It helps clinic leaders and administrators understand how AI is being used, where it is performing well, where it needs clarification, where users are giving feedback, and which interaction patterns may require improvement.

This is not just usage reporting. It is the control room for AI accountability.

Use AI Assistant Analytics to track assistant performance, review failed or unclear responses, monitor user satisfaction, inspect past requests, and identify learning opportunities that can improve future AI behavior.

What AI Assistant Analytics Does

AI Assistant Analytics helps your clinic:

• Track total AI assistant interactions

• Measure success rate

• Measure first-try success

• Review requests that needed clarification

• Review failed AI requests

• Monitor average response time

• Track positive and negative feedback

• Review interaction history

• Inspect user messages and AI responses

• Identify common issues and categories

• Detect learning opportunities

• Support AI awareness, feedback, and auditability

This is where the clinic can see whether AI is actually helping or just wearing a lab coat and waving at buttons.

Main Areas

AI Assistant Analytics includes:

• Overview metrics

• Distribution

• Quick Insights

• Learning Opportunities

• Interaction history

• Expanded interaction details

• Feedback indicators

• Date range controls

• Refresh controls

Date Range

The date range selector controls which AI interactions are included in the analytics view.

Common options may include:

• Last 7 days

• Last 30 days

• Last 90 days

• Custom range

Changing the date range updates the dashboard metrics, interaction history, feedback totals, and learning opportunity analysis.

Refresh

Refresh reloads the latest AI analytics data.

Use Refresh when:

• New AI interactions were recently completed

• Feedback was added

• You changed the date range

• A recent issue needs to appear in reporting

• The dashboard appears stale

Overview

The Overview section shows the top-level health of AI assistant performance.

It includes:

• Total interactions

• Success rate

• First try success

• Average response time

• Needs clarification

• Failed requests

• Positive feedback

• Negative feedback

This gives leaders a quick read on whether the assistant is working smoothly or creating friction.

Total Interactions

Total Interactions is the number of AI assistant requests during the selected date range.

An interaction may include:

• Asking a business question

• Asking for patient information

• Requesting appointment data

• Asking for revenue or labor metrics

• Asking the AI to create or retrieve information

• Asking for operational support

• Asking for scheduling help

Total interactions show adoption and usage volume.

High usage is not automatically good. The useful question is whether those interactions are successful.

Success Rate

Success Rate shows the percentage of AI interactions that completed successfully.

A successful interaction means the assistant was able to process the request and provide a usable response or action.

Success Rate helps identify whether the AI is reliable enough for daily workflows.

First Try Success

First Try Success shows how often the assistant succeeded without needing clarification or correction.

This is one of the most important quality metrics.

A high first-try success rate means users are getting useful answers quickly.

A low first-try success rate may mean:

• User prompts are unclear

• AI instructions need improvement

• Data access is incomplete

• Tool routing needs adjustment

• The assistant is asking too many follow-up questions

• The clinic’s knowledge or settings need cleanup

First-try success is the difference between helpful AI and a vending machine that asks philosophical questions before giving you chips.

Average Response Time

Average Response Time shows how long the assistant takes to respond.

This may be shown in seconds or milliseconds.

Use this to monitor speed and workflow friction.

Slow responses may indicate:

• Complex requests

• Tool delays

• Data retrieval issues

• Model latency

• Integration bottlenecks

• Large context processing

Needs Clarification

Needs Clarification shows interactions where the assistant could not confidently answer or act without more information.

Examples:

• The user asked for “that patient” without naming one

• The assistant needed a date range

• Multiple clients or patients matched the request

• The request was too broad

• The assistant needed confirmation before taking action

Clarification is not always bad. Sometimes it protects the clinic from wrong actions.

But too much clarification means the system is creating drag.

Failed

Failed shows interactions the assistant could not process.

Failures may happen when:

• A tool or integration fails

• Required data is missing

• The request is unsupported

• The assistant cannot access the needed record

• The system returns an error

• The request exceeds current permissions

• The AI cannot safely complete the action

Failed interactions should be reviewed because they often reveal workflow gaps, data issues, or system bugs.

Positive Feedback

Positive Feedback shows how many AI responses received positive user feedback.

This may be represented by a thumbs-up icon.

Positive feedback helps identify what the assistant is doing well.

Examples:

• Correct answer

• Useful summary

• Good action recommendation

• Fast response

• Accurate data retrieval

• Helpful workflow completion

Negative Feedback

Negative Feedback shows how many AI responses received negative user feedback.

This may be represented by a thumbs-down icon.

Negative feedback helps identify where AI needs improvement.

Examples:

• Wrong answer

• Missing information

• Slow response

• Confusing response

• Failed request

• Wrong category

• Needed too much clarification

• Did not understand the user’s intent

Negative feedback is not failure by itself. It is training smoke.

Distribution

The Distribution section shows issue and category breakdowns.

Use this section to understand where AI activity is concentrated.

Distribution may show:

• Common request categories

• Common issue types

• Failed interaction clusters

• Clarification-heavy categories

• Feedback patterns

If no data is available, the section may show No data available yet.

Issue Breakdown

Issue Breakdown groups interactions by problem type.

Examples may include:

• Failed request

• Clarification needed

• Slow response

• Missing data

• Unsupported action

• Ambiguous prompt

• Tool error

• User correction

This helps teams identify where the assistant needs improvement.

Category Breakdown

Category Breakdown groups interactions by request category.

Examples may include:

• General

• Appointment

• Patient

• Client

• Revenue

• Labor

• Transactions

• Medical records

• Workflow

• Scheduling

This helps leadership see what users are asking AI to do most often.

Quick Insights

Quick Insights summarize the most important AI performance signals.

This area may include:

• Most common issue

• Most common category

• Feedback received

• User satisfaction

Most Common Issue

Most Common Issue shows the issue type that appears most often during the selected period.

If no issues are detected, this may show None.

Use this to prioritize improvement work.

Most Common Category

Most Common Category shows the category with the most AI interactions.

Example:

General

This helps identify where users rely on AI most.

Feedback Received

Feedback Received shows how many feedback responses were submitted.

This includes positive and negative feedback.

Low feedback does not always mean users are satisfied. It may simply mean users are not using feedback buttons.

User Satisfaction

User Satisfaction summarizes feedback sentiment when enough feedback exists.

This may show a score, percentage, or N/A when there is not enough data.

Learning Opportunities

Learning Opportunities are detected patterns that could improve AI performance.

This section may identify:

• Repeated failed requests

• Common unclear prompts

• Categories with poor success rates

• Slow response patterns

• Repeated negative feedback

• Missing data areas

• Tool or integration issues

• Knowledge gaps

• Workflows where AI needs better instructions

Learning Opportunities are where analytics turns into improvement.

Analyze

Analyze runs a review of AI interaction patterns to detect possible improvement areas.

Use Analyze when:

• Failures are increasing

• Users are giving negative feedback

• Clarification is high

• A new workflow was launched

• You want to review AI performance

• The clinic added new knowledge or settings

No Learning Opportunities Detected

This means the system did not detect clear improvement patterns for the selected period.

This may mean:

• AI is performing well

• There is not enough interaction data yet

• Feedback volume is too low

• The selected date range is too narrow

Interaction History

The interaction history shows individual AI assistant requests.

Each row may include:

• Date

• Time

• Status

• User message

• Category

• Response time

• Feedback

• Expand / collapse control

This section gives the clinic audit-level visibility into what users asked and how AI responded.

Date and Time

The date and time show when the interaction happened.

This helps review activity by day, shift, user behavior, or incident timeline.

Status

Status shows the result of the AI interaction.

Common statuses include:

• Completed

• Failed

• Clarify

Completed

Completed means the assistant successfully processed the request and returned a response or completed the action.

Completed interactions may still receive negative feedback if the user did not find the result useful.

Completed does not always mean perfect. It means the request was technically processed.

Failed

Failed means the assistant could not process the request.

A failed row should be reviewed when failures repeat or affect important workflows.

Clarify

Clarify means the assistant needed more information before answering or acting.

Examples:

• Missing date range

• Missing patient name

• Multiple possible matches

• Ambiguous request

• Action required confirmation

Clarify status protects against wrong action, but too much clarification slows the team down.

User Message

User Message shows what the user asked the assistant.

Examples:

• “What is our revenue this week?”

• “How efficient is our labor?”

• “Can you book a follow-up appointment?”

• “What patients have we seen in the last 2 weeks?”

• “Give me information on Macavity.”

Reviewing user messages helps identify real-world usage patterns.

Category

Category identifies the type of request.

Examples:

• General

• Appointment

• Patient

• Revenue

• Labor

• Workflow

Categories help leadership understand where AI is being used and where it may need more support.

Response Time

Response Time shows how long the AI took to process the request.

This may appear in milliseconds.

Use this to identify slow categories, slow tools, or complex request types.

Feedback Icons

Feedback icons show whether users rated the response positively or negatively.

• Thumbs up means positive feedback

• Thumbs down means negative feedback

• A dash means no feedback was submitted

Feedback helps the clinic understand whether responses were useful, not just whether they completed.

Expand Interaction

The expand control opens the full details of an AI interaction.

Use this when reviewing a specific request, failed interaction, or feedback issue.

Expanded Interaction Details

Opening an interaction shows deeper audit detail.

This may include:

• User message

• AI response

• User

• Clarification rounds

• Status

• Category

• Response time

• Feedback

This is the auditability layer.

It allows administrators to understand exactly what was asked, what the assistant answered, who asked it, and whether the request needed clarification.

User Message Detail

This shows the full prompt submitted by the user.

Use this to determine whether the issue came from the assistant, the available data, or the way the question was asked.

AI Response

This shows the assistant’s response.

Examples may include:

• A completed answer

• A partial answer

• A clarification request

• An error message

• A workflow result

Reviewing AI responses is critical when investigating user feedback or failed interactions.

User

This shows which user made the request.

Use this for accountability, coaching, support, and audit review.

Clarification Rounds

Clarification Rounds show how many follow-up clarification steps were needed.

For example:

• 0 means the assistant answered on the first try

• 1 means one clarification was needed

• 2 means two clarification rounds were needed

High clarification rounds may indicate unclear prompts, missing clinic data, or AI workflow issues.

Awareness

AI Awareness means the system provides visibility into AI activity instead of hiding it.

In PawthosX One, AI should be observable.

Admins should be able to see:

• What users asked

• What AI answered

• Whether the request succeeded

• Whether clarification was required

• Whether the interaction failed

• How long the response took

• Which category the request belonged to

• Whether feedback was positive or negative

Awareness prevents AI from becoming a black box in the clinic.

Feedback

Feedback is how users tell the system whether an AI response was useful.

Feedback may be submitted through thumbs-up or thumbs-down indicators.

Feedback helps identify:

• Good responses

• Bad responses

• Missing context

• Poor wording

• Wrong answers

• Data access issues

• Training opportunities

• Workflow friction

Feedback should be treated as operational signal, not decoration.

Positive Feedback Workflow

When users give positive feedback, it may indicate:

• The assistant answered correctly

• The response was useful

• The action worked

• The result saved time

• The user trusted the answer

Positive feedback helps identify strong patterns to preserve.

Negative Feedback Workflow

When users give negative feedback, admins should review:

• What the user asked

• What the assistant answered

• Whether the answer was factually wrong

• Whether data was missing

• Whether the assistant misunderstood the request

• Whether the request should have triggered a different workflow

• Whether the user needed better prompt guidance

Negative feedback should become improvement work, not dashboard confetti.

Auditability

Auditability means AI interactions can be reviewed after they happen.

This matters because AI may affect clinic workflows, client communication, records, scheduling, operations, and business reporting.

AI Assistant Analytics supports auditability by showing:

• User messages

• AI responses

• User identity

• Interaction status

• Category

• Response time

• Clarification rounds

• Feedback

• Error states

Auditability gives the clinic a record of AI behavior.

This is how PawthosX One keeps AI accountable, reviewable, and operationally safe.

Common Workflows

Review AI Performance

1. Open AI Assistant Analytics.

2. Select the desired date range.

3. Review Total Interactions, Success Rate, and First Try Success.

4. Check Failed and Needs Clarification.

5. Review Positive and Negative Feedback.

6. Open Learning Opportunities if patterns appear.

Investigate a Failed Request

1. Open AI Assistant Analytics.

2. Find the failed interaction.

3. Expand the row.

4. Review the user message.

5. Review the AI response.

6. Check category and response time.

7. Determine whether the issue was missing data, unsupported action, tool failure, or unclear prompt.

Review Negative Feedback

1. Open the interaction history.

2. Locate rows with thumbs-down feedback.

3. Expand the interaction.

4. Compare the user message with the AI response.

5. Identify whether the issue was accuracy, usefulness, speed, or missing context.

6. Add the issue to improvement work if needed.

Monitor Clarification Issues

1. Review the Needs Clarification metric.

2. Open interactions marked Clarify.

3. Look for repeated patterns.

4. Determine whether users need clearer prompt guidance or whether the AI needs better context.

5. Update knowledge, settings, or workflows if needed.

Run Learning Analysis

1. Open the Learning Opportunities section.

2. Select Analyze.

3. Review detected patterns.

4. Prioritize repeated issues.

5. Update AI configuration, clinic knowledge, workflow rules, or system access as needed.

Best Practices

Use AI Assistant Analytics as an ongoing quality system.

• Review failures weekly.

• Review negative feedback weekly.

• Watch first-try success closely.

• Treat clarification rounds as friction.

• Use Learning Opportunities to improve workflows.

• Check response time when users report slowness.

• Do not ignore repeated “general” category failures.

• Audit important AI interactions when they affect records, scheduling, clients, or revenue.

The goal is not perfect AI. The goal is visible, accountable AI that gets better instead of getting weird in a corner.

Final Definition

AI Assistant Analytics is the awareness, feedback, and auditability layer for AI inside PawthosX One.

It shows how AI is being used, how well it is performing, where users are giving feedback, where failures are happening, and where the system can improve.