Clinician-supervised early intervention
ECI prepares structured observations from home video and organizes the evidence for review, so early-intervention clinicians can focus on the decisions only they can make.
Pre-revenue / Validation planned
Edits, rejects and other clinician corrections are recorded as override / supervision signals. No guidance is released before sign-off.
01/ The constraint
The research question
Can ECI reduce clinician time per family without compromising review quality or appropriate escalation?
Early-intervention systems cannot add qualified clinicians on demand. ECI is designed to move preparation away from scarce licensed time, while leaving interpretation and accountability with the clinician.
02/ One case, end to end
Prototype screens, sequenced for illustrationSynthetic demonstration. No patient data.
Step 1 of 7 / Home evidence
A 10-second clip: a child sits on a play mat and reaches for a toy. Press play to watch it.
The video is the input, and the source of truth. Every observation that follows links back to a moment in it.
Advances the demonstration only. It never grants approval.
Step 2 of 7 / AI preparation
Five items drafted from the clip, zero decided. Each links to its evidence. Where the evidence is uncertain, the item says so.
Advances the demonstration only. It never grants approval.
Step 3 of 7 / Clinician review
They watch the clip and choose: accept, edit, reject or escalate. Here the first item is accepted as drafted.
Edits, rejects and other clinician corrections are recorded as override / supervision signals.
Advances the demonstration only. It never grants approval.
Step 4 of 7 / Clinician correction
The AI's original stays visible, struck through. The edit is logged against the item as a supervision signal.
Advances the demonstration only. It never grants approval.
Step 5 of 7 / Clinician sets the activity
ECI does not choose the activity. The product says so on this screen: “Chosen and set by you · the AI does not prescribe activities.” The clinician selects Supported Sit-to-Stand, sets 5 minutes and up to 5 tries, and reviews the wording the family will read.
Nothing is sent until every observation is decided and the clinician adds a record note.
Advances the demonstration only. It never grants approval.
Step 6 of 7 / Approval
The clinician authorizes and signs. Their note is stored for the record and is not shown to the family.
Advances the demonstration only. It never grants approval.
Step 7 of 7 / Family receives guidance
The parent sees only the clinician-approved activity, with the clinician's name. No AI draft or measurement reaches this screen.
Advances the demonstration only. It never grants approval.
Step 1 of 5 / Home evidence
A 10-second clip: a child sits on a play mat and reaches for a toy. Press play to watch it.
The video is the input, and the source of truth. Every observation that follows links back to a moment in it.
Advances the demonstration only. It never grants approval.
Step 2 of 5 / AI preparation
Five items drafted from the clip, zero decided. Each links to its evidence. Where the evidence is uncertain, the item says so.
Advances the demonstration only. It never grants approval.
Step 3 of 5 / Clinician review
They watch the clip and choose: accept, edit, reject or escalate. Here the clinician is unsure how to read the clip.
Advances the demonstration only. It never grants approval.
Step 4 of 5 / Escalation
Escalation is the clinician's decision, raised outside the AI draft. The prototype asks the clinician for a reason and a written note.
Here the reason is that a specialist assessment is needed.
Advances the demonstration only. It never grants approval.
Step 5 of 5 / Escalated
It is written to an append-only decision record, with the clinician's reason. No activity is authorized and nothing reaches the family.
Advances the demonstration only. It never grants approval.
Step 1 of 3 / Review video
A separate synthetic case, shown as a safety check. The child cannot be seen well enough in the clip to work from.
Illustrative workflow state: the prototype's progress bar always lists six steps, so it is left out of this branch.
Advances the demonstration only. It never grants approval.
Step 2 of 3 / Insufficient evidence
No observation, no activity and nothing for the family. The clinician can still accept, edit, reject or escalate the flag.
Advances the demonstration only. It never grants approval.
Step 3 of 3 / Request another video
The branch ends here. Nothing is drafted until a better clip arrives.
Illustrative workflow state, adapted from prototype screens. No family guidance exists, so none is shown.
Advances the demonstration only. It never grants approval.
03/ First deployment
04/ Next proof
The question is not whether AI can draft. The question is whether ECI can safely increase clinical throughput while keeping accountability with the clinician.
Planned feasibility validation. No results claimed.
What will be measured
The goal is not AI efficiency.It is more children reached early enough for intervention to matter.
05/ What compounds
Structured clinician feedback creates a proprietary supervision layer that can improve quality, governance and trust over time.
Could become an asset for
Potential, not established. Defensibility depends on structured capture, appropriate data rights and demonstrated product improvement.
06/ Physician-trained. Parent. Builder.
Dr. Chioma Agha-Ayorinde
“I understand both sides of this system: the clinical constraints, and the reality of trying to carry a child's plan into everyday life.”
Discuss the clinical workflow and the next validation.