Join the event that the leaders of the institutions have been trusted for nearly two decades. VB Transform combines people who build AI’s strategy for real institutions. Learn more
Salesforce Comprehensive improvements were put on their AI’s agent platform on Monday, as they address the largest obstacles facing institutions when digital workers are widely published: knowing what these agents are already doing and ensuring that they can work safely through corporate systems.
Company Agentforce 3 The release provides a comprehensive “Command center“This gives executives in the actual time in the performance of the artificial intelligence agent, in addition to the original support for the emerging inter -operating standards that allow agents to communicate with hundreds of external business tools without dedicated coding.
The timing reflects the increased demand for institutions for artificial intelligence agents. According to the Slesforce data, the use of artificial intelligence agent jumped by 233 % in six months, as more than 8,000 customers registered to spread technology. The first adoption sees measurable returns: engine Reducing the time of dealing with customer cases by 15 %, while 1-800accountant 70 % achieved independent decision for administrative chat requests during the peak tax season.
“We have hundreds of live applications, if not thousands, and they run widely,” said Jayesh Govindarajan, EVP of the Salesforce AI, in an exclusive interview with Venturebeat. He pointed out that the company has moved decisively than experimental publishing operations: “Artificial intelligence agents are no longer experimental. They deeply moved to the fabric of the institution.”
“During the past few months, we have listened deeply to our customers and we have continued in our quick pace to innovate technology,” Adam Evans, Executive Vice President and Director General of Salesforce AI said in announcing the upgrade of the platform during a direct event on Monday. “The result is Agentforce 3, and it is a great leap forward for our platform that brings greater intelligence, higher performance, more confidence and accountability for every agent’s publication.”
How Pepsico Giant World Food leads the AI’s agent’s revolution
Among the companies that build this technology Pepsico,, Which publishes Agentforce as part of a wider transformation of AI in its global operations. In an exclusive interview with Venturebeat, Athina Kanyoura, chief strategy and transformation officials in Pepsico, described the publication of the company’s development in an increasingly complex market.
“As a long -time partner in Salesforce, we have realized an opportunity to combine the way we use its platforms in our business – especially with the development of the customer scene, the trade becomes more complicated, and the need to better integrate our data,” said Kaneoura Venturebeat.
The food and beverage giant, whose products consume more than a billion times a day around the world, believes that artificial intelligence agents are necessary to meet customers “where they – and in the ways they want to deal with us”, while leading the efficiency of the back interface by integrating systems and simplifying operations.
The seven -year -old Pepsico relationship with Salesforce has put the company to quickly move to artificial intelligence agents. “We were excited about how to strengthen Agentforce from our daily field experience-simplifying the workflow and leaking the most intelligent ideas in an actual time,” Caniona explained.
Lost piece: Why do AI need to monitor and control actual time
the Command center The Salesforce response to the critical gap in the AI Enterprise market. While companies rushed to the deployment of artificial intelligence agents for customer service, sales and operational tasks, many of them lack how these digital workers perform or affect business results.
Govindarajan described the challenge facing institutions that have passed experimental programs: “It is one thing to build a experimental offer for AI, but when you already build an agent system and put it in front of users, there is a different standard.” He explained that companies need tools to understand when artificial intelligence agents struggle and when they bring people to the workflow.
“The teams cannot know what the agents do – or their development quickly enough,” the company recognized its announcement. The new observation platform provides detailed analyzes on the agent’s reactions, health monitoring with actual time alerts, and recommendations on behalf of improvement.
The system addresses what Govindarajan calls “today the two problems” – the operational challenges that appear after the initial publication. He said: “You can have many agents for multiple people, and you should be able to note how this already affects the task that must be done widely.” This includes the management of delivery operations between digital agents and human workers when complex decisions or approvals are needed.
The system picks up all the activities of the agent in Salesforce’s Data cloud Using Opentil’s measurement Enable integration with current monitoring tools such as DATADOG And other institutions systems. This addresses the need for AI’s agent’s supervision in the current operating workflow.
Open standards and safe integration: How artificial intelligence agents call through institutions systems
Perhaps most importantly is the embrace of Salesforce Form context protocol (MCP), an open -emerging standard for AI Agent action. The articles of association will include the original MCP support, allowing AgentForce to communicate with any server compatible with MCP without dedicated development work.
“There is a general inter -employment capacity, then there is what we call the interviewable operation at the level of the institution,” explained by Gary Liraubt, Vice President of the Product Architecture at Salesforce, in an exclusive interview with Venturebeat. “If the degree of institutions are not, this is similar to the unreliable Interop.” He said that the main difference lies in the mechanisms of governance and control that institution agents need.
This possibility, which works alongside the expanded Agentexchange market Amazon web servicesand fundand Google cloudand IBMand PayPalAnd tape. Lerhaupt said that the company will be launched with “North 20, Perhaps 25 Plus”, MCP servers have been examined, with partners like PayPal providing capacity and box bills that provide access to documents through their MCP applications.
“In a world full of artificial intelligence tools, the Fores agent not only emerged from the unique technology, but also corresponds to our technological ecosystem, the way we work and the strategy of our artificial intelligence, our standards and workframe.”
Performing Payment: AI Models Faster and Future Security for Organized Industries
The latent behind the new features are what Salesforce calls improved.atlas“Architecture designed for performance and safety at the level of the institution. The statute now provides 50 % less than the arrival time compared to January 2025, the response to the user’s experiences in actual time, and the spontaneous failure between artificial intelligence models to ensure continuous operation.
For organized industries, the Salesforce approach to hosting artificial intelligence models directly in infrastructure address important security concerns. “With Anthropor, the stack will be fully operated within the Salesforce infrastructure,” Govindaraajan explained. “The calls do not come out to Openai, and traffic will work within the Salesforce VPC. For organized industries, this is what we were working on.”
Critically for organized industries, Saleforce is now hosting Human models Directly inside its infrastructure via Amazon BedrockMaintaining sensitive data within the selesforce security perimeter. The company plans to add Google Gemini models Later this year, institutions have given more options to the governance of the artificial intelligence model.
The platform also expanded global availability to Canada, the United Kingdom, India, Japan and Brazil, with the support of six additional languages, including French, German, Spanish, Italian, Japanese and Portuguese.
From zero to Ai Agent: How the industry actions were built in advance
In recognition that institutions need faster revenues on artificial intelligence investments, Saleforce has been built 200 industrial procedures formed in advance – With more than 100 additions this summer alone. This ranges from the patient’s schedule in the field of health care to generate advertising proposals in the media, designed to help companies publish artificial intelligence agents quickly instead of building from zero point.
The results show the ripening of the platform. Besides the average deviation of 1-800kountant by 70 % during the tax season, Govindarajan was martyred with other production publishing operations: “OpenTable 73 % of all the restaurant web inquiries that are dealt with by agents”, and “Grupo Falabella, which is a Colombian customer service process using WhatsApp, has received, and got a background of 71 % In the movement of the phone in only three weeks. ”
The company also provided more flexible prices, including unlimited use licenses for employees ’agents and prices for each business that holds together with the actual AI work instead of the simple conversation size.
The new digital workforce: what the AI’s adoption for commercial operations means
Since the institutions are increasingly seen artificial intelligence agents as digital employees instead of simple automation tools, the risks for publishing properly were not higher than ever. Companies that have succeeded in expanding the scope of artificial intelligence clients will get great competitive advantages, while companies that are struggling with governance and risk of operational disorders.
“New roles appear for people who run a fleet of agents.”. “Information manager may ask,” I have seven agents working in my institution, what is happening widely? “But someone runs a specific marketing agent with a different lens from the same problem.”
Looking forward, LerhaUPT position is the current moment as a transformative: “You have a personal computer, then the Internet, and now it’s multi -agent.” He described the development from one publication to what he calls “a multi -agent revolution and the ability to connect agents together to do new types of work very complex.”
For Pepsico, the shift exceeds efficiency gains. “Artificial intelligence and technology reshape the institution’s operations in unimaginable ways,” Kanyura said. “The work we do with Agentforce is one of the broader transformation of Pepsico as a connected company, which paves the way for a more flexible and adapted work future.”
It stresses the competitive scene as the main technology companies are racing to create the platforms of the artificial intelligence agent. When asked about the competition from Microsoft, Google and Amazon, Govindarajan emphasized the advantages of Salesforce integration: “We are able to track the entire work cycle within the institution’s environmental system,” he said. “We can determine the flows and interactions in the institution, and we were open and expanded in bringing your data and actions effectively.”
the Agentforce 3 The platform is generally available on Monday, with many features, including hosting Human models Full Command center It is put forward until August. But it may be the most amazing mark on the Foundation’s willingness to technology in the list of features – it is with confidence from companies like Pepsico to bet on its digital transformation on artificial intelligence agents that they can finally see, measure and control.
https://venturebeat.com/wp-content/uploads/2025/06/nuneybits_Vector_art_of_AI_agent_handshake_f856c181-d118-4cfb-a49c-5969b31bb862.webp?w=1024?w=1200&strip=all
Source link