为什么公司用不好AI？从焦虑到行动的 3 个关键动作｜对谈百融智能张韶峰

Summary

This Shizilukou Crossing episode with Bairong Intelligence chair and CEO Zhang Shaofeng argues that enterprise AI is a supply-side productivity shift, not only another software tool. The source uses Bairong’s internal Digital Employees practice to explain why companies fail at AI when they ignore workflow politics, employee incentives, system APIs, and outcome measurement. It connects enterprise agents to Dark Office, Contact Center AI, Outcome-Based AI Pricing, AI BPO Roll Up, Business-Led AI Transformation, and Service As Software.

Key Claims

• Zhang Shaofeng says traditional companies feel both excitement and FOMO after DeepSeek and Open Claw, but often do not know where to start, who should own AI transformation, or how to measure value.
• The source frames enterprise AI as a production-side shift: if manufacturing had the “dark factory,” office work may move toward a Dark Office where white-collar and gray-collar workflows are performed by agents.
• Bairong Intelligence claims to have turned customer service and marketing, HR, reimbursement, legal review, approval, and risk-control work into Digital Employees rather than only personal productivity copilots.
• Bairong measures digital work against a “standard person” benchmark and asks digital employees to beat standard human output; the source treats output ratio and task completion as more important than agent count alone.
• The contact-center demo shows why Contact Center AI is a natural early agent scene: the AI can continue prior context, hand off to a human or another role, follow financial compliance rules, and refuse improper guarantees.
• Bairong intentionally simulates existing roles and handoff processes first, because real processes encode power, incentives, authorization boundaries, complaint handling, and abuse resistance.
• The source says one early mistake was preaching AI as a trend without giving human employees a reason to cooperate; Bairong later added performance records, rewards, and penalties for people who teach digital employees.
• A “digital employee home” gives agents names, job numbers, onboarding records, email, and performance information, making Digital Employees an HR and management problem as much as a model problem.
• Legacy software must expose APIs so agents can call CRM, order-management, and office systems; this reinforces Agent-Facing Interfaces as an enterprise requirement.
• Zhang argues that Chinese SaaS struggled because companies resist paying for process tools and prefer resources, hardware, custom projects, person-days, and maintenance fees; Outcome-Based AI Pricing is presented as a way out.
• Bairong’s pricing patterns include charging against standard-employee output, charging by work volume or hours, and taking a service fee based on transaction scale.
• The source identifies programming and Contact Center AI as early agent landing scenes because the work is bounded, measurable, and easier to close in a loop.
• Zhang is bullish on AI BPO Roll Up: recruiting, consulting, accounting, law, tax, finance, and other professional-service workflows have clear deliverables and may be rebuilt with agents.
• Baijian is presented as a professional-service platform that gives independent experts digital employees and office systems, then recombines legal, business consulting, tax, and finance expertise around company globalization needs.
• For enterprise leaders, the practical advice is to take AI seriously, avoid treating it as a trivial plug-in, start from simple high-frequency bounded tasks, design incentives, and build confidence before broader process redesign.

Key Quotes

“AI 更像劳动力” — enterprise AI should be managed as work capacity, not only a tool.

“不要一上来就改流程” — Bairong’s rollout advice for reducing organizational resistance.

“供给侧、生产力侧变化” — the source’s framing of AI as a production-side shift.

Connections

• Bairong Intelligence — company whose internal AI deployment and customer offerings anchor the episode.
• Zhang Shaofeng — guest and Bairong chair/CEO explaining the operating model.
• Baijian — Bairong’s professional-service platform example.
• Shizilukou Crossing — podcast/media context for the episode.
• Open Claw and DeepSeek — trigger examples for enterprise AI anxiety and agent expectations.
• Digital Employees, Business-Led AI Transformation, and AI Organization Design — main enterprise operating themes.
• Contact Center AI — contact center as an early measurable agent landing scene.
• Agent-Facing Interfaces and Agentic Workflow — technical workflow requirements for agents to act inside existing business systems.
• Outcome-Based AI Pricing, Product Led Willingness To Pay, and Software Payment Culture — pricing and buyer-behavior themes.
• AI BPO Roll Up, Service As Software, and Service Productization — service-market reconstruction themes.

Contradictions

• No direct contradiction with prior wiki content. The source reinforces the Rolling AI and Forward Deployed Engineer thread, while adding a more conservative rollout tactic: first fit agents into existing workflows and incentives before attempting deeper process redesign.