Skills, workflows and governance: What businesses need to make AI work

Moving beyond experimentation means rethinking jobs, employee support and technology guardrails, says Salesforce Asean senior vice-president and general manager Paul Carvouni

Building the right skills and redesigning workflows can help employees work more effectively alongside AI as it becomes a bigger part of the workplace.

Published Sep 11, 2026, 04:00 AM

For decades, Singapore has made a habit of moving early when new technologies emerge. Digital services have become embedded in everyday life and, more recently, the country has laid out ambitions to deepen its use of artificial intelligence through initiatives such as National AI Strategy 2.0 .

But inside the workplace, the transition may not be moving quite as quickly.

As businesses explore the next phase of agentic AI – where autonomous AI agents can reason, make decisions and carry out tasks – employees are already anticipating how their jobs could change. A Salesforce survey found that 97 per cent of workers expect to work with an AI agent.

This suggests the challenge may be less about outright resistance to AI than uncertainty over what it could mean for individual workers: how it might change their jobs, what tasks it could take on and, crucially, what skills they will need to work alongside it.

For businesses, closing this gap will require more than experimenting with the technology. They will also need to prepare their people and rethink how work gets done.

A Salesforce survey found that 97 per cent of Asean workers expect to use AI agents at work, but only 32 per cent receive company training on how to use them. PHOTO: SALESFORCE

A Salesforce survey found that 97 per cent of Asean workers expect to use AI agents at work, but only 32 per cent receive company training on how to use them.

Why businesses need to move beyond AI pilots

There are understandable reasons for businesses to proceed cautiously with AI.

Although the technology is evolving quickly, organisations must weigh considerations ranging from data privacy and security to the accuracy and safety of AI-generated responses. A lapse in any of these areas can have consequences for customers as well as a company’s reputation.

The risk, however, is that caution turns into a prolonged cycle of small-scale pilots, without organisations taking the next step towards wider adoption.

When that happens, businesses may struggle to realise the value of AI while employees are left knowing that change is coming, but without a clear understanding of how it will affect their work.

Some workers may instead turn to AI tools and use them in their personal capacity to build familiarity. But what works for individual use does not necessarily translate neatly into the workplace, where considerations such as security, privacy and governance become much more important.

Why agentic AI is different from earlier technology shifts

AI should not be treated as just another technology cycle that businesses can wait out until there is greater certainty. There is a generational shift in the nature of work.

Previous waves of workplace technology largely helped people perform existing tasks faster or more efficiently. Agentic AI could go further by taking responsibility for certain tasks and processes itself, changing the role people play within them.

That has particular implications for Singapore. With a small domestic labour force and an ageing population, finding ways for workers to become more productive will remain important to the economy.

The question, then, is not simply whether businesses should adopt AI, but how they can do so in a way that allows their workforce to adapt alongside it.

How businesses can prepare for agentic AI

Regardless of which AI models or tools ultimately prevail, successful adoption is likely to depend on three areas progressing together: workforce skills, redesigned workflows and trust.

First, businesses need to go beyond basic AI literacy.

Knowing how to use an AI tool to summarise a document or draft an email is useful, but the bigger opportunity lies in developing workers who can combine their professional expertise with an understanding of how AI can be applied to their jobs.

Singapore’s National AI Impact Programme , for instance, aims to help companies adopt AI while equipping workers with the skills to use it. Businesses can complement such efforts by moving beyond one-off training sessions and showing employees how AI can be incorporated into their day-to-day work.

That may also mean reviewing job scopes and team structures as people begin working more closely with AI agents.

The second step is to rethink workflows.

Simply adding AI to an existing process may deliver incremental improvements, but larger gains can come from redesigning the process itself – identifying which routine tasks can be handled by AI and where human judgment, expertise and interaction remain most valuable.

Take, for example, the use of customer service in insurance, where a human service representative might use AI to help draft a response. A redesigned workflow could instead allow an AI agent to handle routine queries and complete certain steps independently, before passing more complex cases to a human agent.

Health maintenance organisation Maxicare in the Philippines has adopted such an approach for its Letter of Authorisation process, with AI agents handling parts of the customer journey and escalating cases when necessary. According to the company, this has cut service time by 97.5 per cent.

The third consideration is trust and governance.

Customers need to know that their data is protected, while employees need clarity on how AI systems are being used. Regulators, meanwhile, expect businesses to remain accountable for their systems and processes.

Slackbot, Salesforce’s AI agent for work, can search Salesforce data and surface answers directly in Slack, with governance managed centrally through Agentforce. PHOTO: SALESFORCE

Slackbot, Salesforce’s AI agent for work, can search Salesforce data and surface answers directly in Slack, with governance managed centrally through Agentforce.

This becomes more complex with large language models (LLMs), whose responses are probabilistic rather than determined by a fixed set of rules. The same input may not always produce exactly the same result.

As AI becomes more widely used across organisations, governance therefore needs to scale alongside it. Businesses need ways to set controls, monitor how AI systems are being used and manage issues such as risk, security and compliance.

Salesforce’s Agentforce, for example, is designed to give organisations a centralised way to manage AI agents and the governance surrounding their use.

How business leaders can equip their people for AI

For Singapore’s business leaders, one of the clearest signals from the shift towards agentic AI is that the workforce may be more ready for change than some assume.

But expecting to work with AI is not the same as knowing how to do so effectively.

Businesses will need to give employees a clearer picture of how AI will change their work, equip them with the skills to use it and redesign processes so that people and AI can each play to their strengths.

Companies that remain stuck in experimentation risk missing not only potential productivity gains, but also the opportunity to prepare their workforce for a broader change already under way.

The next phase of AI adoption will therefore be as much about people and the way work is organised as it is about the technology itself.

Find out how you can transform your business into an agentic enterprise with Salesforce.

This article was written by Salesforce Asean senior vice-president and general manager Paul Carvouni.

Artificial Intelligence

Salesforce东盟高级副总裁兼总经理保罗·卡沃尼表示，超越实验阶段意味着重新思考工作岗位、员工支持和技术保障措施。

随着人工智能在工作场所中扮演越来越重要的角色，培养正确的技能和重新设计工作流程可以帮助员工更有效地与人工智能协同工作。

发布于 2026 年 9 月 11 日 上午 4:00

几十年来，新加坡一直习惯于在新技术出现时率先行动。数字服务已融入日常生活，而近年来，新加坡更是雄心勃勃地通过诸如国家人工智能战略2.0等举措，深化人工智能的应用。

但在工作场所内部，这种转变可能不会那么快。

随着企业探索下一阶段的智能体人工智能——即能够自主推理、决策和执行任务的人工智能代理——员工们已经开始预见自身工作可能发生的变化。Salesforce 的一项调查发现，97% 的员工希望未来能与人工智能代理协同工作。

这表明，挑战可能不在于对人工智能的直接抵制，而在于对人工智能可能对个体劳动者意味着什么的不确定性：人工智能可能会如何改变他们的工作，它可以承担哪些任务，以及至关重要的是，他们需要哪些技能才能与人工智能协同工作。

对企业而言，弥合这一差距需要的不仅仅是技术试验，还需要培养员工并重新思考工作方式。

Salesforce的一项调查发现，97%的东盟员工希望在工作中使用人工智能助手，但只有32%的员工接受过公司提供的相关培训。图片来源：Salesforce

Salesforce 的一项调查发现，97% 的东盟员工希望在工作中使用人工智能代理，但只有 32% 的员工接受过公司关于如何使用人工智能代理的培训。

为什么企业需要超越人工智能试点项目

企业在人工智能领域采取谨慎态度是有可以理解的。

尽管技术发展迅速，但企业必须权衡诸多因素，包括数据隐私和安全，以及人工智能生成回复的准确性和安全性。任何一个环节的疏忽都可能对客户和公司声誉造成影响。

然而，这种谨慎可能会演变成小规模试点项目的长期循环，而各组织却没有采取下一步行动，实现更广泛的采用。

当这种情况发生时，企业可能难以实现人工智能的价值，而员工虽然知道变革即将到来，但却不清楚变革将如何影响他们的工作。

有些员工可能会转而使用人工智能工具，以熟悉其功能并进行个人使用。但适用于个人用途的方法并不一定能直接应用于工作场所，因为在工作场所，安全、隐私和治理等因素更为重要。

为什么智能体人工智能与早期的技术变革有所不同

人工智能不应被视为企业可以坐等前景更加明朗的又一个技术周期。它正在引发工作性质的代际转变。

以往的职场技术浪潮主要帮助人们更快、更高效地完成现有任务。而智能体人工智能则更进一步，它能够承担某些任务和流程的自主权，从而改变人们在其中扮演的角色。

这对新加坡来说尤其具有重要意义。新加坡国内劳动力规模较小，人口老龄化严重，因此，如何提高劳动者的生产力对新加坡经济仍然至关重要。

因此，问题不仅仅是企业是否应该采用人工智能，而是如何以一种能够让员工与之同步适应的方式来采用人工智能。

企业如何为智能体人工智能做好准备

无论最终哪种人工智能模型或工具胜出，成功应用可能取决于三个领域的共同进步：劳动力技能、重新设计的工作流程和信任。

首先，企业需要超越基本的AI素养。

了解如何使用人工智能工具来总结文档或撰写电子邮件固然有用，但更大的机遇在于培养能够将专业知识与对人工智能如何应用于其工作的理解相结合的员工。

例如，新加坡的国家人工智能影响计划旨在帮助企业采用人工智能，同时为员工提供使用人工智能的技能。企业可以通过提供更全面的培训，例如不局限于一次性的培训课程，而是向员工展示如何将人工智能融入到日常工作中，来进一步完善这些努力。

随着人们开始与人工智能代理更紧密地合作，这可能也意味着需要重新审视工作范围和团队结构。

第二步是重新思考工作流程。

简单地将人工智能添加到现有流程中可能会带来渐进式的改进，但更大的收益来自于重新设计流程本身——确定哪些日常任务可以由人工智能处理，以及在哪些方面人类的判断、专业知识和互动仍然最有价值。

以保险行业的客户服务为例，客服代表可能会使用人工智能来辅助撰写回复。重新设计的流程可以让人工智能代理处理常规查询并独立完成某些步骤，然后再将更复杂的案例转交给人工客服。

菲律宾的健康维护组织Maxicare在其授权书流程中采用了这种方法，由人工智能代理处理客户流程的部分环节，并在必要时上报问题。据该公司称，此举已将服务时间缩短了97.5%。

第三个考虑因素是信任和治理。

客户需要了解他们的数据是否受到保护，而员工则需要清楚地了解人工智能系统的使用方式。与此同时，监管机构希望企业对其系统和流程负责。

Salesforce 的企业级 AI 代理 Slackbot 可以搜索 Salesforce 数据，并将答案直接显示在 Slack 中，所有管理工作均由 Agentforce 集中完成。图片：Salesforce

Slackbot 是 Salesforce 的 AI 工作代理，可以搜索 Salesforce 数据并在 Slack 中直接显示答案，并通过 Agentforce 进行集中管理。

对于大型语言模型（LLM）而言，情况会变得更加复杂，因为它们的响应是概率性的，而非由一组固定的规则决定。相同的输入未必总能​​产生完全相同的结果。

随着人工智能在各组织机构中得到更广泛的应用，治理也需要随之扩展。企业需要采取措施来建立控制机制，监控人工智能系统的使用情况，并管理风险、安全和合规性等问题。

例如，Salesforce 的 Agentforce 旨在为组织提供一种集中管理 AI 代理及其使用相关治理的方式。

企业领导者如何才能让员工做好应对人工智能的准备？

对于新加坡的商业领袖来说，向智能人工智能转变所发出的最明确的信号之一是，员工可能比一些人想象的更能接受变革。

但期望与人工智能合作并不等同于知道如何有效地与人工智能合作。

企业需要让员工更清楚地了解人工智能将如何改变他们的工作，让他们掌握使用人工智能的技能，并重新设计流程，以便人和人工智能都能发挥各自的优势。

仍然固守实验模式的公司不仅可能错失潜在的生产力提升机会，还可能错失让员工为正在进行的更广泛变革做好准备的机会。

因此，人工智能应用的下一阶段不仅关乎技术本身，也关乎人以及工作的组织方式。

了解如何利用 Salesforce 将您的企业转型为代理型企业。

本文由 Salesforce 东盟高级副总裁兼总经理 Paul Carvouni 撰写。

人工智能