The next battleground in AI is not about replacing humans but about scaling what they can achieve, with AI agents handling routine tasks, analysing information and acting autonomously so people can focus on strategy, innovation and growth.
Salesforce calls this shift the rise of the “agentic enterprise,” where AI works as a tireless, always-on partner that never forgets a detail or keeps a customer waiting. Organisations that deploy such AI agents across the business will move faster, serve better and outpace competitors still relying on human speed alone.
Arun Kumar Parameswaran, Salesforce’s executive vice president and managing director for South and Southeast Asia, identifies three categories of AI impact: efficiency and productivity, customer experience enhancement and top-line growth through market expansion and share gains.
“The biggest opportunity will be in top-line growth,” he says. “Those use cases get green-lighted by the chief financial officer (CFO) much faster than efficiency plays, because the moment you say I’m going to show two points of market share gain, CFOs are more likely to approve that.”
Augmentation, not automation
What distinguishes the agentic enterprise vision from previous automation efforts is the explicit framing of AI as augmentation rather than replacement.
See also: Is Alibaba among the most undervalued AI companies?
Singapore insurer Singlife exemplifies this approach. The company is deploying Agentforce to give customer service staff instant access to product information that previously required time-consuming searches through dense manuals and policy documents.
“At Singlife, AI is more than just a tool. It is becoming a key part of how we run the business,” says Romil Sharma, Singlife’s group head of technology and operations. “Collaborating with Salesforce allows us to bring AI into the hands of our customer service executives in a practical way, helping them respond faster and with greater confidence.”
Sharma emphasises that it’s augmentation, not elimination. “Singlife takes a very human-centric approach to AI, wherein we use AI to enable employees to be more efficient and productive. If people end up saving 30% of their time on a task, it doesn’t mean I will cut down 30% of the workforce. Instead, it means these people can service more.”
See also: Univers opens global AI lab in Singapore to help drive down energy use and emissions
The strategic value extends beyond internal efficiency. “In the insurance business, sales and service go hand in hand. If my sales team is able to cut down 30% of their time, they can sell more. And I don’t need to increase the capacity of my servicing team because they’re also 30% more productive through AI,” adds Sharma.
This multiplication effect represents what companies are chasing. Not replacing humans with machines, but creating human-machine teams that accomplish what neither could alone.
Singlife plans to extend Agentforce beyond its customer service executives to its network of financial adviser representatives. “Ultimately, our aim is to institutionalise AI as a core enabler for business growth, operational excellence and customer-centricity,” says Sharma.
The platform requirements
Making AI agents genuinely helpful requires solving technical challenges that have stymied previous automation efforts. Salesforce positions its Agentforce 360 as addressing three critical requirements: unified data, reliable reasoning and enterprise-grade governance.
At the foundation sits Data Cloud, the company’s data unification engine. Unlike traditional approaches requiring companies to consolidate information into a single repository, Salesforce’s “zero-copy” architecture leaves data where it lives in policy systems, customer relationship management (CRM) databases and document repositories while creating a unified view that agents can query.
“We’re not asking you to move all your data into the Data Cloud. Leave the data where it is at rest. We’ll surface the data, take the actions we need to take, and we’re very happy to work with the understanding that there will be multiple data lakes,” says Salesforce’s Parameswaran.
To stay ahead of the latest tech trends, click here for DigitalEdge Section
He adds that this matters because enterprises will never consolidate everything into one system, despite vendor promises. So, the question becomes how to work with that fragmented reality while still delivering intelligent outcomes.
On top of that data foundation, Agentforce 360 provides the reasoning engine that powers autonomous behaviour. New capabilities include conversational development tools for building agents using natural language and hybrid reasoning that combines deterministic workflows with flexible AI decision-making.
The third critical layer is governance and observability. Dashboards allow companies to monitor how AI agents reason through decisions, track accuracy over time and ensure compliance with regulations and internal policies.
Multi-modal and multi-agent
To enhance contextual understanding, Salesforce has equipped Agentforce 360 with voice capabilities, enabling AI agents to interpret and act on a mix of inputs such as text, voice, images and video, just like humans.
The practical applications span industries. A field service technician could record a sputtering car engine and ask an AI agent to diagnose the problem. Marketing agents could analyse social media video trends in real time. Customer service agents could detect negative sentiment from a caller’s tone of voice apart from their words.
The agentic enterprise vision goes further. Salesforce envisions specialised AI agents collaborating seamlessly, with orchestrator agents directing and coordinating activities and managing multistep tasks through to completion.
In an enterprise context, a service AI agent could process a customer’s inquiry while an inventory AI agent checks product availability. A logistics AI agent calculates shipping, and a billing agent reviews payment options. The AI orchestrator coordinates all these inputs into a coherent response.
The interoperability and memory challenge
This multi-agent collaboration could extend across organisations. Imagine a global supply chain where AI agents from different companies coordinate logistics, source materials, and manage distribution, reacting in real time to delays and disruptions.
However, that cross-company coordination requires interoperability, particularly as enterprises will most likely adopt AI agents from multiple technology vendors. “We’re assuming it will be a multi-agent world, so we’ll have to interoperate with AI agents from other platforms. We’ve always been an open architecture, and we’re not going into this with the view of world domination,” says Parameswaran.
Besides interoperability, AI agents must also overcome a deeper constraint: short-term memory. Most still struggle to retain and share context across interactions, limiting their ability to collaborate effectively.
Salesforce’s Data Cloud addresses this by storing both short-term and long-term memory, with the latter referring to context built during individual transactions that expands as AI agents interact with each other over time. This enables AI agents to handle multistep processes while building better customer relationships by recalling preferences, problems and past interactions.
Long-term memory also helps teams of agents collaborate more effectively by sharing knowledge across functions. A legal AI agent and logistics AI agent could work together to onboard a new partner without duplicating requests or missing steps.
“That’s the power of our Agentforce 360 platform,” claims Parameswaran. “We’re able to build that context over many engagements, so when agents start interacting with each other, you get the power of that context and memory.”
While true long-term coherence isn’t fully achieved yet, Salesforce researchers say progress is accelerating. As AI becomes more autonomous, human oversight remains critical. According to Salesforce AI Research experts, these systems will require sophisticated coordination, orchestration and governance by both humans and AI, with humans evolving to focus on strategic guidance.
The regional reality check
For Southeast Asia, adoption timelines face a fundamental economic challenge. Parameswaran notes that “the cost of the human agent is still lower than the cost of AI… in many countries in the region”, unlike in North America, where technology delivers clear cost savings.
Yet, he expects the equation to flip as AI costs continue falling. This cost reality reinforces why top-line growth use cases, like Singlife’s market share strategy, will drive regional adoption more than pure efficiency plays.
The region’s less mature digital infrastructure could paradoxically prove advantageous. “We’re talking to organisations that never had a CRM or a real, scalable contact centre solution. So, they want those solutions AI-enabled from [the first day they adopt them]. That leapfrog is an advantage we will see,” says Parameswaran.
Whether the agentic enterprise succeeds depends on companies balancing speed with accuracy, governance and employee adoption. The payoff for those who get it right is a fundamental competitive repositioning rather than just productivity gains.
