CREATECH NORTH: AI LEAP

KNOWLEDGE REPOSITORY

09

Operations

Model Evaluation Sprint

Compare AI models on the work your team actually does before changing a production workflow.

Outcome
A documented model selection
Time
Half day
Level
Starter

BEFORE YOU BEGIN

Set up the work.

Suggested toolkit

  • Test set
  • Models
  • Scorecard
  • Review panel

You will make

  • Representative test set
  • Weighted scorecard
  • Cost and latency notes
  • Selection rationale

STEP BY STEP

Run the workflow.

Work in sequence and keep each intermediate artifact. Review the result at every stage before spending time on the next one.

  1. 01

    Sample real work

    Create a small test set from representative tasks, difficult edge cases and known past failures. Remove sensitive information.

  2. 02

    Define the rubric

    Weight accuracy, usefulness, style, speed, cost and safety according to the workflow. Decide pass criteria before seeing results.

  3. 03

    Run blind comparisons

    Use the same inputs and conditions for every model. Hide model names from reviewers where possible to reduce preference bias.

  4. 04

    Choose with evidence

    Document results, failure patterns and operating cost. Keep the test set so the decision can be repeated when models change.

QUALITY GATE

Before you call it finished

01The output matches the stated outcome.

02Sources, settings and intermediate files are saved.

03A person has reviewed the result and edge cases.

04The next person can repeat the process.