深圳新闻网2026年9月20日讯(记者 朱佳若)9月18日下午,华为全联接大会2026“人工智能赋能城市高质量发展”分论坛在上海顺利举办。深圳今日人才、夸夸菁领创始人、董事长兼CEO胡伟出席论坛,并带来题为《从算力到“人力”——数智员工在城市治理与产业情形的规模化上岗》的主题分享,现场正式对外发布销售大师、零售大师两款增收型智能体机型。
在本次大会前一日举办的华为全球出色合作伙伴颁奖典礼上,今日人才成功斩获“华为全球出色合作伙伴奖(公司级)”。据了解,今日人才与夸夸菁领均由胡伟创办,两家企业独立运营,聚焦“AI+数智员工”赛道深耕产业。其中,夸夸菁领为DinTal数智员工研发主体,当下该数智员工体系已在全国26座城市、2000余个岗位达成规模化落地应用,覆盖情形从城市治理逐步延伸至产业增收领域。胡伟在演讲中表示,随着算力底座持续成熟,AI价值竞争重心正在发生转移,可以创造实际业务价值的,是算力之上可承接真实岗位、直接产出业绩的“数智人力”层。
当前,海外增收型智能体赛道已迎来资本集中布局,美国Sierra、Clay两家相关企业先后完成大额融资,估值分别达158亿美元、71亿美元,资本市面走势印证企业AI应用正从传统的降本提效阶段,逐步迈入业务增收的全新发展阶段。本次发布的销售大师与零售大师,是国内AI赛道落地增收模式的关键实践。其中,销售大师依托顶尖销售经验沉淀,可为销售团队给出智能化专业支撑;零售大师聚焦新能源汽车门店零售情形,对接门店实际营收目标。两款智能体机型可完整覆盖线索发现、线索验证、外联培育、机会推进、合同执行、收入回款六大业务流程,全面衔接ML线索管理、MO机会点管理、MCE合同执行三段核心业务链路。
“提效型智能体省的是成本,增收型智能体挣的是收入。前者有上限,后者没有。”胡伟表示,数智员工的核心价值并非单纯协助企业缩减人力,而是深度嵌入企业收入链路,推动企业拓展业务、斩获更多订单。
DinTal数智员工的机型能力,源自长期严苛政务情形的实战打磨。此前,该数智员工已在深圳福田区落地43类AI岗位,覆盖民生诉求、公文处理、政务咨询等高频政务业务。福田区出台全国首创的“机器人雇员”管理办法,明确智能体政务运行规范,为机型落地给出标准化依据。当下,相关工艺与模式已复制落地至全国60余个政务项目、240余个政务情形,月均处理民生工单4.5万件、公文1.4万件,月对话量超252万次。政务情形高安全、高精准的严苛要求,持续打磨优化机型表现,为其产业化落地筑牢工艺根基。
论坛分享结束后,胡伟接受媒体专访,围绕增收型智能体赛道机遇、落地痛点、数据安全等行业热点难题作出解读。谈及国内赛道的发展机遇与挑战,胡伟表示,增收型智能体概念起源于硅谷,海外已诞生多家十亿美金级头部企业。国内赛道契合企业增收、内需提振的市面环境,当下仍处于发展早期,成长空间广阔。同时,国内企业业务流程、销售体系具备鲜明本土化特征,本土团队更熟悉国内商业逻辑,可构建适配本土企业发展的智能化机型体系。他坦言,赛道发展仍存在认知升级的挑战,需要企业转变思维,接纳AI智能体赋能业务增长,而非仅将其视作基础工具。同时他提到,近三个月市面反馈向好,企业经营者对增收类AI机型接受度持续提高,且智能体以辅助业务拓展为主、不替代人力,取得一线员工广泛认可,行业总体发展态势良好。
针对增收型智能体全链路落地的现实难点,胡伟介绍,贯通线索到回款全业务链路的核心壁垒在于业务深度融合,而非简单的大模型对话能力。增收型智能体需要深度适配岗位SOP、吃透全业务流程、沉淀行业销售数据,达成人机高质协同、标准化执行业务动作。过往多数行业AI应用仅聚焦单点提效,而全链路增收模式需要串联分散业务环节,搭建客户洞察、商机研判、思路输出、合同跟进、回款管理的完整能力体系。他表示,当前赛道竞争核心聚焦落地能力,国内AI基础工艺已日趋成熟,关键在于产业端主动拥抱AI变革、推动工艺落地业务情形。依托复杂政务情形的实战历练,团队已攻克全链路协同多项难题,相关机型已在银行对公业务、新能源汽车零售等情形落地应用,有效提高业务运营效能。
在数据安全与合规保障维度,胡伟介绍,依托长期政务情形服务经验,机型从底层架构搭建了完善的安全防护体系。DinTal数智员工达成100%国产自研内核,不依赖海外开源框架,搭建“横向三重安全保障、纵向三层安全策略”的立体化安全架构,使用沙箱隔离运行模式,达成业务全流程可管控、可追溯。当下机型已完成国家生成式人工智能服务备案,也是国内少数完成数智员工专项备案的机型体系。横向维度涵盖国家标准合规、平台权威认证、数据保险箱私有化部署三重保障;纵向维度借助业务外观、工程防护、指令护栏三层管控策略,达成输入合规管控、权限分级管理、敏感操作阻断、全流程日志审计,全方位保障企业业务数据、客户信息安全,适配政务、商业多情形合规需要。
据了解,今日人才与夸夸菁领深耕华为生态合作已有四年,从华为云出色应对思路伙伴、华为政企核心伙伴,逐步成长为华为全球出色合作伙伴。DinTal政务数智员工与华为深度联合交付,依托昇腾算力、CodeArts研发安全底座、AgentArts智能体运行时,构建起算力支撑、研发保障、平台运行、业务落地的完整产业闭环。其开源项目DinTal Claw(GenericAgent)曾登上GitHub全球Trending榜单榜首,工艺实力取得全球开发者社区认可。
胡伟表示,将来企业将延续“让数智员工做打工人,让人类员工做温度人”的发展理念,将政务情形沉淀的成熟工艺能力持续赋能产业市面。依托华为生态资源长处,持续深耕增收型智能体赛道,探索全职、外包、灵活用工之外的新型用工模式,以AI工艺推动千行百业达成高质量业务增长。
enpproperty本文来源:https://www.sznews.com/news/content/2026-09/20/content_32176125.htm
若有版权问题,请来信告知:13951431913#139.com 或在右方提交评论。
AI Vision (The following content is AI-generated and for reference only)
Shenzhen News Network, September 20, 2026 (Reporter Zhu Jiaruo) — On the afternoon of September 18, the “Artificial Intelligence Empowers High-Quality Urban Development” sub-forum of Huawei Connect 2026 was successfully held in Shanghai. Hu Wei, founder, chairman, and CEO of Shenzhen Today’s Talent and KuaKua QingLing, attended the forum and delivered a keynote speech titled “From Computing Power to ‘Manpower’ — Large-Scale Onboarding of Digital Intelligent Employees in Urban Governance and Industrial Scenarios”, officially launching the Sales Master and Retail Master intelligent body models to the public.
On the day before the conference, at the Huawei Global Outstanding Partner Awards Ceremony, Today’s Talent successfully won the “Huawei Global Outstanding Partner Award (Company Level)”. It is reported that both Today’s Talent and KuaKua QingLing were founded by Hu Wei, and both companies operate independently, focusing on the “AI + Digital Intelligent Employees” track to serve the industry. KuaKua QingLing is the R&D entity for DinTal digital intelligent employees, which have now achieved large-scale deployment in over 2000 positions across 26 cities nationwide, covering scenarios ranging from urban governance to industrial revenue generation. In his speech, Hu Wei stated that as the computing power foundation continues to mature, the center of AI value competition is shifting; what can create actual business value is the “digital intelligent manpower” layer, which sits above computing power and can directly undertake real positions to produce results.
Currently, the overseas market for revenue-generating intelligent bodies has seen concentrated capital investment, with U.S. companies Sierra and Clay securing large financings and valuations reaching $15.8 billion and $7.1 billion respectively. This market trend confirms that enterprise AI applications are transitioning from traditional cost-saving and efficiency-improving stages to a new phase of business revenue generation. The Sales Master and Retail Master models released this time represent a critical practice in the domestic AI track for revenue-generation mode landing. Sales Master leverages top-level sales experience to provide intelligent, professional support to sales teams, while Retail Master focuses on new energy vehicle store retail scenarios, aligning directly with store revenue targets. Both models fully cover the six major business processes of lead discovery, lead verification, outbound cultivation, opportunity promotion, contract execution, and revenue collection, seamlessly connecting the core ML lead management, MO opportunity management, and MCE contract execution links.
“Efficiency-improving AI saves costs, revenue-generating AI earns income. The former has an upper limit, while the latter is limitless.” Hu Wei stated that the core value of digital intelligent employees is not merely helping companies cut labor costs, but rather deeply embedding into the company’s revenue chain to drive business expansion and secure more orders.
The capability of DinTal digital intelligent employees stems from long-term, rigorous testing in governmental scenarios. Previously, the model has been deployed in 43 AI positions in Futian District, Shenzhen, covering high-frequency government services such as public complaint handling, document processing, and administrative consultation. Futian District introduced the nation’s first “Robot Employee” management policy, clearly defining the operational norms for intelligent bodies in government, providing a standardized basis for their deployment. Currently, related technologies and models have been replicated to over 60 government projects and 200+ service scenarios nationwide, processing an average of 45,000 public service tickets and 14,000 documents per month, with a monthly dialogue volume exceeding 2.52 million. The high-security, high-precision requirements of government scenarios have continuously refined model performance, laying a solid technological foundation for their industrialization.
After the forum, Hu Wei accepted a media interview to discuss the opportunities in the revenue-generating intelligent body track, implementation pain points, and data security. Regarding the development opportunities and challenges in the domestic track, Hu Wei noted that the concept of revenue-generating intelligent bodies originated in Silicon Valley, where several billion-dollar companies have already emerged. In the Chinese market, this aligns with the need for corporate revenue growth and domestic demand stimulation, presenting vast growth potential even though the sector is still in its early stages. While domestic businesses have distinct sales and process structures, local teams understand Chinese commercial logic better, allowing them to build models tailored to local enterprises. He candidly acknowledged that the track’s development still faces the challenge of mindset upgrading — companies need to accept that AI intelligent bodies are not just productivity tools but enablers of business growth. However, he reported positive market feedback over the past three months, with business leaders increasingly accepting revenue-focused AI models, and frontline staff appreciating that AI assists rather than replaces them, indicating a healthy overall trend.
On the practical challenges of full-link revenue-generation implementation, Hu Wei explained that the core barrier lies in deep business integration, not just conversational AI. A revenue-generating model must deeply adapt to job SOPs, master the entire business flow, and aggregate industry sales data to achieve high-quality human-machine collaboration and standardized execution. While past AI applications often focused on single-point efficiency, full-link revenue requires bridging separate business phases to build a complete system for customer insight, opportunity assessment, strategy output, contract follow-up, and payment collection. He emphasized that the current competition in this track is centered on implementation capability — although domestic AI tech is maturing, the key is industrial uptake and scenario-specific deployment. With experience from complex government projects, his team has overcome many challenges in full-link coordination, successfully applying these models in fields like corporate banking and new energy retail to enhance operational efficiency.
In terms of data security and compliance, Hu Wei highlighted that the model’s design, derived from governmental experience, includes robust security defenses. DinTal uses a 100% domestically developed kernel free from external open-source dependencies, employing a “three-tier horizontal security” and “three-layered vertical strategy” architecture with sandboxed execution to ensure all operations are traceable and controllable. The model has completed national generative AI service registration and is one of few with dedicated employee-bot certification. The horizontal security encompasses national standard compliance, platform certification, and private data vault deployment; the vertical layer includes interface monitoring, engineering protection, and command barriers to strictly manage permissions, block sensitive operations, and maintain comprehensive logging — ensuring compliance across both governmental and commercial environments.
Today’s Talent and KuaKua QingLing have partnered with Huawei for over four years, starting as a cloud service partner and growing into a global outstanding partner. DinTal’s governmental AI employees are jointly delivered with Huawei, leveraging Ascend compute power, CodeArts for secure development, and AgentArts for runtime support to form a complete ecosystem covering computation, development, execution, and application. Their open-source project, DinTal Claw (GenericAgent), once topped GitHub’s global trending list, affirming their technical prowess in the developer community.
Looking ahead, Hu Wei stated that the company will continue to adhere to the ethos of “letting digital employees do the work while humans bring the warmth”, extending governmental-grade capabilities to the industrial sector. By tapping into Huawei’s ecosystem strengths, they aim to deepen the revenue-generating AI track, exploring new work models beyond full-time, outsourced, or gig employment — ultimately driving high-quality business growth across industries through cutting-edge technology.
Reporter: Zhu Jiaruo Reviewer: Lu Dongbo Corrector: Zhuo Jing Editor-in-Chief: Liu Xiaoyu

