The human behind the machine: Salesforce’s vision for AI-powered customer service
Salesforce sees AI agents reshaping customer service by handling routine work while humans focus on complex cases, judgement, empathy, and customer relationships.
Dataquest Bureau18 Sep 2026
15:44IST
New Update
As artificial intelligence (AI) agents take on routine customer-service interactions, the role of human service representatives is beginning to change. Rather than replacing people, the shift is creating a model in which AI handles repetitive work while employees focus on complex problems, judgement, and empathy.
On the sidelines of Dreamforce 2026, Shipra Sinha, Senior Analyst at CyberMedia Research (CMR), spoke with Prasad Raje, SVP, Service Cloud, Product Management at Salesforce, about how AI could reshape customer service, from agent-assisted workflows to autonomous interactions.
Raje compares the transition with the adoption of Excel. While there were concerns that spreadsheets could reduce the need for financial analysts, the technology instead became a tool that allowed professionals to work more efficiently.
“AI is just a power tool in the hands of humans to do greater things,” he says.
Connecting AI with customer context
For Salesforce, the move towards an agentic enterprise depends on combining the capabilities of large language models (LLMs) with the structured information available in customer relationship management (CRM) systems.
Raje sees Agentforce as the layer connecting these capabilities. AI agents can draw on CRM data, company knowledge, policies, and procedures to generate responses based on a specific customer and business context rather than relying only on general-purpose AI capabilities.
This distinction becomes important as enterprises move from AI experimentation to deployment. An agent that can access relevant business information can potentially perform tasks within defined organisational processes, rather than simply generate an answer.
Building a more complete customer view
Fragmented data remains a challenge for customer-service teams. Information can be distributed across CRM platforms, databases, data warehouses, data lakes, and unstructured repositories.
Salesforce’s Data 360 is designed to bring these sources together into a more consistent customer view. Its data graph can connect information such as marketing interactions, purchases, service tickets, and complaints, including information that remains outside Salesforce.
The same context can support both AI and human representatives. Instead of asking customers to repeatedly explain previous interactions, service teams can work from a broader history of the relationship.
That could also make personalisation more practical, particularly when an interaction involves multiple departments or previous service issues.
India adds a language challenge
India presents a significant opportunity for AI-enabled customer service because of its scale and linguistic diversity. Raje says Salesforce is expanding its AI capabilities across languages, including Hindi, while working towards broader coverage.
The challenge, however, extends beyond translation. Customer expectations and business practices can vary across regions, while some Tier 2 and Tier 3 Indian languages remain underrepresented in AI training datasets.
According to Raje, the underlying technology can support these languages, but reliable deployment depends on access to sufficient, high-quality, region-specific data.
This makes language capability a data problem as much as a model problem. For customer-service applications, accuracy and contextual understanding can be critical because an incorrect response can directly affect the customer relationship.
AI changes how service is measured
Traditional customer-service metrics are unlikely to disappear as AI adoption increases. Customer satisfaction (CSAT) and Net Promoter Score (NPS) continue to provide insight into customer experience, while average handle time and time to first response remain relevant operational measures.
AI, however, creates another way to analyse interactions.
Instead of relying exclusively on surveys after an interaction, AI can analyse conversations to identify customer sentiment. Salesforce is also developing moment-by-moment sentiment capabilities that can help businesses understand how customer sentiment changes during interactions.
For service organisations, this could shift measurement from a primarily post-interaction exercise towards more continuous analysis of customer conversations.
Trust becomes a prerequisite for autonomy
Greater autonomy also brings questions around security, privacy, and accuracy. Raje argues that these considerations need to be built into the foundation of AI deployment.
Salesforce’s trust layer includes controls around data security, permissions, and privacy. The company says it uses Zero Data Retention (ZDR) arrangements with its large language model providers, while extending enterprise access controls to AI agents.
Businesses can determine what information an agent can access and which actions it is permitted to take. The platform also monitors interactions for risks including prompt injection and inappropriate content.
The underlying principle is that greater autonomy needs to operate within defined enterprise boundaries. For businesses, the question is therefore not simply what an AI agent can do, but what it should be allowed to do.
From service desk to customer relationship
The changing role of Service Cloud is particularly relevant to sectors such as automotive, banking, financial services and insurance (BFSI), and retail, where customer relationships extend beyond individual service interactions.
Salesforce’s industry-specific offerings bring sector-specific information into the service experience. This can help businesses move beyond resolving individual cases towards understanding the broader customer relationship.
In this model, customer service becomes part of a longer-term engagement strategy rather than remaining a standalone support function.
What customer service could look like in 2030
Raje expects customer-service organisations to become increasingly “agentified” by 2030, with autonomous AI handling a larger share of interactions across channels.
Websites could also evolve from traditional search-driven interfaces towards conversational and agentic experiences.
Human representatives, meanwhile, could increasingly work alongside AI assistants. One potential benefit is reducing the experience gap between new and experienced employees. An AI assistant could provide a newer representative with access to organisational knowledge, processes, and guidance that would otherwise take years to develop through experience.
This could be particularly relevant for organisations that rely on seasonal or rapidly expanding service workforces.
For Indian customer-service leaders beginning their AI journey, Raje’s advice is direct: “Start building agents.”
The broader shift is not necessarily a choice between humans and AI. Instead, customer-service organisations are beginning to explore how AI can provide scale and speed while human employees focus on interactions that require context, judgement, and empathy.
