Solving Hard Industrial Problems with Fast AI Deployment - with Kriti Sharma of IFS Nexus Black

Kriti introduces Nexus Black's controversial claim of delivering AI value in just three weeks by combining boots-on-ground problem discovery with elite AI engineering teams and deep domain expertise. The approach involves rapid product development using AI to build production-grade solutions, contrasting sharply with traditional two-year implementation cycles.

• •Deploy AI engineers and domain experts directly to factory floors, plants, and field sites to understand real operational challenges
• •Combine subject matter expertise with elite AI engineering teams to rapidly productize solutions
• •Use AI tools to build AI solutions faster, enabling production-grade deployment within weeks
• •Validate approach through customer results rather than vendor claims - let peers do the talking

Detailed breakdown of how Nexus Black tackled production batch losses at William Grant & Sons (Hendrix Gin, Glenfiddich Whiskey) by extracting data from piping and instrumentation diagrams, SCADA systems, and historical maintenance records. The solution addressed the fact that 38% of repair work was emergency or corrective, representing significant downtime opportunity.

• •Train AI systems to read complex piping and instrumentation diagrams (P&IDs) by combining industry expertise with AI capabilities
• •Integrate multiple disparate data sources: SCADA systems, historical work orders, sensor data (pressure, temperature, vibration)
• •Target emergency/corrective maintenance reduction to prevent unplanned downtime and batch losses
• •Leverage advanced techniques to interpret data from multiple modalities: images, video, diagrams, sensor streams

Example of how increasing complexity in seemingly simple tasks (windscreen replacement now involves advanced driver assistance systems calibration) requires rapid workforce upskilling. The case demonstrates how AI can help technicians handle more complex jobs per day while maintaining margins as vehicles become more connected and sophisticated.

• •Address workforce challenges where traditional skills (glass replacement) now require advanced technical knowledge (ADAS calibration)
• •Enable technicians to handle increasing vehicle complexity while maintaining or improving job throughput
• •Recognize that IoT integration is transforming even 'mundane' manufacturing and service tasks into tech-intensive operations
• •Focus on upskilling existing workforce rather than complete replacement

Strategic guidance on choosing AI projects that balance mission-critical impact with safe sandboxing. Production planning emerges as an ideal candidate - still largely unsolved, handles real-world chaos (rush orders, sick employees, repair emergencies), and offers opportunities to dial up autonomy gradually while keeping eyes on big-picture value.

• •Target hard, unsolved problems like production planning that account for real-world constraints and last-minute chaos
• •Start with controlled deployments and higher human intervention, then gradually increase autonomy as success builds
• •Avoid small incremental steps or generic chatbots - focus on meaningful bottom-line impact
• •Apply 'head in the clouds, feet in the mud' philosophy: dream big but solve through small, concrete steps

Discussion of how even 10% improvements in high-value workflows like downtime reduction, throughput optimization, or technician efficiency can deliver significant ROI across entire industrial environments. Emphasis on learning from peer success stories rather than vendor promises, with validation through customer testimonials at Industrial X event.

• •Focus on moving the needle 10% on critical metrics rather than solving 100% of smaller problems
• •Prioritize high-value workflows where modest percentage improvements translate to substantial operational gains
• •Validate AI approaches by listening to peer organizations and customer success stories, not vendor marketing
• •Target specific use cases in downtime reduction, production throughput, and workforce efficiency for measurable impact

Final synthesis of actionable principles for industrial operations leaders: start with frontline reality by embedding with technicians, understand where data lives in aging assets, treat domain expertise as design input, and tackle hard mission-critical workflows rather than safe experiments. Rapid impact comes from boots-on-ground methodology, not theoretical transformation programs.

• •Embed AI teams with frontline technicians to understand where data actually exists in aging infrastructure
• •Treat domain expertise as a core design input, not an afterthought in AI development
• •Begin with constrained, safe-to-sandbox use cases but focus on mission-critical workflows with real business impact
• •Shorten time to value by starting with operational reality rather than theoretical transformation roadmaps