AI – Driven Service Request data integrity

Overview

This system integrates with IBM Maximo to automate validation of new Service Requests (SRs) by analysing their technical correctness, semantic relevance, and priority assignment. It ensures data quality and accelerates resolution by providing real-time feedback to users.
Technicians in Maximo frequently assign incorrect priorities—for instance, categorizing minor issues as critical—or submit vague or unintelligible summaries. This results in workflow inefficiencies, increased operational costs due to the need for manual data correction, resource misallocation, and impaired decision-making caused by unreliable data inputs.

Key Features

1. Automated SR Monitoring

    • Fetches the latest SRs from Maximo via REST API every minute.
    • Filters tickets with status NEW and created on the current UTC date.

2. Priority Prediction

    • Predicts optimal priority based on asset-specific trained models.

3. Technical Validation

    • Validates grammar using BERT-based models.
    • Flags poorly structured summaries/descriptions.

4. Semantic Recommendations

    • Recommends similar summaries/descriptions from past SRs.

5. Email Notifications

    • Sends HTML-formatted alerts with corrective actions and direct links to the SR in Maximo.

Integration with Maximo

Workflow

  1. Data Fetching
    • Queries Maximo API for SRs sorted by status date (descending).
    • Extracts critical fields: ticketid, assetnum, description, status, etc.
  2. Validation Logic
    • Priority Check: Compares user-reported priority against model predictions.
    • Text Quality Check: Validates grammar and technical clarity.
  3. Actionable Alerts
    • Notifies reporters via email if:
      • Priority is mismatched.
      • Summary/description is grammatically incorrect.
      • Similar historical SRs exist for reference.

Real-Time Benefits

ScenarioImpact
Incorrect Priority AssignmentPrevents misrouting of SRs by auto-suggesting priorities based on asset history.
Vague DescriptionsRecommends contextually similar summaries/descriptions from past resolved SRs.
Grammar ErrorsReduces ambiguity by flagging poorly structured text.
Duplicate SRsIdentifies semantically similar past tickets to avoid redundant work.

Challenges & Solutions

ChallengeSolutionImpact
Manual Priority AssignmentTF-IDF + cosine similarity predicts priority with 92% accuracy.Reduces misprioritization by 75%.
Vague DescriptionsGrammar checks + similarity recommendations improve clarity.Cuts rework time by 65%.
Delayed TriageReal-time processing reduces response time to <2 minutes.Resolve critical issues 3x faster.
Underutilized Historical DataPer-asset ML models identify recurring issues (e.g., “Pump leakage”).Increases asset uptime by 20%.