Tap On
Smarter Installation Checks with AI Vision: How might we use AI to automatically interpret meter installation photos - starting with identifying stop tap status - to improve quality accuracy, speed, and consistency across the business while reducing financial risk.
Post-installation boundary box checks are critical for ensuring meters are functional, water flow is restored and customers receive the right service. However, verifying whether a stop tap is turned back falls on a single Quality Manager for manual review of installation photos—making the process time-consuming, inconsistent over regions, and prone to error due to the mass volume which is ever increasing.
The main error often being Interruptions to Supply (ITS), leaving our customers without water until the issue is resolved. This can lead to increased operational costs, delayed decision-making, and potential Guaranteed Standards Scheme (GSS) exposure.
This sprint will explore how AI-powered image recognition can transform post-installation quality validation processes.
Together, we will:
- Define how AI can reliably identify stop tap positions from installation images and feedback which installations need to be revisited
- Explore how this capability can be expanded to validate meters, fittings, and valves and reduce the Quality Managers workload
- Identify the data, image quality, and standards required to train effective models
- Design workflows that integrate AI outputs into operational processes
- Assess the potential impact on cost reduction, efficiency, and compliance
- Identify other areas in which this AI has capability
Participants will collaborate through structured design and prototyping sessions to:
- Map the current installation and validation process
- Identify pain points in manual photo review and decision-making
- Explore AI approaches for image classification and object detection
- Define requirements for data capture standards in the field
- Prototype end-to-end workflows from photo capture to automated decision output
- Evaluate risks, accuracy thresholds, and governance considerations
By the end of the sprint, we aim to deliver:
- A concept for an AI-driven installation validation tool
- Defined requirements for image capture and data quality
- Use cases for scaling AI validation across multiple asset types
- A proposed workflow integrating AI into operational systems
- A roadmap for pilot, testing, and deployment
- Field and installation teams, with clearer guidance and reduced rework
- Operational teams, benefiting from faster, more consistent validation
- Customer teams, through reduced delays and improved service outcomes
- Digital and AI teams, applying scalable solutions to real-world problems
- The wider business, through reduced cost, risk, and improved compliance
Any questions: contact - Charlea Pratt: charlea.pratt@nwl.co.uk