Doyen Sahoo and Sundar Vedula break down how Salesforce is translating research breakthroughs into real-world impact for enterprise AI.
Shaping the future of AI means turning what’s possible into what’s real. At Salesforce, that vision is taking form through Enterprise General Intelligence (EGI), AI designed specifically for business applications that is powerful, precise, dependable, and deeply embedded into enterprise workflows.
In this exclusive conversation, Doyen Sahoo, Director of AI Research, and Sundar Vedula, Senior Director, Software Engineering, AI Cloud, share how AI research informs the product roadmap, with innovations scaled through platforms like Agentforce, and how Salesforce is operationalising AI at scale, with India playing a pivotal role.
Q. Let’s start from the core. How is Salesforce AI Research shaping the future of AI for the enterprise?
Doyen: Salesforce AI Research is actively shaping the future of enterprise-grade AI as we work to achieve Enterprise General Intelligence (EGI) – practical, trustworthy, and scalable AI agents suited for real-world business. This term was coined by Salesforce’s Chief Scientist, Silvio Savarese.
Itai Asseo, Senior Director of Incubation and Brand Strategy, AI Research, on advancing Enterprise General Intelligence (EGI)
Enterprise-grade agents must prioritise capability and consistency, reducing so‑called “jagged intelligence”, where Large Language Models (LLMs) can answer PhD-level questions but fail on simple ones. To achieve this, Salesforce AI Research has been conducting foundational research. A notable example includes the CRMArena-Pro paper, an agent simulation environment that uses synthetic data to evaluate generic LLM agent performance on CRM workflows, helping quantify jagged intelligence to guide model refinement.
Enterprise-grade agents must prioritise capability and consistency, reducing so‑called “jagged intelligence”, where Large Language Models (LLMs) can answer PhD-level questions but fail on simple ones.
Doyen Sahoo, Director of AI Research
Q. How does your team partner with AI Research to bring innovations from lab to live deployments?
Sundar: Innovation and customer success are core values at Salesforce. Our product team works closely with the AI Research team to identify innovations ready for scaling. Once promising results are demonstrated through demos and customer pilots, we design scalable architectures which are then reviewed by stakeholders before the full software development lifecycle, including quality and performance testing, before deployment.
Examples include models for text embeddings and retrieval augmented generation (RAG), intelligent reasoning, code generation, and model benchmarking. These efforts are driving real-world capabilities—improving reliability, scalability, and delivering tangible value to customers.
Q. Can you give us a closer look at how these agents are solving complex tasks?
Doyen: Since its launch on Salesforce Help in October 2024, Agentforce has handled over 1 million support requests, delivering instant, accurate answers 24/7 with no hold times or departmental transfers. It is projected to save $50 million annually, freeing human support engineers to focus on complex cases.
This impact stems from Salesforce AI Research feeding directly into the product. Breakthroughs in RAG, intelligent prompt orchestration, and enterprise-scale AI governance enable Agentforce to ground responses in trusted documentation, knowledge articles, and customer context.
These capabilities allow Agentforce to break down complex requests into smaller steps, process them efficiently, and adapt in real time, helping to ensure speed, accuracy, and reliability in enterprise environments.
Q. How is the India team adapting Agentforce to meet local enterprise needs while supporting global scale?
Sundar: Salesforce’s AI org in India is designing and building agentic AI systems and platform components for Agentforce from India, for the world. At Salesforce, professional services, customer success, solution engineering, and other customer-facing teams work together to best serve customers by gaining a deeper understanding of their unique needs and incorporating these insights into the product roadmap.
The scale, diversity, price sensitivity, and explainability requirements in India provide invaluable feedback that helps shape solutions designed for global scalability.
Q. Can you share any examples of fundamental AI Research breakthroughs that have shaped the future of enterprise applications?
Doyen: Salesforce has long invested in foundational AI research, from early work in prompt engineering to developing families of models such as BLIP, CodeGen, xGen , and xLAM, contributions that advance the state of AI research and, in some cases, inform capabilities now embedded in Salesforce products.
For example, our reasoning research has influenced orchestration capabilities in our AI agents, helping them plan and execute multi-step business workflows with adaptive decision-making. Internally, this research powers the Atlas Reasoning Engine—the brain behind Agentforce—used to explore advanced planning and orchestration for future enterprise AI applications.
Our safety and governance research powers the Trust Layer, which enforces data privacy, compliance, and policy controls in every AI interaction. We have also advanced retrieval-augmented generation so that Agentforce is grounded in trusted enterprise data, improving accuracy, relevance, and auditability.
Looking ahead, we continue to explore next-generation research that will inform our roadmap. One example is Moirai—a time-series foundation model for universal forecasting. Moirai is currently used internally for research and prototyping to forecast resource consumption and proactively help with cloud infrastructure management.
These initiatives demonstrate the innovation pipeline that keeps Salesforce on the leading edge while helping to ensure that proven research translates into enterprise-ready capabilities.
Q. What makes India a strong testbed for validating real-world AI solutions?
Sundar: India offers a massive, diverse user base and several high-volume use cases, often within companies operating on razor-thin margins. There is also significant variation in digitisation levels across sectors, along with multilingual and cultural complexity.
This makes India an ideal proving ground for AI products that are robust, efficient, and globally scalable. A large portion of end-to-end quality and performance testing for Salesforce’s products is conducted in India.
The scale, diversity, price sensitivity, and explainability requirements in India provide invaluable feedback that helps shape solutions designed for global scalability.
Sunder Vedula, Senior Director, Software Engineering, AI Cloud
Q. Let’s talk impact. Can you share a standout example of how Agentforce can drive real business outcomes?
Sundar: An increasing number of customers are achieving significant business outcomes with Agentforce. For example, Hero FinCorp, one of the fastest growing loan companies in India, distributes an AA+ rated loan every 30 seconds, but the loan processing was highly manual and error-prone, with an average turnaround time of two days. Hero FinCorp turned to Agentforce to automate two-wheeler loan processing from start to finish.
From when an application is uploaded, Agentforce’s automation agent takes over to extract text from images, populate an opportunity in Sales Cloud, and check for issues that could slow down FinCorp’s processing. After using APIs to connect to Indian government databases to perform legal verification, it uses Mulesoft to tap into third party financial systems to instant credit checks and approvals.
Once completed, Agentforce automatically triggers the final series of flows before the loan is disbursed to the right bank – all done in just 30 minutes. This automation results in an 80% improvement in turnaround time with 75% fewer handoffs and 37% fewer form errors from the operations team, helping their customers achieve their dreams with less stress.
With Enterprise General Intelligence and Agentforce, Salesforce is translating breakthrough innovation into dependable enterprise tools—grounded in trust, shaped by real-world needs, and built for customer success.
The future of work is intelligent, and it’s already here.
Learn more:
- Read more about Salesforce’s expanding presence in India
- Find out more about Salesforce AI Research
- Read Salesforce Chief Scientist Silvio Saverese’s blog on how Enterprise General Intelligence (EGI) will form a new business imperative
- Read more about Agentforce
