The Untapped Potential of Generative AI in Customer-Specific Applications
top of page
Search
Writer's pictureVictor Muntés

The Untapped Potential of Generative AI in Customer-Specific Applications


Large Language Models (LLMs) like OpenAI's GPT and Meta's LLaMA have become transformative tools, proving their value in tasks ranging from content generation to complex data analysis. These models are adept at synthesizing vast amounts of general knowledge, providing insightful answers, and even supporting decision-making processes. However, they face a significant limitation when applied to real-world, customer-specific scenarios: they lack access to live, proprietary data.


As a result, while LLMs can offer broad insights and general responses, they fall short when it comes to delivering precise, customized solutions that reflect the nuanced needs of individual clients. Imagine a construction or manufacturing project with a wealth of real-time data in its ERP, CRM, or project management systems. Without direct access to this data, LLMs cannot deliver the kind of actionable intelligence or predictive insights needed to drive value on the ground.


If generative AI could connect seamlessly with live databases and interpret customer-specific information in real-time, the impact would be revolutionary. Such an integration could enable AI-driven decision-making tailored to an organization’s unique operational landscape, allowing for adaptive responses to changing project conditions, enhanced risk mitigation, and a level of automation that dramatically improves efficiency.


Unveiling the EcoSteelFlow Project with Celsa: Transforming Real-Time Risk Management with Digital Twins and Generative AI


Our collaboration with Celsa Group, a leader in low-emission, circular steel production, exemplifies how generative AI can reach its full potential when linked to live customer data. Through the EcoSteelFlow project (Project ID: ACE105/23/000100), Beawre and Celsa are exploring innovative ways to optimize material management and reduce environmental impact, transforming operational risk management in the steel production process.


EcoSteelFlow introduces a Digital Twin capable of simulating and learning from the approval processes in Celsa's steel delivery pipeline. This digital replica allows Celsa to analyze potential risks associated with order approvals, predicting anomalies and inefficiencies that could lead to material waste. Using AI-driven insights, this twin offers a “risk footprint” of each actor involved, comparing it against industry benchmarks and previous behaviors. This data-driven approach enables early detection of potential issues that could delay production or lead to excess material waste.


At the core of EcoSteelFlow is a Generative AI-powered copilot, integrated with the Digital Twin to offer natural language recommendations and preventive actions tailored to real-time scenarios. This copilot is designed to monitor and interpret the ongoing dynamics within the order approval process, issuing suggestions to streamline production timelines and ensure efficient material usage. By allowing the AI to interact directly with live customer data, we anticipate a significant reduction in waste and, consequently, in Celsa’s CO₂ emissions. In fact, by shortening order approval times, EcoSteelFlow aims to prevent up to 71 tons of material from being repurposed, equating to a 19.5-ton decrease in CO₂ emissions annually.


Making Predictions Actionable: A Database Interface Accessible to All


A unique feature of EcoSteelFlow is how it materializes the predictions and insights from the Digital Twin directly into a live database. This database doesn’t just store real-time data from order processing, approval workflows, and production updates—it also seamlessly integrates predictive insights. Through this setup, Celsa employees can query the database effortlessly, using everyday language instead of complex SQL commands.


The system includes a digital analyst that interprets these natural-language questions, converts them into precise database queries, and retrieves the relevant data. With EcoSteelFlow, team members can simply ask questions like, “Which planner has the highest efficiency on recent projects?” or “How many production sheets are still awaiting approval?” The digital analyst interprets these natural-language queries, converts them into database searches, and delivers clear, actionable insights based on the behavior learned by the Digital Twin. This capability makes insights accessible across the organization, regardless of technical skills, empowering Celsa to make informed, data-driven decisions grounded in the historical and predictive knowledge of their operations.


The integration of digital twins and generative AI in EcoSteelFlow has the potential to set a new standard in operational risk management and sustainability. With actionable insights and predictive analysis powered by live data, we are moving closer to a future where AI serves as a digital co-pilot, guiding industries towards more sustainable and efficient practices.


We are grateful to ACCIÓ - Generalitat de Catalunya for their funding support through this project (ACE105/23/000100), making this innovative initiative possible.


0 comments

© 2024 by Beawre Digital SL

  • Twitter - Gris Círculo
  • LinkedIn - Gris Círculo
bottom of page