Trust is the foundation for successful AI adoption. By addressing skepticism, prioritising data quality and ensuring algorithms are explainable and auditable, AI can become a powerful force-multiplier in manufacturing operations. XpertRule’s Akeel Attar and Iain Crosley show how.
Manufacturers are increasingly looking to AI to boost efficiency, streamline operations and automate routine tasks, with 75% planning to step up their AI spending in 2025. However, much of this attention is focused on Generative AI – something that, as we discussed in a previous article, is poorly suited to factory settings.
Part of this misalignment stems from a lack of understanding of AI’s practical applications in industry. With only 7% of manufacturing leaders feeling “very knowledgeable” about AI applications, skepticism and trust issues loom large.
Feedback from vendors and end-users consistently points to trust as a leading barrier to adoption. Without trust, AI cannot deliver on its full potential, leaving many manufacturers hesitant to go beyond pilot projects.
Overcoming the AI ‘Fear Factor’
The portrayal of AI in the media has long been dominated by dystopian headlines and Hollywood blockbusters, with fears of mass unemployment and doomsday narratives. For manufacturers, this continuous, subliminal bombardment creates a trust deficit before any AI project even begins.
Business leaders are having to overcome not only technical hurdles but also the deep-seated skepticism that AI solutions are uncontrollable or inherently risky. To counter this, companies must approach AI with transparency and explainability at every stage, showing that AI is a tool to amplify human capability not replace it.
You can read our deep dive on the importance of Explainable AI here but a useful analogy is cruise control in a car.
Traditional cruise control maintains a set speed but that’s all. Compare that to adaptive cruise control, which considers real-time conditions, adapts to your driving preferences and responds intelligently. Similarly, AI in manufacturing must adapt to the unique needs and complexities of each operation.
For those implementing these systems, understanding the ‘mechanics’ – how algorithms interact with data inputs and external influences – is a vital part of building trust. Explainable AI bridges the gap between automation and operator oversight, providing a clear view of how the system reacts and adapts. This clarity increases confidence among users, fostering trust in AI’s outputs.
Building trust also requires a mindset shift – from a data-centric focus to a decision-centric approach.
Trust Starts with Decisions, Not Data
A common misstep in AI adoption is starting with the data instead of focusing on the desired outcomes. Many manufacturers think, We have all this data – what can we do with it? However, this approach often leads to complex systems that lack focus, transparency, fail to deliver meaningful outcomes and reinforce doubt over AI’s value.
A decision-centric approach begins by asking, What do we want to achieve, and What decisions need to be made to deliver those outcomes? Only then should you ask, What data supports those decisions and what are the models linking these decisions to this data?
From there, manufacturers must focus on ensuring data quality – calibrating sensors, cleaning data streams, validating inputs and standardizing formats. Remember, 90% of AI success lies in data preparation and only 10% in the modeling itself.
Imagine a manufacturer aiming to improve quality control. They might gather extensive data from every step of the production process to find possible defects, leading to an overwhelming volume of disjointed data with no clear path to action.
Using a decision-centric approach, they would:
- Define the Goal: Improve product quality and reduce defects by 10% over the next quarter.
- Identify Key Decisions: What factors directly impact product quality? What parameters should trigger quality checks? How can inspection processes be optimised to catch defects earlier? What actions should be taken when deviations are detected?
- Use AI to model the Outcomes: Build AI models that analyse historic production data , to discover explainable patterns relating outcomes to metrics like machine settings, material consistency or environmental conditions. The system can then use these models in real time to flag anomalies that indicate potential defects and recommends adjustments to maintain product quality.
This clarity in purpose makes AI implementations transparent, explainable and, ultimately, more trustworthy. It also provides a clear framework for measuring success, helping to build greater confidence from engineers, users and management alike.
Decision Intelligence – The Missing Link
A key factor in building trust is recognising that AI doesn’t replace human insights and experience – quite the opposite. Human operators and engineers bring a level of expertise, contextual knowledge and intuition that machines cannot replicate. Having a ‘human in the loop’ is therefore critical to an AI system’s effectiveness.
Decision Intelligence connects Explainable AI principles with operational trustworthiness by embedding human oversight at its core. For example, experienced technicians possess knowledge built up over years of practice. While they can’t be everywhere at once, their expertise can be integrated into AI systems to automate routine decisions while reserving complex or ambiguous scenarios for human intervention.
This balance between human and machine intelligence ensures AI systems remain transparent, reliable and dynamic. It also enables manufacturers to scale the knowledge of their experts, reducing variability across shifts and locations while maintaining trust and accountability.
How Decision-Centric AI Builds Trust
- Alignment with Business Goals
Decision-centric AI is purpose-driven. It starts with clear objectives, making it easier to evaluate success and demonstrate value – crucial for gaining buy-in from stakeholders. - Transparency in Decision-Making
By focusing on specific decisions, this approach provides clear reasoning for AI outputs. When engineers and operators understand why an AI system recommends a particular action, they are more likely to trust and act on its insights. - Scalability and Continuous Improvement
A decision-centric approach allows manufacturers to begin with well-defined, targeted projects. Each success builds trust, paving the way for broader AI adoption across operations. - Better Integration with Existing Workflows
Decision-focused AI aligns more naturally with current business processes, reducing resistance from employees and supporting smoother implementations.
From Pilots to Trusted Partner
For AI adoption to move from pilot projects to the heart of manufacturing operations, trust must come first. A decision-centric approach offers a practical pathway to achieve this, ensuring AI systems are transparent, aligned with business goals and designed to augment human expertise.
When manufacturers trust their AI systems, they can harness the technology’s full potential, creating new opportunities for efficiency, resilience and competitive advantage. Decision Intelligence becomes the connector between Explainable AI and operational trust, moving AI from being perceived as a risk to becoming a trusted partner.
This article is part of our ongoing series, Reality Check: What AI Really Means for Manufacturing, designed to inform, inspire and help you implement AI in your manufacturing operation.
Next in the series, we explore how manufacturers can build on this foundation of trust by adopting a Responsible AI strategy, focusing on transparency, accountability and fairness.
Dec 12, 2024 2:12:47 PM