
The core driving force behind the first wave of artificial intelligence applications was market excitement. Organizations of all kinds accelerated the deployment of generative AI tools, advanced process automation, and actively explored application scenarios for AI in improving work efficiency. Leadership teams generally encouraged experimental applications driven by curiosity, with the core objective of avoiding falling behind industry trends.
Now, as AI applications continue to scale, enterprises are increasingly facing new realities – controlling computing costs, managing environmental impact, and ensuring business model consistency are just as important as technological innovation.
The next stage of AI application focuses on "intelligent optimization" rather than "speed of adoption." Internally, this means developing precise strategies to maximize the value of AI computing, essentially achieving efficient management of AI costs; externally, it requires AI-driven products and services to generate meaningful and sustainable revenue. These two dimensions together constitute the core meaning of AI applications evolving towards a smarter and more sustainable direction.
I. The AI Cost Paradox: The Imbalance Between Usage and Cost
Historically, the growth of the SaaS model followed a clear logic: the higher the user engagement, the greater the value created by the product. However, the emergence of generative AI has disrupted this pattern – the usage process generates direct and rapidly increasing computing costs. According to Gartner research, if enterprises fail to correctly model AI usage scenarios, their estimates of AI costs may be off by as much as 500% to 1000%. Every prompt, every token, and every model call in AI applications incurs a cost: a single developer using an AI coding assistant can generate thousands of dollars in computing costs in just a few weeks. If this cost is amplified across various departments of an enterprise, the initial enthusiasm for adoption can easily turn into an uncontrolled budget crisis.
Ironically, the most active AI users within an enterprise often become the most expensive users.
II. Internal Optimization: Value-Oriented AI Cost Management
Organizations that fully embraced generative AI early on are now generally facing multiple practical cost pressures, including computing resource usage costs, model fine-tuning costs, infrastructure expansion costs, and training costs required to help teams use AI efficiently.
The primary task of internal AI optimization is to conduct cost education and manage usage data transparently. Employees need to clearly understand that processing each token incurs a real cost. For example, companies can build data dashboards that allow employees to view the cumulative AI costs generated by their work in real time. Furthermore, selecting appropriate models for different tasks is a key way to optimize AI usage costs – not all tasks require the most advanced, high-end models.
Business leaders should prioritize AI application scenarios that yield tangible results, rather than simply pursuing technological novelty. If an AI application cannot drive business growth in any way, resources should not be invested in its deployment. The core application scenarios of AI should focus on three areas: accelerating results delivery, automating workflows, and providing business insights that are difficult for humans to obtain. Gartner defines this strategy as "strategic IT cost optimization"—its core is not simply cutting costs, but shifting spending to high-impact projects and continuously monitoring the value output of those projects.
It's worth noting that AI optimization can not only reduce operating costs and accelerate return on investment, but also have a positive impact on sustainable development. Processing each AI prompt consumes energy, and each round of model training consumes a large amount of computing resources. The cost of AI is reflected both economically and environmentally.
Even small efficiency improvements can bring significant overall benefits. Simpler prompt phrasing, fewer repetitive iterations, and smarter model selection can all effectively reduce the negative environmental impact of AI applications.
III. External Restructuring: Business Model Innovation Adapting to AI Characteristics
While promoting internal AI optimization, companies also need to rethink the customer delivery model for AI services. The previous model of bundling AI technology into standard subscription services or charging a fixed extra fee is becoming increasingly unsustainable—after all, AI computing costs increase proportionally with usage. Currently, four emerging pricing models are gradually gaining traction, with the core objective of ensuring flexibility while balancing cost and revenue predictability:
1. Pay-per-use: Businesses pay based on actual AI usage (such as the number of tokens consumed), ensuring that costs accurately match the scale of usage;
2. Tiered pricing: Gradient discounts are set based on usage volume, providing customers with cost predictability and helping them efficiently scale their AI usage;
3. Value-based pricing: Pricing is linked to quantifiable business results (such as productivity improvements and time savings), ensuring that pricing accurately reflects the actual business value of AI;
4. Hybrid pricing: This model combines subscription fees and usage-based billing, helping businesses better plan their budgets while providing the flexible option of paying extra only when usage increases.
A healthy AI business model hinges on achieving a balance between "value creation" and "cost generation" for both suppliers and customers – customers pay accordingly when they receive more high-quality services, and suppliers earn reasonable profits when they deliver actual business results. This balance ensures a win-win situation for both supply and demand sides as AI applications continue to become more widespread.
IV. Core Conclusion: The Balance of Intelligent Adoption and Optimization
The next stage of AI applications is centered on finding a balance between "innovation advancement" and "efficiency and sustainability assurance." The success metrics for enterprise AI applications have shifted from simply "scale of use" to "actual value output," focusing on three key dimensions: "cost per insight," "computational resources per outcome," and "environmental impact." Ultimately, the true value of AI lies not in the breadth of its application, but in its ability to achieve precise and efficient intelligent optimization within internal organizational operations and the broader business environment.