Marker Learning / weaving AI into the long, careful work of writing psycho-educational reports.
Marker Learning builds tools for clinicians who write long, careful, evidence-based reports about learners. We designed where AI assists and where it stays out of the way.
Clinicians write long, careful reports. AI is useful in some parts of that work, in the way and not useful at all in others.
Marker Learning provides accessible psychoeducational evaluations to students and families. The existing student-evaluation report workflow was functional but outdated: difficult to navigate, slow to fill out, and unable to leverage AI in any part of the long-form writing clinicians do.
Marker needed an experience that felt intuitive and adaptable to clinicians, not a generic AI editor bolted onto a clinical workflow. They also needed it shipped in time for an important client launch. The pressure on the timeline was real and the consequences of getting AI wrong in a clinical context were larger than the consequences of being late.
We partnered with Marker to redesign the student psychoeducational evaluation report workflow and ship AI-powered features that improved usability and client satisfaction without changing the parts of the work that clinicians needed to fully own.
Measured after launch.
- 35% faster workflow, an average of 20 minutes saved per report.
- +40% user satisfaction, based on post-launch survey scores.
- +25% active users, growth in daily engagement within the app.
- +30% retention, clinicians more likely to return and reuse the platform.
The AI does not invent. It traces.
When a clinician adds a student, they can upload prior documents: a previous assessment, a referral packet, school records. The system scans the content and suggests where the new report could pull from existing material, but every suggestion is anchored to its source. Click a suggested edit, see exactly which paragraph of which document it came from.
- Suggests text replacements or additions, never silently inserts them.
- Highlights suggested edits directly in the report draft for review.
- Click-through traceability from every AI suggestion back to its source document.
- Saves real manual effort (20 min per report on average) without removing the clinician's editorial control.
The traceability is the trust mechanism. Clinicians do not accept AI suggestions because the AI is impressive, they accept them because they can verify the citation in one click.
Built on the rails clinicians already trust.
We redesigned the editor surface using TipTap and LiveBlocks, frameworks Marker's engineering team was already comfortable shipping with. The visual update created a more professional, user-friendly experience that resonated with both clinicians and the enterprise clients they work with, without forcing the engineering team into an unfamiliar tooling rewrite.
Picking the right rails matters here. A rewrite would have shipped late. The chosen frameworks shipped on time and stayed maintainable after handoff.
The "new report" flow that actually matches how clinicians start work.
The old flow had one path: start from scratch. The new flow has three, because clinicians do not all start the same way:
- Start from a template, for standard evaluation types.
- Copy from an existing student report, for similar cases or longitudinal work.
- Start from scratch, when the situation does not match either of the above.
AI in clinical workflows works when it traces, not when it generates. Every minute we saved a clinician was a minute we let them spend on the parts of the report where their judgment is the actual product.