Decades before the entry of artificial intelligence (AI), manufacturing innovation occurred incrementally, whether through mechanical enhancements or digital upgrades. Manufacturing companies have since discontinued the incremental model and have switched to using AI not just to upgrade but to optimize how they operate and make decisions.
A case in point is CITIC Pacific Special Steel, a Chinese steel manufacturer that uses AI to consistently monitor the internal workings of its blast furnaces. It leverages AI-generated data to adjust its process parameters across factories, thereby increasing throughput. It now produces 15% more steel than before while consuming 11% less energy.
This kind of example, where companies are using AI to boost efficiency and value from their operations is impressive but rare. Many companies that employ AI are unable to move beyond targeted pilots or individual AI use cases, such as predictive analysis or schedule optimization.
A McKinsey survey reports that many companies are still stuck at the “exploration” or “targeted-implementation” stages, with only 2% of COOs reporting that AI is fully embedded across operations. As such, most manufacturers are struggling to make the jump from one-off AI pilots to enterprise-wide AI scaling.
In this post, we’ll explore the gap manufacturers face when implementing AI and look at what high-performing companies are doing differently.
What Are the Barriers to Enterprise-Wide AI Implementation?
Using AI in manufacturing is not about isolated deployments, such as predicting equipment failure or using computer vision to improve quality control. Meaningful AI innovation happens when it is integrated into operations. Yet several barriers stand in the way of successfully scaling AI.
Siloed Systems
Companies limit the use and impact of AI by treating it as a set of isolated tools for quick wins, rather than a cohesive system that can transform operations. Shortsighted pilot projects can lead to localized achievements on the factory floor. However, unless AI becomes part of a connected system or broader workflow, companies will be unable to fully extract the value it was meant to provide.
For example, a factory using predictive analysis can successfully forecast that a particular machine needs maintenance. However, since AI doesn’t drive the rest of the systems, they are not notified in time about the maintenance event.
There is also no knowledge of inventory stockouts, so spare parts aren’t available on time. Or, teams may be unable to adjust production schedules in time. These disparate systems cause the machine to break down, despite initial successful predictions.
Weak Data Quality
AI is only as powerful as the data that feeds it. That is why it’s challenging to scale an AI system on a foundation of inconsistent or inaccurate data. This is a major barrier that prevents companies from adopting AI models successfully into operations.
According to a McKinsey report, 46% of COOs believe that limitations in their data and IT/OT systems are major roadblocks to company-wide AI transformations. Of these, 19% cite obsolete infrastructure, while 18% cite poor data quality as an impediment to building scalable AI systems.
Gaps in Training and Lack of AI-Skilled Workforce
Simply introducing AI models into workflows isn’t enough to mobilize manufacturing operations. Companies also need to effectively train employees, shop-floor technicians, and supervisors who will work with AI. However, there is evidence to the contrary. According to a Boston Consulting Group (BCG) report, only 51% of frontline employees regularly use AI in their workflows, even though more than 75% of leaders claim to have integrated generative AI into their processes.
To ensure the manufacturing workforce is ready for AI adoption, leaders need to reshape their culture to make it more AI-ready through:
- Reskilling and training sprints that provide practical, hands-on training with AI applications
- Moving beyond external vendor reliance to building internal expertise, so that knowledge stays within the organization
According to a McKinsey report, at least three-quarters of COOs plan to adopt a hybrid “build-buy-partner” model, which involves working with tech partners and also strengthening in-house capabilities through AI training.
How to Scale AI for a Company-Wide Transformation
1. Reshape Workflows From The Ground Up
High-performing companies are three times more likely to “fundamentally redesign their workflows during AI deployment,” according to a McKinsey report. This shows that industry players who have successfully adopted AI across company systems take a ground-up approach, i.e., they avoid integrating isolated AI tools into existing systems.
For example, Hyundai’s HMGICS smart factory uses a “digital twin” to mimic its real-time processes. In doing so, it helps guide root-cause analyses and detects defects proactively, creating continuous feedback loops in core operations.
In another scenario, pharmaceutical company Novartis adopted a fully automated production line at its Swiss facility. It used Internet of Things (IoT) sensors in its machines to gather real-time data, which it analyzed using real-time analytics. As a result, it saw a 30% decrease in downtime and even released batches 60% faster than before.
2. Treat Employees and AI Training As Strategic Priorities
Culture change and reskilling are significant barriers to scaling AI, and high-performing companies acknowledge this fact. As such, they’re more proactive about championing AI use and often lead by example. Operations leaders at such companies set clear expectations for AI adoption and invest in training even non-technical employees about AI engagement.
The study of a pharmaceutical Lighthouse factory site by McKinsey revealed the following:
- Aside from deploying AI tools, they trained over 25 managers and more than 100 frontline staff to apply AI through agile sprints.
- This led to a 10% gain in labor productivity and also reduced reliance on external vendors or consultants, as new AI-led digital roles were filled internally.
3. Outline Clear KPIs for AI Impact
Often, AI pilots stall and fail to scale due to a lack of clear performance metrics. A McKinsey report highlights that nearly 60% of manufacturers “lack clear targets” for measuring the impact of their AI integrations. This makes it harder for companies to track value and scale models that yield success, or rework those that don’t.
You’re more likely to scale those use cases beyond the pilot stage when you start tying AI investments to clear business outcomes such as improved yield or reduced downtime.
Experts recommend setting an objective that outlines the primary functions of your AI integration, i.e., where you want AI to add the greatest value. This could include reducing downtime or increasing throughput.
Next, you can tie these objectives to concrete performance measures based on the SMART framework: Specific, Measurable, Achievable, Relevant, and Time-Bound. For instance, a steel manufacturing company that wants to adopt and scale AI may set these targets:
- Lower defect rates by 20% with the help of computer vision technology to detect inconsistencies in finished steel products.
- Shorten production cycles and accelerate product launches.
- Use predictive maintenance for key furnace machinery to reduce unplanned downtime by 15% over the next six months.
4. Invest Beyond the Factory Floor
AI-powered innovation is integral to actual production, but should not be limited to it. To properly scale AI, companies must apply it across the entire value chain. This includes optimizing it for supply chain activities, boosting sustainability, and even leveraging it before production happens.
This means companies can also use it to facilitate quicker R&D and simulation, ultimately helping shorten the time-to-market for new products.
For example, US-based retailer Walmart has collaborated with NVIDIA to create digital twins to plan distribution and layouts for more than 1,700 stores. The result? Its teams can simulate and test optimal environments safely to improve revenue and sales.
How Alpha Software Can Support Your AI Scaling Endeavors
Scaling AI is no small feat – it is only possible when systems bring data, workflows, and teams together in one place. To do this, you need a reliable partner like Alpha Software that can help centralize data and leverage AI insights in daily operations without overextending IT teams. With Alpha Software’s low-code application builder, you can build applications that can update schedules or alert teams in real-time when there are last-minute changes to production. The app can also trigger automated maintenance tasks when your AI system predicts a machine failure.
Don’t let your AI insights languish in dashboards. Use Alpha Software to build applications that can integrate with your company’s workflows, so your staff are always in sync and working with the latest information. Talk to our expert consultants today to begin leveraging AI in your organization.
Comment