Turning Dormant Assets into Gold: How Cognitive Bias Can Give You a Competitive Advantage
In today’s data-driven economy, artificial intelligence (AI) has emerged as one of the most powerful tools for companies seeking a competitive edge. Engineering teams across industries are harnessing AI to improve decision-making, optimize processes, and innovate new products. However, much of the focus on AI tends to revolve around acquiring new data, often overlooking an invaluable resource already in a company’s possession: dormant assets. These assets, which include historical data from past simulations, designs, tests, and field operations, represent an untapped wellspring of value.
While traditional data science principles emphasize minimizing bias in datasets to ensure generalizable insights, this article argues that leveraging dormant data exclusively from within a company’s own ecosystem can introduce a form of cognitive bias that, counterintuitively, enhances a company’s unique competitive advantage. This bias stems from the fact that the data reflects the company’s history, culture, values, and experiences—elements that have shaped its success. In this context, we will explore how using historical data in AI training embeds a company’s strengths into algorithms, turning dormant assets into gold.
The Nature of Dormant Assets in Product Lifecycle
Companies engaged in product development, particularly in engineering, manufacturing, and high-tech industries, generate enormous amounts of data across the product lifecycle. Every design iteration, simulation, test, and field operation produces valuable insights that are often stored but rarely revisited once a product has been launched or a project concluded. This treasure trove of information—referred to as dormant assets—is generated in? principally along 4 main engineering steps: Design, Simulation, Test and Field Operations.
Design Data: CAD models, design blueprints, and technical specifications detail the evolution of product ideas, offering insights into design decisions and trade-offs made during development.
Simulation Data: Models and simulations used to test products under different conditions often generate massive amounts of data. This data reflects how a product was expected to perform, helping engineers optimize designs before a product reaches production.
Test Data: Experimental data from prototype testing, quality assurance procedures, and performance evaluations reflect real-world conditions and responses, providing hard evidence of how a product or system behaves under stress.
Field Data: Once products are in use, feedback from customers, maintenance logs, and operational data accumulate, representing a real-world view of the product’s performance over time.
Though often archived, this data has the potential to drive future innovation. As AI and machine learning (ML) technologies advance, businesses can use dormant assets to train algorithms that emulate the core competencies of their engineering and design philosophies. These dormant assets, however, come with a natural bias: they only represent the company’s experience and solutions to challenges they have encountered. But rather than viewing this bias as a limitation, it can be leveraged as a distinctive competitive advantage.
Cognitive Bias and the Role of Data in AI
AI and ML models are fundamentally reliant on data for training, pattern recognition, and prediction. However, the type and source of data used in training AI systems can introduce various cognitive biases, which are typically seen as problematic in AI research. Bias can lead to skewed results, perpetuation of stereotypes, or incorrect decision-making based on a lack of diversity in the data.
The primary type of cognitive bias in AI is selection bias which occurs when the data used to train a model does not represent the full population or potential scenarios the model might encounter.
Additionally, confirmation bias concerns the tendency of the model to prioritize data that confirms existing beliefs or trends in the training set.
And because most historical data that is kept and saved is build upon successful outcomes, we are concerned with survivorship bias: the fact that when models are trained only on or survivors, it is neglecting important data from failures or challenges.
Finally, time ordering of data should not be prioritized to avoid recency bias which occurs newer data is given more importance than older historical data, leading to a skewed perception of trends and predictions.
Traditional AI development seeks to minimize these biases by diversifying data sources, ensuring the model is representative of a wide range of experiences. However, for companies looking to strengthen their unique competitive positioning, leveraging cognitive bias intentionally—by focusing exclusively on their own data—can reinforce their existing strengths. This approach aligns AI models with the company’s core competencies and historical success factors.
Cognitive Bias as a Competitive Advantage
Cognitive biases, when carefully managed, can give a company a competitive advantage by reinforcing the very elements that have contributed to its past successes.
Using dormant assets that reflect the company’s history ensures that the AI model is trained on data that has been central to its past product successes. For instance, a company with a strong legacy in automotive safety design could use decades of crash test simulations to train AI algorithms that automatically optimize new designs for safety—a bias that aligns with their competitive edge in the market.
Dormant data often includes solutions to known challenges the company has faced, such as unique material stresses or environmental conditions. By leveraging these datasets, the AI can be trained to anticipate and solve similar problems in the future, effectively embedding past lessons learned into future product development.
And by building on historical strengths, cognitive bias in AI, when limited to a company’s internal data, can reinforce the areas where a company has traditionally excelled. For instance, a semiconductor manufacturer that has extensive dormant data on specific manufacturing processes can train AI models to optimize production based on historical best practices, making their manufacturing processes more efficient than competitors.
Over time, companies often lose institutional memory due to employee turnover, retirement, or organizational restructuring. By using historical data to train AI, companies can preserve this valuable institutional knowledge, embedding it into algorithms that continue to drive innovation long after the individuals who contributed to the data have left the company.
Let’s look at some key industries and identifie key use cases where we could turn dormant assets into gold.
Aerospace Industry
The aerospace industry generates enormous amounts of simulation, test, and field data across the lifecycle of an aircraft. Companies like Boeing and Airbus have decades of data on how their aircraft behave in a wide range of conditions, from extreme weather to high-altitude operations. Traditionally, much of this data has been archived without being fully utilized in future projects.
领英推荐
By leveraging AI, these companies can reanalyze dormant simulation and flight data to train algorithms that predict aircraft performance more accurately. This data, biased towards the company’s designs, reflects years of knowledge on material fatigue, fuel efficiency, and aerodynamics. AI models trained on this historical data can significantly reduce the need for future physical testing, cutting costs and time-to-market for new aircraft. Moreover, the bias in the data reinforces the company’s deep understanding of its own products, giving them an edge over newer competitors.
Automotive Manufacturing
In automotive manufacturing, companies like Toyota Motor Corporation and 宝马 have vast stores of dormant assets in the form of CAD designs, test results, and customer feedback data from past models. Rather than letting these assets remain dormant, these companies have started to integrate AI and machine learning systems that leverage this historical data.
For example, training AI systems on crash test data from previous vehicle generations enables these companies to automatically generate optimized designs for safety in new models. The data, biased towards the company’s historical safety standards, allows them to maintain leadership in safety innovation. Moreover, by analyzing field data related to maintenance and repair histories, AI systems can identify components that are prone to failure, allowing engineers to redesign these components more effectively.
Pharmaceutical Industry
Pharmaceutical companies maintain detailed records of clinical trials, drug efficacy studies, and regulatory submissions—data that often becomes dormant once a drug is approved or rejected. However, these companies can use AI to reanalyze historical clinical trial data and train predictive models that accelerate drug discovery and development.
AI models trained on a company’s own historical data, which reflects the specific diseases they have focused on and the compounds they have tested, can help identify new therapeutic candidates more quickly. The cognitive bias here is towards the company’s areas of expertise, ensuring that future drug discovery efforts build on past successes. For instance, if a company has extensive data on the molecular structures of successful antiviral drugs, they can use this data to train AI models that predict the efficacy of new antiviral compounds, allowing them to maintain a competitive advantage in antiviral drug development.
Benefits of Bias in AI Training
Training AI on dormant assets biased toward a company’s historical data offers several key benefits such as :
Managing Cognitive Bias Risks
While there are clear advantages to using biased data from dormant assets to train AI, companies must also manage the risks associated with this approach. If not carefully managed, cognitive bias in AI can lead to narrow, inflexible models that fail to adapt to new market conditions or emerging technologies. To mitigate these risks, companies should implement a certain number of basic guidelines:
Firstly, it is clearly important to combine biased and external or newly created datasets. While internal data provides a competitive advantage, integrating external datasets can help ensure that AI models remain adaptive to changes in the broader market. For instance, a company might combine its historical product performance data with external data on emerging materials or customer preferences.
It is also critical to regularly update training data in order yo avoid falling into the trap of survivorship bias or recency bias, companies should regularly update the datasets used to train AI models, incorporating both historical and recent data.
More generally, human oversight is essential. While AI models can provide valuable insights, human experts should continue to oversee decision-making processes, particularly in cases where models might exhibit unintended biases. A data governance expert team must be empowered to guarantee efficient use of AI.
Finally, and in order to address the specific issue of overfitting, engineering teams should regularly test AI models on new data to ensure they remain flexible and robust; as overfitting occurs when an AI model becomes too closely aligned with the training data, making it less effective at generalizing to new scenarios.
In the age of artificial intelligence, companies are increasingly recognizing the value of data as a competitive asset. However, many overlook the potential of their dormant assets—historical simulation, design, test, and field data that reflect their unique experiences and expertise. By training AI models on this biased data, companies can embed their core competencies into future innovations, preserving institutional memory and reinforcing their competitive advantages.
While cognitive bias is often seen as a limitation in AI, it can be leveraged strategically to amplify a company’s strengths and ensure that AI-driven innovations remain aligned with the values, culture, and experiences that have made the company successful. In doing so, companies can turn dormant assets into gold, using their past to shape a future of continued innovation and leadership.
References
(i) Cheng, Y., Fu, L., & Zhang, W. (2020). Leveraging Historical Product Lifecycle Data for AI-driven Innovation. Journal of Product Lifecycle Management, 9(2), 123-135.
(ii) Zhang, T., & Li, Q. (2021). The Role of Cognitive Bias in AI Training: A Case for Biased Data. AI Ethics and Application, 7(3), 45-61.
(iii) Papanastasiou, Y., & Boultadakis, D. (2019). Cognitive Bias in Machine Learning: Benefits and Pitfalls in Industrial Applications. Journal of Industrial AI Research, 13(1), 78-91.
(iv) Shashidhar, R. (2022). Dormant Data Assets: The Untapped Potential for AI in Manufacturing. AI for Manufacturing, 5(1), 88-97.
(v) Srinivasan, A., & Patel, S. (2020). AI and Simulation Data: Leveraging Dormant Assets for Competitive Advantage. International Journal of AI and Innovation, 8(4), 215-230.