How Salesforce is Rethinking Its ‘Agentic Workforce’ to Deliver ROI
By Siddharth Jindal ·
Enterprise AI is moving beyond copilots and chat interfaces into systems that can actually execute work. The attention is now on agentic AI, which has evolved from assisting users to completing tasks end-to-end.
For companies, this means faster operations, better customer outcomes, and software that actively participates in business processes rather than waiting for instructions.
Salesforce is positioning itself at the centre of this transition, building tools that allow developers to create intelligent agents grounded in enterprise data.
In an exclusive interaction with AIM, Deepak Pargaonkar, Vice President of Solution Engineering at Salesforce, explains how the company is moving development beyond writing static code to orchestrating dynamic, intelligent agents.
Building Agents, Not Just Code
At Salesforce, developers are increasingly tasked with creating systems that can understand business context, execute workflows, and interact across multiple systems.
A key component of this vision is Agentforce Vibes. Launched in October last year, the enterprise-grade vibe coding tool allows developers to describe tasks in natural language. From there, the platform generates apps, defines workflows, and executes business processes.
Pargaonkar says developers already combine coding with business understanding, but agentic systems push them towards defining intent instead of implementation. With organisation-specific data, these agents operate with strong contextual awareness.
The most immediate impact is being seen in customer-facing services, which are directly tied to revenue and operational efficiency.
“Customer engagement and customer experience processes are where organisations are seeing the largest impact,” he says. “If you solve customer issues quickly and deliver a better experience, it directly links to both top line and bottom line.”
Pargaonkar points to use cases across customer acquisition, service, and marketing as early beneficiaries. Faster resolution times, improved personalisation, and lower service costs are driving tangible business outcomes.
For instance, Hero FinCorp is using Agentforce to cut loan approval times from two days to just 30 minutes. The system automates the entire workflow from application to underwriting and payment, with 75% of cases requiring no manual intervention.
Meanwhile, Air India Air has deployed Agentforce to handle refund workflows. The tool manages customer queries, collects details, shares updates, and initiates payments with minimal human involvement. As a result, Pargaonkar says about 90% of cases are being resolved without manual intervention, reducing refund timelines and improving the overall customer experience.
However, the executive clarifies that these agents are not going to take away the jobs. “We always talk about humans plus agents,” he says, adding that the intent is to offload routine work rather than replace people.
What Sets Salesforce Apart
When asked how Salesforce differentiates itself from competitors like Microsoft Copilot, Pargaonkar emphasises the importance of data context and trust.
“One of the most important things in agentic capability is the criticality of context and data,” he says. “Our entire experience is grounded in Data 360, which brings all organisational data together.”
Salesforce Data 360 (formerly Data Cloud) is a native customer data platform that unifies fragmented customer data—including from legacy systems, websites, and data lakes—to provide personalised experiences.
Pargaonkar explains that this grounding ensures that AI agents operate with highly relevant, organisation-specific information.
Beyond that, Salesforce has built a strong emphasis on security and governance through its trust layer. “We anonymise customer data before it is sent to LLMs, and we have zero-retention agreements with model providers,” he says. “That ensures your data is not used to train external models.”
The combination of contextual grounding, built-in guardrails, and data privacy forms a core part of Salesforce’s enterprise pitch.
Rethinking ITSM and Expanding to Voice
Salesforce is also targeting the IT service management (ITSM) space, which it believes is due for a rethink. The company argues that the legacy market, projected to grow to $30 billion by 2030, according to Grand View Research, and dominated by players like ServiceNow, still relies heavily on slow, ticket-based systems that delay issue resolution.
Salesforce’s answer is Agentforce IT Service, a Slack-native, agent-first platform launched in October last year and designed to automate employee support, resolve issues through conversational AI, and manage IT infrastructure. It enables employees to interact directly with agents across channels like Slack, portals, or web chat.
“The entire organisational front end becomes Slack,” Pargaonkar says, positioning it as a central hub for employee interactions. He adds that this approach not only improves response times but also helps prevent issues proactively while reducing support costs.
The offering features over 100 pre-built integrations, connectors, and workflows. By automating incident management, asset tracking, and resource allocation, the platform aims to reduce manual workload and allows teams to focus on higher-value tasks such as problem-solving and innovation.
Salesforce is now taking its agentic push into newer areas. One of the key launches is Agentforce Voice, which brings these capabilities into voice interactions, something that matters a lot in markets like India.
“Voice remains a critical channel,” Pargaonkar says. “We want customers to leverage agentic capabilities through voice as well.”
Salesforce is also investing in multi-agent ecosystems through offerings like MuleSoft, which connects SaaS and enterprise applications in the cloud and on-premise. “MuleSoft Agent Fabric is an amazing solution for enterprises to operate in multi-agent environments and become truly agentic,” the executive says.
Taken together, Salesforce’s strategy is to move software beyond being just a tool to actually taking part in business processes.