AI Enters the Deep End of Industry: My Latest Assessment of Digital…

Title: AI Enters the Deep End of Industry: My Latest Assessment of Digital Transformation, Smart Manufacturing, and Digital Employees

Hello

An unfolding reality: AI is no longer a "tool upgrade" but a rewrite of business structure.

Over the past few years, I've observed many companies talking about AI mostly at the level of "efficiency software," "automated content generation," and "cost-reduction assistants." But at this point, my conviction is increasingly clear: AI's real impact is not replacing a specific job action but restructuring the enterprise's decision chain, execution chain, and value chain.

That's why, once AI is combined with digital transformation, smart manufacturing, and digital employees, the focus of discussion should no longer be "whether to adopt" but "which segment to change first, who is responsible, and how soon a closed loop can be formed."

In popular discussions, AI, digital transformation, smart manufacturing, and digital employees are already on the same table. This is not just a pile of topics; it's the logic of industry:

1. AI is responsible for cognitive enhancement and decision automation;
2. Digital transformation is responsible for turning processes that were once fragmented, ambiguous, and experience-dependent into computable objects;
3. Smart manufacturing is responsible for connecting algorithmic capabilities to production lines, equipment, quality, and supply chains;
4. Digital employees are the key interface for enterprises to organize, assign roles to, and hold AI accountable.

I've always believed that enterprises don't lack a "large model"; they lack a mechanism for the model to truly enter the capillaries of the business.

The real watershed I see: moving from "showcase AI" to "production AI".

Many companies' first-phase AI projects look great: customer service Q&A, meeting minutes, marketing copy, knowledge base retrieval—flawless in demos. But when I review project successes and failures, I find that the real differentiator is the second phase: whether AI enters production systems, whether it is accountable for KPIs, and whether it can be audited and held responsible.

Let me break this down with a story.

A discrete manufacturing company initially introduced AI for knowledge Q&A. Engineers' efficiency in looking up SOPs improved significantly, and internal satisfaction was high. But six months later, management found that such projects had "applause but no reports." The problem wasn't technology; it was that value hadn't penetrated to core business metrics. So they moved AI capabilities upstream to three areas:

• Order forecasting and production scheduling coordination
• Equipment anomaly early warning and spare parts planning
• Image recognition of quality defects and root cause attribution

The result wasn't simply that "everyone got smarter"; it was shorter delivery cycles, reduced downtime losses, and controllable rework rates. In other words, it moved from experiential value to financial value.

This gave me a very clear insight: the core of digital transformation is not putting processes onto systems, but turning business variables into objects that machines can participate in optimizing.

Red and green flag analysis: opportunities are huge, but risks are also amplifying at the industry level.

In any technology wave, the real danger is not the technology itself, but enterprises misjudging the path of technology adoption. I prefer to analyze with "red and green flags" rather than one-sided optimism.

Green flags: four directions worth accelerating investment in

• Digital employee role specialization: turning AI from "everyone can use it a bit" into "certain roles first develop standard operating capabilities"—for example, procurement assistants, financial auditors, after-sales support specialists, and equipment diagnostic assistants.
• Closed-loop smart manufacturing: enabling AI not only to identify problems but also to participate in process optimization, scheduling recommendations, and quality traceability.
• Data asset front-loading: whoever completes master data governance, knowledge structuring, and process standardization first will more easily capture the AI dividend.
• Real-time decision support: shifting from monthly-report-driven to minute- and hour-level operational response, especially in scenarios of supply chain volatility and personalized orders.

Red flags: five types of risk most easily overlooked

• Hallucinations entering production environments: tolerable in customer service scenarios, but in process, quality inspection, and compliance, they can cause high-cost errors.
• Processes unchanged, AI wasted: if the old process is inherently lengthy and distorted, AI only amplifies inefficiency.
• Severe data fragmentation: gaps between ERP, MES, PLM, and CRM reduce AI to surface-level Q&A, unable to perform deep optimization.
• Unclear responsibility boundaries: after digital employees make recommendations, who approves, who takes responsibility, and who reviews must be institutionalized.
• Organizational psychological resistance: frontline employees fear replacement, middle managers fear loss of power, executives fear runaway investment—this affects deployment more than model accuracy.

A defamiliarized perspective: digital employees are not 'substitutes for people' but 'rewriting job roles'

I don't quite agree with simplifying digital employees as 'employees who never complain' or '24/7 assistants.' Such phrasing is catchy but easily misleads management.

In my view, the real value of digital employees lies in decomposing a job into four layers:

1. Information acquisition
2. Rule-based judgment
3. Cross-system execution
4. Exception reporting and escalation

Once a job is decomposed this way, enterprises can see which parts are suitable for automation and which must retain human judgment. That is, a digital employee is first not a 'person' but a new type of job architecture.

This is why I emphasize the Medici collision: manufacturing, software engineering, organizational behavior, and financial management must be viewed together. Viewing AI purely from an IT perspective tends to turn it into a system project; viewing it purely from a business perspective tends to make it a local patch; only cross-domain collaboration can turn it into an operational capability.

Data-driven predictions: three new thresholds for enterprise competition in the next two to three years

Based on current industry evolution, I make three judgments.

First, AI will shift from 'department-level pilots' to 'company-level operating systems.' Whoever first connects knowledge, processes, permissions, and audit systems will enable digital employees to scale up.

Second, the competitive focus of smart manufacturing will shift from point automation to full-chain collaborative optimization. The future competition is not just about equipment connectivity rates, but whether orders, plans, processes, quality inspection, and supply chain can be linked within the same decision-making framework.

Third, the ROI evaluation criteria for digital transformation will become stricter. I expect the industry to place more weight on three types of metrics: unit labor efficiency improvement, reduction in anomaly response time, and convergence of inventory and delivery volatility. AI projects that cannot map to these metrics will find it increasingly difficult to secure budgets.

For enterprise managers, I suggest doing only three things:

• First, select scenarios with high frequency, high loss, and high rule density
• First build data and responsibility mechanisms, then talk about full rollout
• First prove the business value of one closed loop, then replicate it to more job roles

Final note: don't treat AI as a trend; treat it as organizational capability building

I am increasingly convinced that AI will not reward all enterprises equally. It will only reward those willing to confront real process problems, real data problems, and real organizational problems.

If AI is still treated as a display window today, it will likely fade into silence after the excitement; but if it is placed in the main channel of digital transformation, in the core site of smart manufacturing, and in the institutional design of digital employees, it will evolve from a new technology term into a new profit structure for the enterprise.

This is no longer about 'whether you understand AI,' but 'whether you can turn AI into your company's language of productivity.'

With best regards