As tech giants race to adopt AI, Salesforce is keeping pace. Just as Microsoft CEO Satya Nadella and Google CEO Sundar Pichai revealed that AI now writes about 30% of their companies’ code, Salesforce reports a similar trend. AI generates roughly a third of the code today, with expectations to hit 50% by year-end 2025. That shift has already translated into a rise in productivity.
“We’re already seeing 30% productivity improvements, but it’s not just about code generation,” said Muralidhar Krishnaprasad, president and CTO of Salesforce, in an exclusive interview with AIM, adding that the company is also using AI for test case generation. He explained that AI can now generate these test cases automatically by understanding the code’s logic and user flows. This saves time and helps developers catch bugs earlier in the development process.
Earlier this year, Salesforce CEO Marc Benioff announced that the company would not hire any more software engineers in 2025. He explained that the company is able to accomplish much more with fewer engineers thanks to AI-driven automation and productivity gains.
Krisnaprasad also highlighted the impact of AI even in post-deployment stages. While automated deployment is one aspect, AI can also monitor performance, detect issues, and take corrective actions, such as generating incident reports. According to Krishnaprasad, this kind of AI support spans the entire development lifecycle—from writing code and generating test cases to deployment and post-deployment monitoring.
He added that these days, he no longer has to chase down engineers for answers. Instead, he simply goes to Slack, where all design documents and videos are indexed. He asks the agent a question—like “How does this feature work?”—and gets an instant response. “These are immeasurable, brutal improvements.”
Salesforce is using AI internally, but interestingly, it announced last year that it would hire several thousand employees to sell its AI products.
Data Cloud is the Heart
Krishnaprasad dismissed the narrative that software as a service (SaaS) is under threat. “People try to sensationalise because they don’t have a good SaaS platform themselves,” he said, taking a dig at Microsoft.
He further added that every company needs a certain set of five to six functions that Salesforce provides, and they will evolve with the help of agents. He added that Salesforce has transformed from software as a service to agent as a service (AaaS).
The company launched Agentforce 2 in December last year, offering out-of-the-box autonomous agents that integrate easily with Salesforce Customer 360 platform, Data Cloud, Slack, Tableau, and third-party systems via MuleSoft and the AppExchange.
Speaking of Data Cloud, Krishnaprasad explained that it lets users aggregate structured data and index unstructured data. “We also integrated MuleSoft so that any API you have in the enterprise will now be available to agents,” he said. He added that by combining data and APIs within the platform, agents become “super smart” and context-aware across domains like sales, service, marketing, health, and automotive, making them genuinely useful for end users.
Data Cloud reached $900 million in annual recurring revenue (ARR) for fiscal year 2025. Since October last year, the company has closed 5,000 Agentforce deals, including more than 3,000 paid ones.
Krishnaparasad said Salesforce agents help with everything from identifying segments, creating campaigns, generating leads (via SDRs/BDRs), nurturing them, and closing sales.
In order to improve accuracy and reduce hallucinations, Salesforce uses its Atlas Reasoning Engine and Retrieval Augmented Generation (RAG) to deliver personalised responses grounded in both structured and unstructured data.
Data Cloud plays a key role by integrating disparate data sources to create a unified view, breaking down silos without requiring full data duplication.
Beyond integration, Data Cloud also creates unified customer profiles, helping avoid fragmented experiences, like being targeted with ads for products one has already bought.
Krishnaprasad noted that the platform serves as the foundational system for logging all interactions, including agent flows and conversations. “We also use Data Cloud as our logging substrate for all agent flows. Every single conversation you’re having is stored in Data Cloud,” he said.
Model Context Protocol and Google A2A
Krishnaprasad further shared that Salesforce is experimenting with the Model Context Protocol (MCP) and has launched new software that allows customers to convert any API into an MCP with the click of a button.
Moreover, he added that Heroku can now host and run any MCP server, and that the company’s Agentforce platform supports MCP. Heroku is a cloud platform as a service (PaaS) that allows developers to build, run, and scale applications quickly without managing infrastructure.
Salesforce is working with Google on the Agent2Agent (A2A) protocol to set industry standards. The company has also partnered with Google to integrate Gemini AI into Salesforce’s Agentforce, allowing agents to process images, audio, and video, handle complex tasks, and provide real-time insights using Google Search and Vertex AI.
Reflecting on these advancements, Krishnaprasad compared the rise of AI agents to the early days of cloud computing. He said AI agents are now at a similar inflection point as the early days of cloud computing. He believes they could deliver even greater impact—up to 10 times more—in terms of productivity and job creation.
The post Salesforce Targets 50% AI-Generated Code by End of 2025 appeared first on Analytics India Magazine.
