A Tale of Crushed Optimism and the Hidden Weight of Domain Knowledge
In the AI gold rush, where algorithms shimmer like fool’s gold, our cautionary story exposes how incomplete data and the absence of domain expertise can turn even the most promising AI projects in manufacturing into unsuccessful outcomes.
The Ambitious Goal
An iron-producing company’s mission to improve blast furnace efficiency with the help of AI in manufacturing. Their aim was to save millions on raw material costs without sacrificing the grade of their iron production. This ambitious project aimed to cut coke consumption by over 5%, boost throughput by 6%, and increase end-to-end product yields by 15%.
The Quest for Optimization
Our enthusiastic data science team worked relentlessly for months to optimize the implementation of AI for manufacturing. They were scrutinizing data and running countless experiments – each intended to refine the AI model. Confident in their meticulous approach, the team believed their work was on the verge of unlocking cost-reduction opportunities.
However, lacking insights from blast furnace specialists, the team missed the metrics crucial for the model’s success. Therefore, the AI model, operating with incomplete information, delivered inaccurate predictions and fell short of its potential.
Bam! Reality Slams the Door: Optimism met a harsh reality check.
Challenges in Implementing AI for Manufacturing
Domain expert’s absence:
The data science experts embarked into the unknown territory of the blast furnace alone. They faced typical challenges of implementing AI in manufacturing. The consequences were swift and severe.
Missing Key Data:
The data science team failed to identify the critical process parameters.These are essential to train the AI model effectively.
Language Complexities:
Miscommunication strained collaboration. This created confusion
The Foundation Of Trust:
Lack of transparency in communicating the thought process behind the solution and model evaluation contributed to a lack of confidence and disengagement.
Lessons Learned!
Our story serves as a reminder for careful consideration of best practices for organizations considering AI for manufacturing solutions to unlock transformative developments.
Domain expertise is the linchpin to AI success:
Actively seek inputs from domain experts to validate model outputs. This helps adapt the solution to business objectives by ensuring that your AI models are built on a solid understanding of real-world nuances.
Data Science as an Applied Business Problem:
Even seasoned data science experts find it difficult to unleash the full potential of data science unless guided by the deep and nuanced insights that business experts acquire over years of experience in the field.
Build Trust Through Cross-Functional Teams:
It is essential to bridge the gap with clear communication, collaboration, and transparency at every stage of an AI project lifecycle for successful AI implementations.
Learn from our mistakes, and let them guide your path to AI success. The future is bright when AI and domain knowledge work hand in hand.
The Key Takeaway:
In the end, the crucial lesson from this failure is clear: even seasoned AI experts in the field of manufacturing AI cannot unleash the full potential of their skills unless they intimately understand the intricate workings of the industry they are serving. Indeed, decades of experience in AI implementation can only go so far. Furthermore, the true power of AI can only be harnessed when it is guided by the deep and nuanced insights that business experts, developed over years in the field, can provide.
In this era where AI in manufacturing is transforming industries, it is a clarion call to all organizations considering AI solutions. Domain expertise is the linchpin to AI success. Consequently, the integration of AI and business acumen is the key to unlocking the full potential of artificial intelligence and achieving transformative results.
A relevant example is presented by Michael Agostini in his MATLAB Expo keynote in 2019, where he discussed a case study of a consortium of New Zealand universities using AI to optimize the production of powdered milk. Despite having a large amount of data and advanced algorithms, the project was ultimately unsuccessful. thus, this case study highlighted that for AI projects to succeed, it’s not enough to have vast amounts of data and skilled programmers; domain experts who understand the specific industry and its processes are also essential. This serves as a cautionary tale of how AI projects can fail if the right insights and knowledge are not applied.
We’ve previously explored a similar theme in our story about how neglected maintenance led to failure in AI implementation. You can read more about that here.
Question For you?
Beginners → Where does domain expertise fit in your AI strategy?
Experts → Share your thoughts on how organizations can integrate domain expertise into their AI strategy.
For any queries, you can reach out to us at support@emlylabs.com.