AI-Generated Insights

Flow Myna's AI continuously analyzes your process to automatically identify bottlenecks, anomalies, and optimization opportunities. No complex queries or statistical analysis required—just open the Insights page to see what the AI discovered.

What Are AI Insights?

AI Insights are automated analyses that surface important patterns, problems, and opportunities in your process. Think of them as having an expert process analyst continuously reviewing your data and highlighting what matters most.

Why AI Insights Matter

Without AI Insights, you'd need to:

• Manually examine duration statistics for every transition
• Calculate performance metrics across variants
• Identify outliers and anomalies
• Compare different process segments
• Formulate improvement hypotheses

With AI Insights, the platform:

• Automatically runs comprehensive analysis
• Identifies the most significant findings
• Explains what's happening in plain language
• Provides actionable recommendations
• Generates starter questions for deeper exploration

Example Insight

Insight: Bottleneck in Manager Approval Step

The transition from Risk Assessment to Manager Approval takes an average of 4.2 days, which is 3.5x longer than other approval steps. This affects 175 loans (70% of cases).

Impact:
- Adds 2.8 days to overall process time
- Creates backlog during month-end periods
- 15% of cases exceed SLA due to this delay

Recommendations:
- Implement automated approval for low-risk cases under $30K
- Add additional approval capacity during peak periods
- Review approval criteria to reduce manual review requirements

Explore Further:
- "Show me cases delayed by this bottleneck"
- "Compare approval times across different loan amounts"
- "What's the approval backlog over time?"

How Insights Are Generated

The Insights Generator runs automated analysis combining multiple data sources:

Analysis Pipeline

1. Process Graph Analysis

• Examines all transitions
• Calculates duration statistics
• Identifies slow paths
• Detects anomalous patterns

2. Variant Analysis

• Compares variant performance
• Identifies rework patterns
• Detects conformance issues
• Measures variant frequency changes

3. Attribute Influence

• Analyzes which attributes drive routing decisions
• Identifies factors affecting duration
• Detects patterns in approvals/rejections
• Measures impact of different attribute values

4. Edge-Level Analysis

• Calculates "lift" for attribute values
• Identifies paths that specific cases take
• Measures deviation from expected paths

5. LLM Synthesis

• Converts technical metrics to business language
• Generates plain-English explanations
• Creates actionable recommendations
• Formulates follow-up questions

When Insights Are Generated

Automatic Generation:

• When you first create a project
• After uploading new data
• On a scheduled basis (daily/weekly)

Manual Generation:

• Click "Generate New Insights" anytime
• Useful after applying filters or making changes
• Re-analyzes based on current data view

Types of Insights

Flow Myna generates several categories of insights:

Bottleneck Insights

What they identify: Activities or transitions that slow down the process

Example:

Title: Risk Assessment Processing Delay
Type: Bottleneck
Severity: High

Finding: Risk Assessment takes an average of 3.2 days, compared to 0.5 days for similar assessment steps.

Impact: Affects 80% of cases, adding an average of 2.7 days to total cycle time.

Causes:
- Manual review required for all cases
- Limited analyst capacity (only 3 reviewers)
- No prioritization mechanism

Recommendations:
- Implement risk-based triage (auto-approve low-risk)
- Add capacity or redistribute workload
- Create fast-track for time-sensitive cases

Optimization Opportunities

What they identify: Changes that could improve performance

Example:

Title: Parallel Processing Opportunity
Type: Optimization
Severity: Medium

Finding: Credit Check and Employment Verification currently run sequentially but could be parallelized.

Impact: Potential time savings of 0.8 days per case (195 cases affected).

Current Flow:
Application → Credit Check (2h) → Employment Verification (1.5h) → Next Step

Optimized Flow:
Application → [Credit Check + Employment Verification in parallel] (2h max) → Next Step

Savings: 1.5 hours per case

Anomaly Insights

What they identify: Unusual patterns or unexpected behavior

Example:

Title: Approval Before Credit Check Pattern
Type: Anomaly
Severity: Low

Finding: 8 cases (3%) were approved before completing credit checks.

Risk: Potential compliance issue or data quality problem.

Cases: L-156, L-203, L-287, L-341, L-445, L-502, L-671, L-788

Investigation Needed:
- Verify if this is a data error or actual process deviation
- Check if these are special case types
- Review compliance requirements

Attribute Influence Insights

What they identify: Factors that drive process routing or outcomes

Example:

Title: Loan Amount Drives Approval Path
Type: Attribute Influence
Severity: Medium

Finding: Loan amount significantly influences which approval path is taken.

Patterns:
- Loans under $25K: 95% auto-approved (avg 1.2 days)
- Loans $25K-$75K: 80% standard review (avg 4.1 days)
- Loans over $75K: 100% require director approval (avg 7.3 days)

Insight: Clear threshold-based routing exists. Consider if these thresholds are optimal for current risk tolerance.

Rework Insights

What they identify: Cases that loop back or repeat steps

Example:

Title: Document Resubmission Loop
Type: Rework
Severity: High

Finding: 45 cases (18%) required document resubmission, adding an average of 5.2 days.

Common Reasons:
- Missing signatures (60% of rework cases)
- Incomplete employment info (25%)
- Outdated documents (15%)

Recommendations:
- Add upfront document checklist
- Implement real-time validation during upload
- Provide clearer requirements in initial communication

Reading Insight Details

Click any insight to see the full analysis:

Insight Structure

Header Section

• Insight title
• Type and severity
• Date generated
• Cases affected

Finding Description

• What the AI discovered
• Plain-language explanation
• Why it matters

Impact Analysis

• How many cases affected
• Time or cost impact
• Frequency of occurrence
• Business implications

Supporting Data

• Relevant statistics
• Comparison metrics
• Visualizations (charts, graphs)
• Example cases

Recommendations

• Specific, actionable suggestions
• Priority ordering
• Expected impact
• Implementation considerations

Explore Further

• Starter questions for AI Co-Pilot
• Related insights
• Links to relevant data views

Acting on Recommendations

Insights are only valuable if you act on them:

Priority Framework

High Severity + High Frequency = Top Priority

• Affects many cases
• Significant impact
• Clear improvement path
• Act immediately

High Severity + Low Frequency = Investigate

• Serious when it occurs
• But rare
• Understand root cause first
• Plan mitigation

Low Severity + High Frequency = Optimize

• Small impact per case
• But affects many
• Cumulative benefits significant
• Good candidate for automation

Low Severity + Low Frequency = Monitor

• Minor impact
• Rare occurrence
• Keep an eye on it
• Address if pattern worsens

Implementation Steps

1. Validate the Finding

• Use AI Co-Pilot to explore further
• Apply filters to see affected cases
• Verify the data supports the insight

2. Understand Root Causes

• Ask "Why does this happen?"
• Examine edge cases
• Talk to process owners

3. Plan Improvements

• Consider recommendations
• Assess feasibility
• Estimate resources needed
• Define success metrics

4. Implement Changes

• Make process modifications
• Update systems or procedures
• Train team members

5. Measure Impact

• Generate new insights after changes
• Compare before/after metrics
• Verify improvements achieved
• Document learnings

Generating New Insights

Trigger fresh analysis anytime:

When to Regenerate

• After significant new data uploads
• When you've made process changes
• To analyze a specific time period
• After applying filters to focus analysis

How to Generate

1. Go to Insights page
2. Click "Generate New Insights"
3. AI runs full analysis (takes 1-2 minutes)
4. New insights appear, replacing old ones
5. Previous insights are archived

Insight History

Access past insights to:

• Track how issues evolved
• Measure improvement over time
• See which recommendations were acted on
• Compare process performance across periods

Best Practices

Review Insights Regularly

• Weekly: Quick scan of new high-severity insights
• Monthly: Deep dive into all insights, plan actions
• Quarterly: Review archived insights, measure progress

Prioritize Ruthlessly

Don't try to fix everything at once:

• Focus on 1-2 high-impact improvements
• Complete them fully before moving on
• Measure results to confirm improvement

Combine with Co-Pilot

Use insights as starting points:

• Click "Ask Co-Pilot" from insight detail
• Explore underlying data
• Test hypotheses
• Validate recommendations

Share with Stakeholders

• Export insights for presentations
• Discuss in team meetings
• Use to justify process changes
• Track implementation progress

Next Steps

Insights identify opportunities—now explore them in depth:

Ask the AI Co-Pilot

• AI Co-Pilot - Explore insights conversationally

Investigate on Process Map

• Process Map - Visualize the bottlenecks and patterns
• Process Variants - Compare performance across paths

Apply Filters

• Filters & Exploration - Drill down into specific scenarios