Questions manufacturers should ask AI vendors but rarely do

Key Takeaways

• Manufacturers should look beyond traditional software purchase criteria when evaluating AI vendors
• Data quality, governance, integration, accountability, and long-term operational realities are crucial for successful AI deployment
• Vendors' assumptions about data quality and handling of data drift are essential considerations
• Manufacturers should ask about the percentage of customers who have reached production scale and the reasons for unsuccessful deployments

Introduction to AI in Manufacturing

The potential of artificial intelligence (AI) in manufacturing is vast, with applications in predictive maintenance, automated quality inspection, production scheduling optimization, and more. However, executives' enthusiasm for AI can lead to hasty decisions, and manufacturers often evaluate AI purchases using traditional software purchase criteria, such as implementation timelines, licensing costs, and return on investment.

Critical Questions to Ask AI Vendors

To ensure successful AI deployment, manufacturers should ask more nuanced questions, including:

What Assumptions Are You Making About Our Data?

AI vendors often claim their platforms can connect to existing systems, but few explain their assumptions about the data quality, completeness, and accessibility. Manufacturers should insist that vendors identify every assumption being made about data quality before deployment begins. For example, is sensor data complete, and are maintenance records consistent? A study by McKinsey found that poor data quality can lead to a 10-20% reduction in AI model accuracy.

How Does the Model Handle Data Drift?

Factories are dynamic environments, and AI models can become less accurate over time due to changing conditions. Manufacturers should ask how performance is monitored, how retraining is handled, and whether ongoing model maintenance is included in the contract or billed separately. According to a report by Gartner, 80% of AI projects will require ongoing model maintenance to ensure accuracy.

What Percentage of Your Customers Have Reached Production Scale?

Many AI pilots are successful, but few become enterprise-wide deployments. Manufacturers should ask about the percentage of customers who have reached production scale and the reasons for unsuccessful deployments. This can reveal obstacles such as integration challenges, organizational resistance, and disappointing returns. The following comparison table highlights the key differences between successful and unsuccessful AI deployments:

Bottom Line

Manufacturers should look beyond traditional software purchase criteria when evaluating AI vendors and consider critical questions such as data quality, handling of data drift, and production scale deployment. By asking the right questions and understanding the complexities of AI deployment, manufacturers can increase their chances of successful AI implementation and achieve significant returns on investment, with some studies suggesting up to 30% increase in productivity.