Chasing perfection can do more harm than good, according to Ozgur Tohumcu, the general manager of automotive and manufacturing at Amazon Web Services.
In this Q&A with Manufacturing.net, Tohumcu explains how manufacturers can make calculated decisions based on the best available information, rather than waiting for a perfect opportunity.
Nolan Beilstein (NB): Why do you believe chasing perfection is holding back manufacturing leaders?
Ozgur Tohumcu (OT): Perfection should not be viewed as a prerequisite needed for innovation to begin. Particularly in today’s rapidly changing and technology-oriented manufacturing environment, the ability to innovate quickly is foundational for success and maintaining a competitive advantage.
I’ve seen this work firsthand at Amazon Web Services with our “minimum loveable product” approach. We partner directly with customers to co-design solutions and test them in real environments.
Instead of waiting for a perfect, all-encompassing solution, we release products that solve specific, important customer challenges effectively. We then iterate based on actual feedback, embodying the “move fast and adjust” mindset that’s crucial for manufacturing agility.
The most important step is simply beginning the process. Manufacturing leaders should empower their teams to experiment with existing data assets. Success starts with a modern industrial data strategy that democratizes secure data access across the organization and encourages rapid proof-of-concept testing focused on the most pressing production challenges.
NB: What is the 70% rule for decision makers?
OT: Fundamentally, the 70% rule is a mechanism to help organizations develop a culture of experimentation and continuous improvement. It prioritizes speed and action taking, by empowering organizations and individuals to make calculated decisions based on the best available information, instead of waiting for 100% certainty.
At Amazon Web Services, we use a mental model we call “one-way and two-way doors” to assist in making high-quality, high-velocity decisions. One-way door decisions have significant, often irreversible consequences—like building a fulfillment center or data center, which requires substantial capital expenditure and careful analysis.
Two-way door decisions, however, have limited and reversible consequences—such as A/B testing a feature on a website or mobile app.
By recognizing which decisions are easily reversible, companies can lower the cost of failure and gain insights that inform and accelerate their next innovation.
NB: What is an example of how generative AI has impacted engineering and product development?
OT: A powerful example of generative AI’s impact on engineering and product development is our collaboration with Siemens. They now offer their Industrial Copilot for Engineering on AWS through their Totally Integrated Automation Portal (TIA Portal), which serves more than 120,000 users worldwide.
This solution empowers engineers of all experience levels to write accurate, trustworthy code in a fraction of the time previously required.
The Industrial Copilot helps address the shortage of skilled automation engineers with domain-specific programming expertise by providing programming assistance for Programmable Logic Controller (PLC) code in the Structured Control Language (SCL) programming language or structured text.
By leveraging Amazon Bedrock, TIA Portal users can code more quickly, automatically creating code instructions for repetitive tasks. This allows engineers to focus on more complex, value-adding tasks.
This collaboration illustrates how generative AI is making specialized knowledge accessible, elevating the capabilities of junior engineers while enabling more experienced engineers to focus on more complex, value-adding opportunities.
NB: What is driving the adoption of generative AI and how can manufacturers prepare?
OT: The adoption of generative AI in manufacturing is being driven by skilled workforce shortages and the need to increase the pace of innovation. With generative AI, manufacturers are able to preserve the institutional knowledge of experienced workers, optimize machine availability, improve efficiency and maintain consistent product quality.
For example, through Georgia-Pacific’s implementation of Amazon Bedrock, they were able to create an internal chatbot to respond to machine operators’ questions in near real-time.
This generative AI solution has streamlined operations, providing operators with access to the data they need to make immediate adjustments. This has resulted in fewer off-quality products, decreased machine downtime and increased productivity.
The company estimates that this generative AI solution alone has had potential annual savings in the millions across their facilities.
By identifying use cases that address specific operational challenges, manufacturers can quickly implement generative AI solutions, while simultaneously developing the capabilities and expertise for more complex, widespread AI adoption.