AuriqLens

Independent AI evaluation

Does your model beat the baseline?

Compare an AI model with the approach you use today. The proposed AuriqLens pilot gives you a report your team can check, using data you control.

Synthetic example

Missed visits identified

Five appointments were missed. Each method could flag six appointments for a call.

Candidate model
4 of 5
Existing rule
3 of 5
12 fictional appointments ยท Same six-call workload

How the pilot works

Agree on a decision

Choose the model, the existing rule, and the outcome that would change what your team does. Set the test rules before looking at the results.

Compare on the same cases

Apply both methods to the same approved historical records. Keep source data in the agreed workspace and document any exclusions.

Review the evidence

Read the counts, errors, and limits together. Reviewers can use the recorded rules to repeat the calculation and check the conclusion.

Sample report

Appointment outreach

At the same six-call workload, the candidate identified one more missed visit than the existing rule in this fictional sample.

Results across 12 fictional appointments
MeasureCandidateExisting rule
Appointments flagged66
Correct flags43
False alarms23
No-shows missed12
Overall accuracy9/12 (75.0%)7/12 (58.3%)
Read the method and limits

Each method flagged a case when its supplied score was 0.50 or higher. The reported counts were recalculated independently from fictional scores and known outcomes. This exercise checks the arithmetic. It does not establish who generated the predictions, how a model performs in a real clinic, or whether an outreach call prevents a missed visit.

Pilot inquiry

What decision are you facing?

Tell us what you want to compare. A pilot starts by agreeing on the evaluation question, access, and deliverables.

Describe the decision in general terms. Keep names of individuals, customer records, and confidential model details out of this form.