Assessment for the AI era

AI exposed assessment. So we changed what assessment looks like.

The situations are simulated.
The reasoning is real.

Every decision carries weight and consequence. Move beyond what your learners can recall. Start assessing how they apply it.

  • Encounter
  • Consult
  • Studio
  • Insights
  • Sentinel

Built in Newcastle, Australia

A scenario, in four moments

A static quiz tests recall. A Meandrix Encounter tests judgement.

Your learner enters a simulated clinical situation. Every decision changes what happens next. You see exactly how they reasoned through it.

Question 1 of 50 of 5 answered
0:04

Scenario update

Brian Chen, 58-year-old male, admitted this morning for cardiac investigations. He reports central chest tightness rated 7/10, onset approximately 20 minutes ago, accompanied by mild nausea. Respiratory rate 18, SpO2 96% on room air. No ECG has been performed on the ward today. The ward is busy; it is late afternoon.

HR0
BP146/92
RR0
SpO20%
Pain7/10

What is the immediate priority action?

  • Obtain a 12-lead ECG↑ Recovery
  • Perform a full set of routine observations→ Maintain
  • Administer GTN sublingual tablet⚠ Critical
  • Take a full pain history from Mr Chen↓ Deterioration

Scenario updated

GTN was administered without a current blood pressure reading and without a medical officer order. Mr Chen's BP drops to 78/50. A colleague activates the emergency call system.

Critical ErrorAssessment Terminated

GTN is a potent vasodilator. NSW Health and Australian Heart Foundation guidelines require a current BP reading confirming systolic ≥ 90 mmHg before administration. No ECG has been performed, no medical officer notified, and no standing order confirmed.

Insights · Pathways

Most common routeCritical errorRecovered

01 / Open

A scenario opens. The clock is running.

The learner encounters a situation that simulates real practice. They have what a clinician on shift would have: a patient presentation, vital signs, background history, and a decision to make.

"Brian Chen, 58, admitted this morning for cardiac investigations. Background: hypertension, type 2 diabetes. He's diaphoretic and pale. Central chest tightness 7/10, onset 20 minutes ago. The ward is busy; it is late afternoon."

02 / Decide

Every choice leads somewhere different.

Every option is a defensible response a real clinician might make. There is no obvious right answer. The choice opens a pathway and the reasoning behind it becomes the data.

A correct first decision opens the optimal pathway. A reasonable but slower choice opens a delayed one with measurable consequences. A wrong choice changes the scene entirely.

03 / Consequence

Decisions have consequences. The scenario adapts.

The patient deteriorates, stabilises, or surprises based on what the learner chose to do. In a Consult, a critical clinical error ends the assessment immediately. The feedback arrives while the consequence is still on screen.

04 / See

And you see exactly how they thought.

See every decision your cohort made and where reasoning diverged. The gold line traces the most common route. Filters reveal who recovered, who deteriorated, and where the clinical thinking broke down.

What competitors don't doThe next scene rewrites itself based on how the learner arrived at it. A pathway map that knows the difference between "got there by recovery" and "got there by deterioration."

The LogicPath engine

Not just whether they passed. How they got there.

Insights starts with the cohort view: completion, score distribution, time, and pathway patterns. Open any learner record to follow the encounter from start to finish, then export the report as a PDF.

RecoveryGood choices can offset early missteps. The route back is visible, and so is who took it.
DeteriorationA reasonable but slower choice opens a delayed pathway with measurable consequences.
Viva promptsEvery learner record generates prompts you can use for a viva, review, or follow-up.

Assessment that captures the reasoning behind every decision.

The reform window

The sector responded to AI with more surveillance. That was never the real problem. A credential only means something if it reflects what a graduate can actually do under pressure. Meandrix was built to make reasoning visible.

Assessment integrity

Integrity by design. Not surveillance.

When a learner works through an Encounter in real time, there is nothing to retrieve. The design is the first line of integrity. Sentinel is the second. It covers what remains, invisible to any learner doing the work honestly.

01
Design

The assessment itself. The learner's decisions change the course of the encounter, shaping the situation, the information available, the risks that emerge, and the decisions that follow.

02
Sentinel

Opt in. Only when it matters. Adds integrity friction when the stakes require it, and produces a timestamped evidence trail for your Academic Integrity Committee.

Learn how Sentinel works

A learner who engages genuinely will never notice Sentinel is there.

That is not an accident. That is the point.

  • No webcam access. Sentinel never sees the learner.
  • No microphone access. The learner is never recorded.
  • No facial recognition or biometric identification.
  • No AI scoring of behaviour or inferred intent.
  • No location tracking. No GPS, no IP geolocation profiling.

Every session produces a timestamped integrity record. When your Academic Integrity Committee asks for evidence, you have it.

AI governance

Helpful, not autonomous.

AI has one job in Meandrix. Authoring. An educator describes the situation. AI drafts a scaffold, once. The educator edits, validates and approves it. After deployment, AI is not in the assessment loop.

  • Nevergrades a learner response or decides whether a learner passes.
  • Neverchooses which pathway a learner follows. LogicPath executes educator rules, deterministically.
  • Neversees learner submissions, attempts, names, or results.
  • Neverspeaks to a learner. No chatbot, no AI tutor, no AI examiner.
  • Nevertrains on Meandrix data. Contractually enforced.

Built for the disciplines that need it most

Where decisions actually matter.

Also relevant for medicine, allied health, psychology, and any regulated profession where judgement is the assessment.

How it works

Build it. Deploy it. See exactly how they reasoned.

01

Build

Studio is where you build. Choose Encounter, Consult, or Quiz. Use the built-in AI assistance to accelerate authoring, or take full control in manual mode. Preview both sides: what the learner sees, and the logic behind it.

02

Deploy

Meet your learners where they already are. Embed into Canvas, Moodle, or Blackboard. Share a direct link with a class. Set the window, the attempts, the time limit. Enable Sentinel when the assessment requires it.

<iframe src="meandrix.com/embed/n4xq-mp7k" />
03

See

Insights shows the cohort, then the learner. Engagement shows whether learners are practising, retrying, improving. Pathways shows the routes they took, where they recovered, and where reasoning may have broken down.

Dr Gary Crowfoot, Registered Nurse and founder of Meandrix

Built by a clinician, for clinical educators

I couldn't find it. So I built it.

I'm a Registered Nurse, founder of Meandrix, and Senior Fellow of Advance HE. I kept looking for a way to assess clinical reasoning digitally, at scale, without losing the authenticity that makes assessment meaningful.

Book a thirty-minute call. We'll build an Encounter or Consult together, using a scenario from your discipline. It's faster than most educators expect.

Dr Gary CrowfootRegistered Nurse · PhD (Nursing) · Founder, Meandrix Education Systems

Build an Encounter with me