From Excel to Neural Networks: How AI Is Transforming Production Planning and When to Expect Full Autonomy

AI in production planning is evolving from a tool that replaces humans to a powerful assistant that augments them, as illustrated by the 'digital planning panel' concept from Fraunhofer Institute researchers, where the algorithm checks thousands of hidden constraints and offers optimal schedules, but the final decision stays with humans. In Russia, a joint study by the Skolkovo Foundation and the Digital Economy ANO reveals a cautious interest in generative AI, with key barriers including staff resistance (cited by about half of companies), poor data quality, and difficulties in evaluating ROI and information security. Despite these challenges, engineers are building hybrid systems that combine generative models with graph representations of production networks, use large language models (LLMs) to translate unstructured queries, and employ mathematical linear programming (as in the 'AI-Kantorovich' system, which reduced planning time to 2-3 minutes and cut costs by over 3% in a cosmetic production pilot). The Skolkovo researchers predict a mass transition: within 1-2 years, adaptive models for scenario analysis; within 2-3 years, high-precision models with digital twins; within 3-5 years, industry solutions and competence centers. Practical cases include US Steel and Google's MineMind reducing equipment downtime by 20%, Botswana Diamonds finding seven new diamond prospects from 380 GB of geological data, Severstal's AI chatbot cutting maintenance audit time from 60 person-days to one, NLMK's AI assistants speeding up software development by 53%, SIBUR and Sber designing new polymers via digital modeling, Tatneft and ITMO's 'Akela' assistant improving planning efficiency by 30% and reducing labor by 16-20 times, Rolls-Royce reducing physical experiments by 90% in alloy development, LEAP 71 testing a neural-network-designed Aerospike rocket engine in three weeks, Siemens' industrial copilot saving up to 25% maintenance time, Rosneft and ITMO cutting conceptual design for Arctic ports by tenfold, and the DOM.RF and Rocket Group generating over 20 development concepts daily. The company Creative Way, which reported this, sees that the biggest bottlenecks are a lack of competencies and dirty data, while that same area offers the fastest growth that pays off within the first months after a project.