Apollo Foods partnered with RossOps to transform shift handovers and eliminate knowledge gaps across its 24/5 beverage manufacturing operation. In just 30 days, the team achieved faster troubleshooting, improved onboarding, contractor autonomy, and a 3.2× increase in captured operational knowledge. This case study highlights how AI-enabled knowledge transfer drives productivity, reduces downtime, and boosts team confidence.
The Challenge
Apollo Foods engineers' top focus is to maintain optimum operating conditions and solve problems quickly. However, like so many other factories, they face the hidden costs of knowledge fragmentation. When all original engineers from the 7-year-old facility left, the current team started from scratch with only 2 days of handover, forcing painful first-time troubleshoots and surprise breakdowns from undocumented practices and modifications. To support knowledge transfer, Apollo used a paper-based shift handover system with 30-minute overlaps three times daily.
Specific Challenges:
- Scattered information: Time lost hunting through binders and correlating information across multiple messaging platforms.
- Communication gaps: Manual completion at end-of-shift meant details were often missed due to poor handwriting or fatigue.
- No historical insight: Inability to access past fixes or spot recurring patterns before they caused production disruption.
- Isolated expertise: No way for experts to share their reasoning, forcing teams to relearn everything independently.
The Impact
The difference between a minor stop and a multi-hour breakdown often came down to having the right information. This directly impacted OEE and team morale, as engineers felt they were constantly fighting the same battles without the benefit of lessons learned.
For instance, a recent disruption to an asset’s main compressed air supply caused by a corroded safety burst disc led to a 2am call-in, as the root cause was hidden inside pipework and not immediately apparent. Yet, with this newfound knowledge, the engineer would diagnose the issue in minutes instead of hours and avoid the callout altogether.
