Frost & Sullivan Releases '2026 China Decision Intelligence: The "AI Third Pole" Reshaping the Industrial Economy'"

The evolution of artificial intelligence capabilities does not involve a single replacement path, but rather a progressive sequence of identification, generation, and optimization. Analytical AI is good at identifying patterns and predicting trends, answering "what happened" and "what might happen"; generative AI focuses on content creation and natural language interaction, answering "how to express" and "how to collaborate"; decision-based AI targets complex business objectives, incorporating predictive results, business rules, resource constraints, and optimization goals into a unified decision-making process, further answering "what should be done". Especially when multiple goals, constraints, and dynamic disturbances exist, simply predicting or generating information is often insufficient to provide actionable answers. The white paper views this as an important evolution direction for artificial intelligence from being a tool to becoming an active force. Its output is not just insights or draft solutions, but action plans such as production plans, scheduling instructions, route planning, and inventory allocation that can directly enter business processes.

Different from probabilistic generation, decision-based AI is usually based on objective functions, constraint conditions, and solving processes, emphasizing the interpretability, verifiability, and auditability of results. In low-fault tolerance scenarios such as power, aviation, and manufacturing, the speed, stability, and numerical accuracy of high-performance solvers often determine whether a solution can be adopted in real business operations.

From a value perspective, decision-based AI does not exist in isolation, but forms a collaborative relationship with analytical and generative AI: the former provides prediction, identification, and attribution capabilities, while the latter reduces the barriers to knowledge interaction, model building, and result interpretation. Decision-based AI completes the closed loop of constraint solving, solution trade-offs, and execution, and continues to iterate based on business feedback. The combination of these three capabilities enables artificial intelligence to move from "assisting in judgment" to "active action", especially suitable for high-value business areas such as complex scheduling, resource allocation, dynamic pricing, production planning, route planning, and supply-demand coordination. Through feedback, a continuous optimizing decision loop is formed, further improving operational efficiency and system resilience.

Source: Frost & Sullivan Analysis

Decision-based AI is not a single model, a single tool, or a fixed technical route. It is an intelligent decision paradigm that focuses on business objectives and execution results, integrating data, models, rules, knowledge, simulation, optimization, and feedback mechanisms. Its capabilities span data understanding, goal modeling, solution generation, simulation evaluation, optimization solving, and execution feedback, aiming to transform complex business problems into executable, evaluable, traceable, and continuously iterative action plans.

The complete technical architecture can be divided into three layers: the underlying computing power base and acceleration library, providing unified data, model, computing power, and governance support for different decision technologies; the middle layer includes multi-path decision capabilities that integrate predictive analysis and causal judgment, intelligent agent collaboration, rule engine and knowledge reasoning, as well as optimization solvers and mathematical programming; the upper layer consists of intelligent decision platforms and industry applications, which are applicable to scenarios such as manufacturing, energy, transportation, aerospace, and finance, transforming technical capabilities into business actions such as production plans, scheduling instructions, route planning, and resource allocation. The key to decision-based AI is not only having various technical capabilities, but also achieving vertical integration of computing power, models, algorithms, platforms, and scenarios through a full-stack architecture, allowing different technical paths to be flexibly combined according to business objectives, forming a complete closed loop of "perception—judgment—generation—optimization—execution—feedback". The full-stack collaborative capability thus becomes an important foundation for improving decision quality, engineering efficiency, and scalability.

The global decision-based AI industry size is showing significant growth. The underlying driver is the enterprises' systematic upgrade needs for complex operational decisions, and market capacity will increase significantly. The market space for decision-based AI is supported by its specific decision-making value that is verifiable, executable, and sustainable optimized in high-complexity scenarios such as energy, manufacturing, transportation, and supply chain. The following calculations are made based on three core dimensions, covering the application value from underlying solving capabilities, platform capabilities, to industry solution delivery.

The global decision-based AI market is expanding from single algorithm tools to a full-stack value space covering platform capabilities and industry solutions. The market size is expected to grow from $526.1 billion in 2025 to $1462.77 billion in 2030, with a CAGR of approximately 22.7% from 2025 to 2030E.

From a market structure perspective, the industry focus is shifting from underlying algorithm software to the decision platform layer and industry solution layer. Underlying solvers and mathematical optimization software provide core computing capabilities, but customers are increasingly focusing on platform products and industry delivery services that integrate modeling, solving, simulation, scheduling, and execution closed loops. The main growth in the market is expected to come from the industry solution layer. As industry scenarios expand, industrial intelligent agents, digital twins, and data governance systems gradually mature, the commercialization of decision-based AI will move from single-point verification to multi-scenario replication, driving the market size to maintain high growth.

Fields such as energy and power, intelligent manufacturing, transportation and logistics, aerospace, and consumer retail generally feature high decision frequency, many constraints, and wide impact ranges, making them representative application areas for decision-based AI.

In the energy and power sector, decision-based AI can be used for economic power grid scheduling, new energy output coordination, and spot trading strategy optimization; in intelligent manufacturing, it can help make overall trade-offs between delivery timeliness, equipment load, and material constraints regarding complex production planning, multi-factory capacity coordination, inventory replenishment, and abnormal re-scheduling; in transportation and logistics, it can be applied to vehicle route planning, flight restoration, capacity allocation, and port scheduling, enabling dynamic resource allocation under sudden disruptions. Related technical practices also show that scheduling, route planning, linear and mixed integer programming are becoming key capabilities continuously expanded by international optimization tools.

For high-reliability scenarios such as aerospace, decision-based AI can convert trajectory windows, dynamics conditions, and safety boundaries into verifiable optimization models; in consumer retail, it can support inventory allocation, dynamic pricing, and marketing budget optimization. In the future, its implementation path is expected to extend from single-point planning tools to a closed-loop decision system covering "prediction—modeling—solving—execution—feedback", and from one-time project applications to continuous optimization across regions and business units. Truly understanding industry physical constraints and business processes and embedding them into reusable models and platform capabilities will be an important foundation for decision-based AI to achieve large-scale value.

From China's competitive landscape, local enterprises have formed diverse layouts at each level of computing power, solvers, and industry applications: large model and intelligent agent manufacturers are close to business processes, but precise optimization often relies on external engines; predictive decision platforms are good at data analysis and business optimization, but still have limitations in optimal solving under all constraints; industry software and vertical modules are close to specific scenarios, but cross-industry replication and standardized delivery still need verification; commercial solvers possess core algorithm capabilities, but their value limit depends on their collaboration with domestic computing power, modeling platforms, and industry delivery. Overall, China's decision-based AI has development conditions such as a complete industrial system, high scenario complexity, and clear demand for domestic substitution. The next stage requires continued accumulation in underlying algorithms, cross-layer interfaces, engineering delivery, and replicable business models.