New Data Cloud features to boost Salesforce’s AI agents

Salesforce added new features to its Data Cloud to help enterprises analyze data from across their divisions and also boost the company’s new autonomous AI agents released under the name Agentforce, the company announced at the ongoing annual Dreamforce conference.

The CRM software provider terms the Data Cloud as a customer data platform, which is essentially its cloud-based software to help enterprises combine data from multiple sources and provide actionable intelligence across functions, such as sales, service, and marketing.  

These new features include support for unstructured data in audio and video format, 50 new connectors, and a low latency data layer to help the new AI agents respond faster, said Rahul Auradkar, general manager of unified data services and Einstein at Salesforce, during a press briefing.

Support for unstructured audio and video data

The addition of support for unstructured audio and video content will help enterprises extract and analyze customer data from previously inaccessible audio and video sources like customer calls, training sessions, product demos, feedback surveys, voicemails, and webinars, Auradkar explained.

Further, the general manager said that support for audio and video content can help enrich customer profiles, reveal deeper intent behind customer preferences, and behavioral patterns, and improve the accuracy of autonomous agents deployed with the help of Agentforce.

Explaining how support for audio and video content can help the accuracy of autonomous AI agents, IDC’s research manager, Hayley Sutherland, said that support for vectorization of video and audio content means that more content can be used for retrieval augmented generation (RAG), making that content accessible to large language models (LLMs) underpinning the agents to augment responses and generated content.

Most enterprise data traditionally, according to Nucleus Research CEO Ian Campbell, stores structured in tables or spreadsheets, but a large amount of valuable information exists in unstructured formats like video, audio, and text.

“With vector processing, Salesforce enables enterprises to convert unstructured data into a structured form that AI can process more effectively,” Campbell explained.

The feature, according to the company, is currently in pilot and is expected to be made generally available next month.

Salesforce has also added at least 50 new connectors to the Data Cloud in order to help enterprises bring in more data to the cloud-based software.

“Salesforce’s goal with more connectors is to create a unified data ecosystem that makes it easier for businesses to integrate diverse data sources and subsequently improve personalization and decision-making,” said Arnal Dayaratna, research vice president at IDC.

The diverse data that will flow in through the new connectors, according to Sutherland, will further help the new AI agents as they can be used for RAG.

The total number of connectors currently stands at 200, the company said.

Support for sub-second data layer

Another key update to Data Cloud includes support for a sub-second data layer, which also has been made generally available.

The sub-second data layer, according to Auradkar, helps enterprises ingest, unify, analyze, and act upon data in real time across Salesforce.

The layer powers Einstein Personalization, which itself is built on Data Cloud, by helping provide real-time AI recommendations, analytics, and automation to enable faster decision-making and instant personalization across customer touchpoints, Auradkar explained.

Tying the update to Agentforce, IDC’s Dayaratna said that the sub-second data layer is crucial for the new autonomous AI agents that need low latency to orchestrate responses from multiple models in real time.

“This ensures faster, more accurate customer interactions. It helps enterprises deliver timely responses that contribute to augmented customer and end user satisfaction,” Dayaratna added.

The other advantage of ensuring low latency responses in the agents is to showcase how Salesforce is providing generative AI-based offerings that are viable for real-world use cases.

On the other hand, The Futurum Group’s research director Keith Kirkpatrick pointed out that the sub-second data layer helps developers deploy AI agents that can handle a variety of tasks and workflows in order to achieve “friction-free” automation, especially due to the low latency.

Governance and security updates

Salesforce’s Data Cloud has also been updated with new governance and security features to prevent the exposure of data to unauthorized parties while using AI.

These features include AI tagging and classification, policy-based governance, customer-managed keys, and private connect for Data Cloud.

While AI tagging and classification automates the labeling and organization of unstructured data as per business policies ensuring authorized access to data, policy-based governance as a feature aims to provide governance at scale.

It does so by creating granular security policies that grant appropriate access to user groups based on tags, metadata, and user attributes, the company said, adding that both features will be in beta starting in November.

Customer-managed keys, which are already generally available, according to the company, allow enterprises to manage their own encryption keys, ensuring data stays encrypted and secure regardless of how it is used.

The private connect for Data Cloud feature has also entered general availability, the company said, adding that it enables enterprises to safely share and integrate their data between Data Cloud and public cloud environments by establishing secure, direct network connections between them.

Data Cloud One, hybrid search, and other updates

As part of its updates to the Data Cloud, Salesforce is releasing what it calls Data Cloud One — an offering that will help enterprises connect multiple Salesforce accounts that may be siloed across different departments, regions, or business units.

Data Cloud One, according to Auradkar, extends all of Data Cloud’s functionalities across all Salesforce instances using a no-code, point-and-click setup.

The rationale here is to ease enterprises’ efforts to run multiple Salesforce accounts by creating a single Data Cloud instance, which could be used to generate more business insights.

Data Cloud One is expected to be generally available next month.

The CRM software provider has also updated Data Cloud with a new hybrid search capability to help enterprises find relevant information faster, such as content specific to an industry vertical.

Hybrid search, according to the company, combines vector search with the exact match capabilities of keyword search. The feature is expected to be made generally available in November.

Other updates include the addition of Tableau Semantics and a new data community.

While Tableau Semantics will help enterprises organize data based on its meaning and relationships, creating a standardized semantic model that allows anyone to understand, use, and act upon data consistently, the data community, an online platform, helps connect IT and business leaders, developers, and Data Cloud enthusiasts.

“The data community provides space to learn, share insights, and stay informed on the latest best practices, trends, and tools to maximize the value of data,” the company said. Tableau Semantics is currently in the pilot phase and is expected to be made available in February next year.