Most learners will never meet a specialist.
Not because the expertise doesn’t exist. Because there is never enough of it, so it gets rationed — to whoever can pay, wait, advocate hardest, or happens to live somewhere with provision.
Edra builds application-layer AI that delivers the work of a learning specialist directly to learners and their families — continuously, in their own language, and free at the point of use.
The problem
Rationed, and then frozen.
A specialist’s time is finite, so a system is built to allocate it. Referral, waiting list, threshold, assessment, report. The machinery is not incompetent — it is doing exactly what scarcity requires, which is deciding who goes without.
For the few who reach the front of the queue, expertise arrives as an episode. An assessment on one day, a document a few weeks later, a review in a year or two. The document is treated as a description of a person, but it is a photograph — accurate for an afternoon, and quietly wrong for a long time afterwards.
And through all of it, the person who knows the most about the learner is the least consulted. In our stakeholder research, the practitioners closest to the work — the most experienced special educators, and clinicians in community settings — were the ones who rated meaningful learner involvement lowest. Not sceptics of the idea. People who have watched it not happen.
And the shortfall is not shared out evenly. It lands on the same people everywhere — those who mask well enough to look fine, and those whose families have least standing with institutions.
What changes
The scarce part was never the judgement. It was the hours.
Most of what happens around a learner is procedural: gathering an account of how they think, turning it into a usable document, noticing when it stops being true, revising it, translating it, explaining it to the next adult who needs to understand. That work is scarce because it consumes expert hours — not because it requires an irreplaceable human.
That is the part we build. Delivered directly to the learner and the people around them, without waiting for an institution to stand in between.
What is left is the part that was always the point: relationship, judgement, teaching, advocacy. A specialist’s reach multiplies. The specialist does not disappear.
Commitments
What we will not trade away
The learner is the author.
Not consulted, not surveyed, not represented. What we produce is built from what they said, in their words, with every claim traceable to something they actually told us. Where someone else contributes, it says so.
Adults are coaches, not gatekeepers.
Teachers and clinicians hold training and judgement no model has. The work is to free that expertise from paperwork, not to route around it. A system that replaces the gatekeeper with a different gatekeeper has failed.
Always current, never frozen.
A description of a person should change when the person does. Episodic assessment produces documents that are out of date before they are filed. The alternative is something that keeps listening.
Native, not translated.
Most localisation is a translated interface over an assumption that did not travel. Our own research turned up families whose children were absent from the data for reasons no translation would have fixed. Building for that is a design problem, not a language problem.
Safety is a constraint, not a feature.
This work involves young people describing difficult things. Any system doing it carries a duty that has to be designed for from the beginning and proven continuously — not asserted. We would rather move slowly here than be trusted before we have earned it.
Free by design
The families who need this most can pay for it least.
That is not an unfortunate detail to solve later. It determines what can be built. A system whose economics require a district budget, a statutory framework, or a family with means will only ever reach people who are already reached.
So the target is free at the point of use, funded upstream by the institutions whose mandate this already is — NGOs, agencies and governments. Costs are held near zero by design rather than subsidised down, which is the only version of free that survives contact with scale.
Where we are
Sol, and a deliberately small first claim
Everything above is a thesis. The first test of it is narrow on purpose.
Sol talks with autistic adolescents and their caregivers — over the messaging apps they already use, in ordinary conversation, with no forms and no questionnaire — and drafts an integrated learner profile from what they say. One task, one population, measured properly.
It is the subject of an active doctoral study at the University of Cambridge, with its own ethical approval. We will publish what we find, including the parts that do not support the argument above.