What’s working in AI? Real-world transformations from the first MINDS cohort

As artificial intelligence (AI) reshapes our world, a quiet revolution is taking place, not in theory but in practice.

The World Economic Forum’s MINDS (Meaningful, Intelligent, Novel, Deployable Solutions) programme was launched to spotlight AI applications already delivering real impact across critical sectors like healthcare, energy and financial services.

From optimizing train operations to improving access to diagnostic care, the first cohort of MINDS winners shows how AI can be applied with purpose to solve complex global challenges. These are not pilots or prototypes. They are real solutions already transforming lives and industries.

The MINDS programme is part of the AI Governance Alliance, in collaboration with several Forum Centres and industry stakeholders, to identify and share trusted and scalable AI use cases.

As applications open for the second cohort of MINDS, here is the first group of real-world transformations and the meaningful change they are driving today.

Real-world examples of how AI can transform industries in the first cohort include:

AIMS members: Yangbing Lou (executive sponsor), Siqi Zhu, Zhiyong Zhang, Guoyang Yan, Tianhua Lu

CATL members: Jun Ni (executive sponsor), Zhiyang Wu, Lin Ma, Chao Guo, Wenjing Jin

Hangzhou Augmented Intelligence Manufacturing Solutions Co., Ltd (AIMS) has introduced a hybrid AI industrial augmented intelligence platform designed for high-stakes industrial settings, where data is limited but decisions matter. By combining physics-informed models, expert knowledge and few-shot learning, the system delivers predictive insights and process optimization, even in complex environments like battery manufacturing at Contemporary Amperex Technology Co., Limited (CATL). It runs in real time, integrates with existing infrastructure and adapts to fluctuating conditions without constant retraining. Already, it has cut quality variation by 50%, halved operator workload and boosted production speed, demonstrating a scalable model for intelligent decision-making across data-scarce industries.

Members: Zhang Junjie (executive sponsor), Liu Junwei, Zhu Dongwei, Feng Lei, Wei Peng, Wang Jian

Ant Group has built a multimodal, privacy-preserving AI system spanning the patient journey — from diagnosis to recovery. Used by 40 million patients and 290,000 doctors across 3,000 institutions, the platform delivers diagnostic accuracy above 90% and has reduced medical literature search times by 80%. With plans to scale towards 800 million users, it offers a model for trusted, nationwide health AI infrastructure.

Members: Yuxiang Zhou (executive sponsor), Yongqiang Ren, Xiaokang Pei, Yihang Cai, Qiuru Xiang

Black Lake turns the chronic mismatch between idle factory capacity and fast-moving e-commerce demand into a real-time, AI-orchestrated marketplace. A stack of role-specific agents matches each incoming order to the best-fit machine line in approximately 30 seconds, autogenerates manufacturable 3-D models from plain-text or CAD prompts, and steers every workstation through a smartphone MES — eliminating 75% of manual scheduling. Early roll-outs lifted utilization from 65% to 83%, cut unit energy use 18% and shrank product cycles from 6–12 months to just 1–3 months, releasing RMB 5.4 billion ($752 million) in annual value. At full scale (330,000 daily users), Black Lake – full name Shanghai Black Lake Technologies Co., Ltd. – expects 3–5 % less scrap, 30-day SKU launches and 15–20% less working-capital tied up in inventory.

Members: Lokwon Kim executive sponsor), Dong Jin Hyun (executive sponsor), Tim Park, Moon Sub Jin

DEEPX and Hyundai Motor Group have partnered to bring advanced edge AI to autonomous robots, a major step forward in aligning AI innovation with low-power, real-time computing needs. Their joint solution replaces energy-intensive hardware with ultra-efficient AI chips that deliver 240% the performance of a 40W GPU at just 5W power, cutting system costs by 90%. Paired with a full-stack SDK, it enables seamless deployment of models for perception, navigation, and control directly on-device. The result: faster, greener and more intelligent robots for complex indoor environments, pushing the frontier of scalable AI at the edge.

Members: Ke (Luke) Hu (executive sponsor), Zhe Li (executive sponsor), Lance Zhao, Weinan Zhou, Linghao Lu, Qiyu Chen, Zhixuan Wu, Liming Chang

Battery cell design that once took up to three years now completes in weeks. Engineers simply type a goal — “raise energy density” — and the cloud platform runs physics-grade AI simulation models across 70 variables, returning an optimal cell blueprint with full citations. Early pilots cut concept-to-prototype time by 3.6 times, and trim material and energy waste by 40%. The partnership between battery design and software services company Electroder and the Battery Design and Manufacturing team at the Tsinghua University aims to give thousands of firms instant access to enterprise-level battery cell R&D tools.

EthonAI members: Julian Senoner (executive sponsor), Bernhard Kratzwald, Davide Vecchione, Hendrik Schmid, Arman Pour Tak Dost.

Siemens members: Peter Koerte (executive sponsor), Sébastien Bey, Alexander Dierolf, Andreas Duerr, Matteo Odermatt, Sascha Leadbetter

Siemens and EthonAI began collaborating to improve factory operations using AI-enabled quality control. EthonAI’s Industrial AI platform empowers manufacturers with autonomous workflows that reduce costs, enhance product quality and increase throughput. Integrated with Siemens Industrial Edge, the platform enables factories across Europe, North America and Asia to automate millions of visual inspections and uncover inefficiencies through autonomous root cause analysis and process mining. The decision to scale this collaboration is backed by clear ROI: Standardizing visual inspection via a unified platform - rather than isolated point solutions – saves Siemens between €30,000-€100,000 ($34,800-$115,900) per inspection station. This impact is amplified by Siemens’ scalable software ecosystem, which ensures rapid, consistent deployment of AI solutions across its global factory network.

Members: Rohit Kapoor (executive sponsor), Deepti Kalra, Anand Logani, Shantanu Paliwal, Shubham Jain, Swati Malhotra

EXL's Agent-Swarm trains a fleet of task-focused generative AI agents – one each for legacy code discovery, dependency mapping, translation, testing and tuning – to migrate SAS, Oracle and 20-plus other stacks straight onto virtually any cloud. Running in an orchestration layer, the swarm wipes out 80% of the grunt work, auto-converts 70–80% of source code and still lifts post-migration performance 15–25%. Projects now finish 4–24 months sooner, while trimming overall costs 20–40% (worth $50,000–$500,000 and 25–40 FTE-months per cut-over). Scaled across migration projects, the platform is on track to hand back 5,000 person-years of productivity that was once trapped in legacy rewrites.

Foxconn has launched Project Genesis, an AI transformation initiative that is redefining the future of electronics manufacturing. Applications like FactoryGPT and Setpoint are automating decision-making across Foxconn’s global operations, handling 80% of the workload and enabling humans to focus on the final 20%. Genesis goes beyond isolated use cases – it's an ecosystem that tokenizes production expertise and orchestrates operations through AI agents. With deployments under way, Genesis is projected to unlock up to $800 million in value.

Members: Yoshinami Takahashi (executive sponsor), Yuki Doi (executive sponsor), Kazuki Hiraiso, Kazuki Nakata, Yoshiaki Nonaka, Mao Terashima

Fujitsu fields a network of task-specific AI agents – procurement, inventory, logistics, and risk – coordinated by a knowledge-graph engine that spins up dozens of forecasts in minutes and flags shortages or overstock in real time. Autonomous where possible and human-escalated when needed, the system has lowered annual inventory costs by $15 million, trimmed stock by $20 million and halved supply-chain headcount, while profit-impact assessments now arrive three hours after major disruptions. Rollouts finish inside three months, with projections of an extra $4 million sales and over 50% FTE savings per enterprise as adoption widens.

Members: Uchikado San (executive sponsor), Koji San, Nadia Mazzino

Rail system’s critical assets require important investment and have a very long lifespan. How these assets perform over time is a key factor. To address this, Hitachi Rail developed an advanced digital platform in collaboration with Hitachi Digital Systems and Services to support the next generation of railway operations. The system integrates data from rolling stock, infrastructure and signalling into a unified AI-powered analytics framework. Using machine learning and a modular edge-cloud architecture, it enables condition-based maintenance, real-time diagnostics, and energy optimization. This approach has helped reduce delays, cut energy consumption at depots and lower maintenance costs. The solution has wider applications, with potential to improve outcomes across metro, regional and freight networks.

Members: Yang Longru (executive sponsor), Jin Haimin, Luo Yi, Gao Hongsheng, Liu Chengyan, Bao Xinyue

The Industrial and Commercial Bank of China (ICBC), one of the world’s largest banks, built a 100-billion-parameter language model tailored for financial operations. Deployed across 15,400 branches and 400,000 employees, the model has delivered more than RMB 500 million ($69.5 million) in added capital allocation profit and is used daily by 20,000 staff. As this is the first batch, it is now scaling to serve 600,000 users and millions of automated resolutions per day, redefining intelligent financial infrastructure.

KMPG member: Hendrick Thörner

SAP members: Philipp Herzig (executive sponsor), Sachin Kaura, Natalie Han, Joachim von Goetz, Christian Karaschewitz, Caroline Ngo

Leveraging SAP’s alliance and AI capabilities, KPMG firms deployed a retrieval-augmented copilot built on a knowledge graph indexing 200,000 SAP documents and terabytes of industry best-practice code. Benchmarked against internal SAP certifications, the AI system is designed to deliver human-level answers, explain ABAP logic, generate code and propose configurations in seconds. SAP projects that consultants can save 1.5 hours per day, rework decreases from 15% to 7.5%, and project timelines can shorten from 18 to 15.5 months, targeting an 18% faster migration rate. KPMG already has thousands of active practitioners and is working towards scaling globally.

Members: Lu Bo (executive sponsor), Che Min (Jammi) Tu (executive sponsor), Guan Wei (executive sponsor), Li Robert (executive sponsor)

Lenovo is redefining supply chain management through an AI-powered agent that serves as the brain of its worldwide operations. iChain, a supply chain AI agent, orchestrates demand forecasting, supplier risk assessment, inventory optimization and logistics routing – powered by real-time data and hybrid AI technologies. Early outcomes include: disruptions are flagged up to two weeks earlier, logistics shipment accuracy is improved by 30% and carbon footprint calculation is accelerated by three weeks. Beyond operational gains, this system signals a new paradigm – where scalable AI decision-making drives resilience, sustainability and competitiveness across global supply chains.

Members: Philippe Rambach, Navdeep Ahuja, Aditya Changavalli, Vanya Ignatova, Jacques Kluska

Schneider Electric bolts a cloud-based MPC optimizer onto a rugged edge controller, turning each facility’s solar, batteries, EV chargers and flexible loads into a self-learning microgrid. The AI retrains on fresh weather, tariffs and demand data every few minutes, ranks actions by cost plus CO₂ intensity, and dispatches set-points that keep sites running – connected or islanded – without bespoke engineering. Across 97 live sites a lean 12-person team has cut external draw by 458 MWh and 109 t CO₂ per site each year (-28%). Scaling the shared “brain” to 1 000 locations would remove about 100 000 t CO₂ annually and put top suppliers on a fast track to net zero.

Members: Rahul Chillar (executive sponsor), Alexander Gloning, Carlos-Eduardo Pazos-Zarain, Monica Arnaudo, Paul Baumann

Siemens is bringing autonomy to building operations with a cloud-based engine that orchestrates heating, ventilation and air conditioning systems in real time. By combining live data, occupancy forecasts and weather insights, the AI optimizes comfort and efficiency—without needing to replace existing infrastructure. At NWT Media in Karlstad, Sweden, their solution improved comfort compliance by over 25% and reduced monthly energy use by more than 6%. The system continuously learns and adapts across domains, running as a closed loop. With results like these and a roadmap to scale across 100-plus sites, Siemens is enabling buildings that manage themselves – and simply feel better to be in.

Members: Chi Wang, Jiabin Lu, Tom Lu, Ray Pang, Wenyuan Bao

TerraQuanta and Horizon Power have deployed China’s first transformer-based weather forecasting system tailored for electricity markets, a major step forward in aligning AI innovation with a weather-driven decarbonizing grid. Trained on 40-plus years of global meteorological data and powered by 5 billion parameters, the AI model delivers 7–15 day forecasts with 50,000× the efficiency of traditional methods. It uses satellite sensing, terrain mapping and GNNs to predict wind, solar radiation and price shifts at 30-metre resolution. Now used across Horizon’s 2,000 GWh annual trading portfolio, It enables faster smarter decisions in volatile electricity spot markets reducing risk and boosting profitability.

Members: Stanley Prusiner (executive sponsor), Jack Hidary (executive sponsor), Gregory Merz, Nadia Harhen, Andrea Bortolato, Ben Shields, Kelly Richdale

UCSF’s Institute for Neurodegenerative Diseases, in partnership with SandboxAQ, is tackling the challenges of Parkinson’s drug discovery using AQBioSim's physics-native AI platform powered by quantum chemistry on GPUs. Breakthroughs stem from integrating advanced cryo-EM imaging, access to brain tissue and SandboxAQ’s prion-targeted software. Using PrionDock, the team virtually screened 5.6 million compounds in weeks, narrowing the pool to 7,000 for lab testing—a 36-fold reduction in effort. This led to a 30-fold improvement in hit rates for phenotypic screening assays over traditional high-throughput screening testing 250,000 compounds. The approach drastically shortens discovery timelines and holds promise to accelerate treatments for more than 10 million Parkinson’s patients and others with neurodegenerative diseases.

Beyond the featured individuals, hundreds of colleagues also contributed to the success of each project. PepsiCo was also part of the first cohort of the MINDS programme.

The MINDS programme comprises partners behind ways of leveraging AI to drive transformation and create measurable impact in their respective fields.

The benefits of joining the programme are multifold. Companies gain recognition as leaders in their field by being able to showcase their AI-enabled business reinvention through global exposure on Forum media channels. They can also benefit from insights and unlock partnership opportunities by joining a community of AI leaders, as well as get greater access and visibility by presenting to a community of 400+ VPs across industries.

When applications for MINDS second cohort open in July, private-sector companies from all industries, along with non-profits and academic institutions are welcome to apply to join our global community of AI leaders seeking to build meaningful, intelligent, novel and deployable solutions to address some of the world's greatest challenges.

More information on how to apply for the MINDS programme can be found on the MINDS website.