Salesforce Inc. recently announced the launch of its new artificial intelligence autonomous agent platform Agentforce that can handle tasks for employees, and for enterprises to do this in a useful way, the AI agents need to connect to external tools and partners.
Today at its annual Dreamforce conference in San Francisco, the company announced the launch of the Agentforce Partner Network, which will allow users to quickly get access to connections to external data sources, specialized third-party actions and unique AI skills and deploy partner-built agents. These agents can plan and take action on their behalf across multiple systems and channels, even outside Salesforce.
“Customers want AI that they can deploy with trust that drives value,” said Clara Shih, chief executive officer of AI at Salesforce. “So far, they haven’t seen ROI from their investment in competitor products, that’s because those products are disconnected from the data and metadata and cannot take action. Our own customer testing shows that Agentforce drives higher accuracy and higher time to value than building on an LLM like OpenAI and Azure.”
Among the new partners, IBM Corp. will launch prebuilt agents for regulated industries. The IBM banking agent will speed up loan approvals by being able to use credit checks, comply with local tax laws and generate onboarding documents. Google LLC has joined the Agentforce Partner Network to allow agents access to its ecosystem to generate Google Docs, search Gmail or trigger from Google calendar events.
“In summary, we are leaving the era of disconnected copilots behind and moving into a future of a broad open connected network of interoperable third-party systems and agent forces,” said Shih.
The same partner network will also enable Agentforce agents to connect through Salesforce’s Zero Copy Partner Network, which permits businesses high data connectivity from partners into Data Cloud. The “zero copy” means that none of the data that moves across the locations of the network is moved or copied between platforms, saving on storage costs. Partners in that network include Amazon Web Services Inc., Databricks Inc., IBM, Microsoft Corp.’s Azure, Google BigQuery and Snowflake Inc.
Data Cloud updates for better AI agents
In order to keep up with the soaring data needs to feed the burgeoning demand for AI agents that Salesforce is already seeing, the company also announced multiple feature updates to its Data Cloud hyperscale data platform today.
“Data Cloud is the heartbeat of the Salesforce platform. With it our customers can build a unified view of every customer of theirs so that they can put their data to work with full fluidity,” Rahul Auradkar, executive vice president and general manager for Data Cloud at Salesforce. “Data Cloud brings humans together with agents, which are powered by AI, data and agents to deliver customer success. All of this is done while enforcing enterprise data governance and security.”
Auradkar added that Data Cloud processed “multiple quadrillion records” during the last quarter and it is only continuing to grow.
Data Cloud is adding more unstructured data types, including audio and video to its vector database capabilities. This will allow customers to easily scan through customer calls, meetings and more. It will also make it possible to create multimodal AI-powered apps that can use audio and video as part of their workflow.
This is important because about 90% of the data produced by a business is unstructured and therefore “dark data” when it comes to search and AI. In any given amount of that is also audio and video, meaning that if it cannot be stored and accessible to AI agents then it’s out of reach. By adding this to Data Cloud, suddenly it becomes a valuable resource.
Salesforce also announced that Data Cloud is getting 50 new data connectors allowing real-time data insights from different types of data streams. This adds to the already existing 150 prebuilt connector pipelines that Data Cloud already has available. Connectors could be fairly generic as something that connects to a hyperscaler such as AWS, Google or Azure, such as a Kinesis data stream, Slack communications or Google documents, or something as specific as a particular financial services implementation.
Data Cloud will also receive subsecond real-time data ingestion capability, which will allow data models to take immediate action when something changes. This means customers can create apps that can handle rapidly changing data streams for instantly critical information within fractions of a second with extremely low latency. Use cases include internet of things devices, telemetry and other situations where having immediately up-to-date information is important.