" Blindly studying AI (chasing technical principles, hoarding conceptual tools) is a bubble in the chase for trends, while pragmatically using AI (solving business pain points, creating quantifiable value) is the shortcut to ride out the cycle
In 2025, the AI industry's giant bets triggered market worries: in September, NVIDIA announced it would invest up to $100 billion in OpenAI
At the same time, MIT's 'The GenAI Divide: State of AI in Business 2025'
And Gartner's August 2025 'Hype Cycle for Artificial Intelligence' released a clear signal: AI has moved from the 'trough of disillusionment'
Is AI a bubble trap that devours resources, or a generational dividend that reshapes business? The answer is already clear: studying AI (obsessing over underlying tech R&D, chasing giant-scale capital
01 Studying AI Is a Bubble: All Gear, No Skill — Technology Can't Save Business Shortcomings
The current state of enterprise AI adoption is precisely a true picture of 'all gear, no skill': many enterprises treat AI as a 'universal patch,' detached from actual business
1. 'AI Gear Piling' Detached from Business Ends Up as Ineffective Waste
MIT reports that 75% of failed AI projects share the common trait of 'blind in-house development, detached from business' (source: MIT's 'The Ge
A retail enterprise with chaotic inventory turnover and missing procurement logic invested 20 million yuan to build its own general large model, yet optimized neither the procurement process nor sorted out sales data
A chain restaurant brand with slow serving and insufficient standardization — whose core problem was unreasonable kitchen workflows and lacking staff training — spent 8 million yuan purchasing AI
A manufacturing enterprise with old equipment and backward processes invested 120 million yuan to form an AI R&D team, tackling 'equipment intelligent monitoring algorithms,' but because the equipment
The common thread of these cases is treating AI as a 'fig leaf for business shortcomings': without solving core business-process and management-mechanism problems, they vainly hope technology will 'bypass
2. The 'Smart Use of Tools' Focused on Business Is the Shortcut
In sharp contrast to 'all gear, no skill,' the core logic of 'top students' like Yum China and Mixue is 'solve the problem first, then pick the gear':
Yum China: first face the business pain points of 'high in-store food waste and unreasonable staffing,' then launch the AI smart operations system 'Q-Smart,' focusing
Mixue: first solve the core problem of 'slow supply-chain delivery and serious inventory backlog,' then build the 'Smart Brain' AI system, covering demand forecasting,
SANY: first lock onto the business shortcoming of 'frequent equipment failures and low customer satisfaction,' then use an AI predictive-maintenance system that warns of faults via sensor data
COFCO: first target the production pain points of 'low flour yield and high energy consumption,' then introduce an AI smart grinding mill that adjusts grinding parameters in real time, achieving higher
The success of these enterprises never came from 'mastering more advanced AI technology,' but because they escaped the 'all gear, no skill' trap — A
3. The Core of the Bubble Is 'Business-Technology Disconnect'
MIT's 'The GenAI Divide: State of AI in Business 2025' clearly states
A Gartner 2025 Hype Cycle for AI related report shows that 57% of enterprises say their data has not yet reached AI-ready standards
Gartner's China official website explicitly advises: 'Enterprises should prioritize proven vertical AI solutions rather than investing heavily in self-development,' which
OpenAI's situation — a net loss of $11.5 billion in a single quarter with revenue covering only one third of the loss — further highlights the risk of 'technology detached from commercial business scenarios
02 The Shortcut of 'Using AI': Solve Problems with Tools, Not Learn Tech to Build Tools
'Using AI' means taking the enterprise's specific business pain points as the starting point and quantifiable business value as the goal, leveraging market-proven AI tools, vertical industry
Business orientation: all AI applications revolve around real business scenarios (such as supply chain, production, operations, service), rejecting tech experiments or concept hype detached from reality
Quantifiable value: the final effect must be presented with clear KPIs (such as cost-reduction ratio, efficiency gain, revenue growth amount), not vague claims like 'improving digitalization level'
Tool dependence: prioritize mature, deployed AI products or solutions; no need to invest in self-developing general technology, avoiding the high risk of 'reinventing the wheel'
Controllable threshold: the core requirement is 'knowing how to use the tool' rather than 'understanding technical principles'; ordinary business staff can get started after short training, no need to form a dedicated AI
The essential difference between 'using AI' and 'studying AI'
03 Breaking the Confusion: A Four-Step AI Adoption Method from Business Strategy to Pragmatic Execution
When formulating AI strategy, many enterprises fall into two confusions: either blindly follow giants in 'AI for the sake of AI,' treating 'doing AI' as the goal
1. Ming Dao (Clarify the Path): Anchor the Business Strategy, Find AI's 'Service Direction'
The core of 'Ming Dao' is to resolve the confusion of 'AI strategy detached from business': AI's 'Dao' is never independent; it must carry the enterprise's core
Specific actions:
Anchor the business strategy: first clarify the enterprise's current core business goals (such as 'cut 10% cost in 2025' or 'expand stores in lower-tier markets'), and AI's application
Filter business pain points: from the four dimensions of cost, efficiency, quality, and revenue, sort out the TOP 3 specific pain points blocking business goals (such as 'inventory backlog leading to
Quantify AI goals: turn pain points into measurable KPIs strongly tied to business — for example, if the business goal is 'cost reduction,' the AI goal can be set to
2. Li Fa (Set the Rules): Build the Adoption Framework, Avoid the 'Blind Expansion' Trap
The core of 'Li Fa' is to resolve the confusion of 'no rules, chaotic layout': many enterprises lack a framework when using AI — either spreading too thin and scattering resources, or
Specific actions:
Prioritize scenarios: rank the screened AI scenarios by the principle of 'high fit with business strategy, low investment cost, quick results,' and adopt first the
Data governance planning: the core bottleneck of AI adoption is data; invest 70% of upfront resources to prepare 'AI-ready data' — first clar
Establish an evaluation mechanism: set an 'effect pass line' and an 'adjustment threshold,' for example 'if after 1 month of trial operation the AI ordering-forecast system has not reduced food waste rate
3. You Shu (Refine the Craft): Master Practical Skills, Solve the 'Don't Know How to Use It, Can't Use It Well' Problem
The core of 'You Shu' is to resolve the confusion of 'high technical threshold, don't know how to use it': many enterprises buy AI tools but, because staff don't understand operation and over-rely
Specific actions:
Scenario-based hands-on training: around the selected AI tools, conduct 'business-oriented' hands-on training that does not explain underlying technical principles, focusing only on 'how to use the tool
Build a human-AI collaboration workflow: clarify the closed loop of 'AI assists + humans decide,' breaking the myth of 'over-relying on AI' — for example, AI generates
Distill a standardized manual: compile high-frequency issues encountered in tool use (such as 'how to handle failed data upload') and optimization tips (such as 'how to adjust forecast weights in peak season'
4. Ze Qi (Choose the Tool): Precisely Match Tools, Avoid the 'Blind Hoarding' Trap
The core of 'Ze Qi' is to resolve the confusion of 'choosing the wrong tool': many enterprises follow the crowd to purchase high-end AI tools and general large models, but because they don't match business scenarios
Specific actions:
Four dimensions for tool screening: tightly linked to business scenarios and the adoption framework, from 'scenario fit (whether it matches business needs, e.g., manufacturers prioritize industrial AI tools)
Prioritize 'ready-made-ism': SMEs need not self-develop; the first choice is mature vertical-industry solutions or public APIs — for example, a restaurant can directly adopt the solution akin to Yum China's 'Q-Smart'
Small-scope trial-and-error validation: first apply the selected tool in 1–2 pilot scenarios / stores to verify whether it hits the preset KPIs — for example, a food enterprise
04 Pragmatism as the Foundation, Ride Out the AI Bubble
AI technology's evolution is unstoppable, but the divide between 'studying AI' and 'using AI' has long determined enterprises' different endings. NVIDIA and Op
The four-step method of Dao, Fa, Shu, and Qi is precisely the key to breaking enterprise AI confusion: 'Ming Dao' makes AI carry the business strategy, avoiding 'AI for the sake of AI
Competition in the AI age was never a contest of 'who understands technology,' but a match of 'who can use technology to solve problems.' When the bubble recedes, those obsessed with tech
