The Business

A precision engineering company in the West Midlands producing custom metal components for automotive and aerospace clients. 45 employees. Three CNC machining cells, manual assembly area, and quality inspection. Annual turnover £3.2 million. ISO 9001 certified.

The Problem

Quality Issues Detected Too Late

Quality inspection happened at the end of the production process. When a defect was found, the entire batch was suspect — meaning rework, scrap, and delays. Their defect rate was 4.2%, which sounds small until you realise each scrapped precision component costs £35-180 in materials and machine time. Monthly scrap cost: approximately £3,800.

Production Scheduling by Instinct

The production manager scheduled jobs using a whiteboard and 20 years of experience. It worked — until it didn't. Rush orders caused chaos. Machine changeovers weren't optimised. Jobs were grouped by customer rather than by tooling, meaning unnecessary setup time.

Maintenance Surprises

CNC machines went down without warning. Unplanned downtime cost approximately £800/hour in lost production. The maintenance schedule was calendar-based (every 3 months) rather than condition-based. Some machines were being serviced too often, others not enough.

What We Built

Phase 1: Quality Prediction (Weeks 1-3)

Connected sensors on the CNC machines to an AI monitoring system that:

• Monitors cutting parameters in real time (spindle load, vibration, temperature, tool wear)
• Compares current parameters against patterns from successful previous runs
• Flags anomalies that correlate with historical defects
• Alerts operators before a bad part is completed — not after

The AI learned what "good" looks like for each component type. When parameters deviate from the known-good envelope, operators get an early warning.

Phase 2: Production Scheduling Optimisation (Weeks 3-5)

Built an AI scheduling system that considers:

• Job priorities and delivery dates
• Tooling commonality between jobs (minimising changeover time)
• Machine capabilities and current loading
• Material availability and lead times
• Staff skills and shift patterns
• Historical run times for accurate completion estimates

The system generates optimised schedules that the production manager reviews and adjusts. It doesn't replace his judgement — it gives him a better starting point.

Phase 3: Predictive Maintenance (Weeks 5-6)

Using the same sensor data from the quality monitoring system, built maintenance prediction models:

• Tool wear prediction based on actual cutting conditions, not just time
• Bearing and spindle degradation trending
• Maintenance scheduling based on actual machine condition, not calendar intervals
• Parts ordering triggered by predicted maintenance needs

Results After 4 Months

• Defect rate reduced from 4.2% to 2.5% — a 40% improvement
• Scrap costs down from £3,800 to £2,200/month
• 15% improvement in machine utilisation from better scheduling
• Unplanned downtime reduced by 55%
• On-time delivery improved from 88% to 96%
• Changeover time reduced by 22% from tooling-optimised scheduling

ROI

Implementation cost: £12,800 (including sensor hardware). Monthly platform and maintenance: £720.

Monthly benefit: approximately £4,200 (scrap reduction + improved utilisation + reduced downtime).

Payback period: 12 weeks.

What Made It Work

• Existing data was gold. The CNC machines already generated detailed parameter logs — they just weren't being analysed. Most of the "AI magic" was connecting existing data to intelligent analysis.
• Shop floor buy-in. Operators were initially wary of monitoring. Framing it as "the AI helps you catch problems early" rather than "the AI watches you" made the difference.
• Production manager stayed in control. The scheduling AI recommends; the human decides. His experience with customer relationships, informal priorities, and gut feelings about machine reliability still matters.
• ISO 9001 alignment. Every AI decision is logged and traceable, which actually improved their quality management documentation.

Related: Read about our AI data analysis and process automation services. See other case studies: retail, professional services, hospitality.