From weeks to 15 minutes — AI that rewrote the report bottleneck in neuroaffirming autism & ADHD assessments

When Human Endeavours came to us, the problem wasn't clinical. It was paperwork — specifically, the diagnostic report that closes every ADHD and autism assessment.

Human Endeavours is a neuroaffirming practice: they don't treat neurodivergence as a deficit to be fixed, but as a different way of thinking that deserves understanding, support, and inclusion. Their assessment journey is thoughtful by design — a free consultation, pre-assessment questionnaires, and one or two two-hour therapeutic interviews that explore a person's challenges, strengths, and lived experience in depth. Then comes the feedback session and the comprehensive diagnostic report summarising the findings and recommendations.

That final report is where the whole system stalled.

The report was the bottleneck

In neuroaffirming (and NDIS-aligned) assessment, the report isn't a formality. It's the artefact a person carries into the rest of their life — to access supports, workplace and study accommodations, and the right kind of help. It has to be thorough, personalised, and clinically sound.

The catch: writing it is slow. Across the allied-health sector, report writing for functional-capacity and diagnostic assessments is consistently cited as the single biggest time sink — often 7–8 hours of drafting per report, and 6–8 weeks from first contact to delivered document. For Human Endeavours, the report generation step alone stretched to two to three weeks of clinician time, stacked on top of an already full caseload.

That delay isn't just an operational nuisance. It's a human one. Someone who has spent years questioning their experiences finally books an assessment, does the interviews, and then waits weeks for the document that unlocks the support they came for.

The clinical insight was already there, captured in the interviews. The bottleneck was the act of turning it into a report.

What we built

We built an AI pipeline that does the one thing clinicians least need to do by hand: assemble the raw material into a coherent, structured first draft.

Here's the flow:

• Ingest everything. The system reads all documents relating to the patient — questionnaires, prior notes, referral material — and the full transcript of the entire patient interview, not a summary.
• Synthesise, grounded in the source. A language model works through that material and produces a high-quality diagnostic report draft: structured, narrative where it should be narrative, and faithful to what was actually said and submitted.
• Clinician reviews and signs. The psychologist reads the draft, edits where needed, and owns the final document.

The headline result: report generation dropped from 2–3 weeks to about 15 minutes.

Why this works (and why it's safe)

The instinctive objection is the right one: you don't let a model decide a diagnosis. We didn't build it to. The system drafts; the clinician decides. That human-in-the-loop design is the whole point.

A few principles made it responsible rather than reckless:

• The model writes from your material. It reads the actual documents and the actual interview transcript. It's not inventing a person — it's organising what's already true.
• A clinician is always the author of record. The 15 minutes produces a draft. The psychologist's judgement, voice, and accountability remain the final word.
• Privacy and consent are non-negotiable. Sensitive health information stays within the controlled environment; participants are told how their data is used. This mirrors the responsible-AI stance NDIS thinkers like Team DSC are pushing for — AI as a tool that frees clinicians for participant-facing work, never as a replacement for human judgement or informed choice.

The pattern behind the win

This isn't really a story about reports. It's a story about finding the real bottleneck and aiming AI at it.

Most teams automate the wrong thing first — a chatbot on the website, a flashy demo for the board. The leverage is almost always in the unglamorous middle: the repetitive synthesis of material a human already collected. Reports, case notes, summaries, triage drafts. The work that is high-volume, structurally predictable, and currently eating your best people's time.

When you point AI there, two things happen. Throughput goes up, and — just as importantly — the skilled clinician gets their attention back. The psychologist at Human Endeavours isn't spending three weeks typing. They're spending it with patients.

AI that ships doesn't replace the expert. It removes the part of the expert's job that was never the point.

The through-line

The fix for Human Endeavours wasn't a bigger model or a clever prompt. It was treating "write the report" as a system problem: ingest the sources, synthesise with grounding, put the human back in front of the result. Weeks became minutes, and the clinical relationship stayed exactly where it should be — between the practitioner and the person they're supporting.

If you're sitting on a workflow where the valuable people are buried in synthesis work, that's where the 15-minute win is hiding.