaluating Skilly for your school and want the logic model, the data we capture, or a conversation about independent evaluation, we’ll bring the detail — not a sales promise.