Create scorecards for Voxjar's AI evaluator exactly as you would for your QA team. No keyword searches, complex queries, or data science team needed.

Every scorecard made in Voxjar is automatically compatible with our AI evaluator for automated quality assurance.

To create a new scorecard, go to the Scorecards page (app.voxjar.com/scorecards).

Once there, you can create a new scorecard in three ways.

Once a scorecard is created you can assign them to evaluators in your Queues or when you generate a one-off AI evaluation.

​
​

If you have an existing scorecard, you can duplicate it to create a new version.

This is usually preferred over deleting or adding questions to your existing scorecard to keep historical scores aligned with the scorecard's current state.

It is also a big time saver when launching similar scorecards.

Click on your existing scorecard in the scorecards list, click the edit button on the preview window, then select "Duplicate" from the drop down menu.

As you're building your scorecard you can test it with Voxjar's AI evaluator.

From the scorecard options menu, click the "AI Test" button.

Voxjar's built in test loop shows you what the AI's responses will look like before launching your Auto QA program.

Check out our guide on how to test your scorecard with the AI evaluator.

And of course, you can build a new scorecard from scratch.

This is usually the path taken when you have an existing scorecard and want to migrate it into Voxjar.

From the scorecards list click "Start from Scratch" and you're off to the races.
​

A new scorecard will initially load as untitled with a placeholder section and question.

To edit your scorecard name, passing criteria, main scorecard prompt, or AI model click on the scorecard header that shows the title and description.

The scorecard editor panel will slide in from the right of your screen.

Scorecard sections are used to group your questions so it's easier to navigate your form.

You can label your sections, and add a description.

Clicking on a section will open the section editor panel on the right side of your screen.
​

The questions you create on your call monitoring forms are the questions that will be shown to your QA team or the AI evaluator for review.

Questions are grouped in Sections. They can be dragged and dropped into new orders and into different Sections.

The question editor panel on the right side of your screen lets you define the question in detail and control the possible answers.

Questions can be answered with multiple choice, single selection, or sliding scale responses that you define. Single selection is the most common response type followed by sliding scale.

All of the options you select for each question are combined as prompts for the AI evaluator. Giving you the ability to tweak and tune for the best AI responses

Every existing question on your saved scorecards is available in your Question Library for quick duplication.

The question is what your human call evaluators or Voxjar's AI evaluator will attempt to answer based on a provided call recording.

It's usually best to ask detailed questions instead of relying on industry specific knowledge or jargon.

You can provide general examples of what you would expect to hear on a phone call.

These will be displayed on the scorecard for human evaluators and are also provided to the AI evaluator to better clarify what you want.

Single selection answers limit the response to one answer.

By default we label the placeholders "yes" and "no" but you can replace them with anything that makes sense.

Each answer can be assigned different point values and be marked as an autofail.

A sliding scale lets the evaluator answer a question with any number between the min and max.

This is especially useful for subjective questions like "How polite was the agent?" etc.

Each number on the scale will be counted as a point on the scorecard.

You can set a threshold for autofails.

Multiple choice answers allow the selection of multiple answers.

The points of each selected answered will be added to the overall score of your scorecard.

Every question on your scorecard has an N/A option.

If you want a question to default to N/A you can select that above your answer options.

When a question is marked as N/A it is excluded from the score calculation of that scorecard.

A scorecard response with a question marked as N/A can still receive a 100% score.

An answer marked as auto-fail will mark the entire scorecard as a fail.

Scorecard actions let your Ai evaluator automatically trigger follow up actions based on its responses, the scorecard's point values, pass/fail, or on every submission.

Your Ai can trigger the following actions based on responses to your scorecards:

Trigger actions based on the outcome of the entire scorecard.

To create a Scorecard Trigger, click "Add Action" from the scorecard header card at the top of the scorecard editor.

Choose your action:

And choose your trigger:

Each will have different rules that you can set to determine which evaluations trigger your selected action.

Trigger actions based on the response to a specific question.

To create a Question Trigger, click "Add Action" from the card of the question that you want to trigger the follow up action.

Choose your action:

And then select the conditions that will trigger your follow up action. For Question Triggers the available conditions will related to the type of question. A Single Selection will have you select which answer will trigger the action, Multiple Choice allows exact match or partial match triggers, and Sliders let you set a range.