Salesforce has launched a rebuilt Slackbot that can search across workplace data, draft documents and carry out selected tasks for employees. The release moves Slackbot from its original role as a lightweight notification and reminder tool toward an AI agent positioned at the center of daily enterprise work.

The product is generally available to customers on Slack’s Business+ and Enterprise+ plans. Its arrival places Salesforce more directly against Microsoft Copilot and Google Gemini, as major software vendors compete to make workplace AI useful inside applications employees already use.

What changed inside Slackbot

The new Slackbot is based on a large language model and a search system that can draw on Slack conversations, Salesforce records, Google Drive files and calendar information, according to Salesforce executives interviewed by VentureBeat AI. It can also create content in Slack Canvas, the company’s collaborative document format.

A product demonstration described a workflow in which Slackbot analyzed customer feedback, examined an uploaded dashboard image, identified Salesforce accounts that could be candidates for an early-access program, assembled the findings in a Canvas and checked stakeholders’ calendar availability. Meeting booking was not available at launch but was expected a few weeks later, according to Slack chief product officer Rob Seaman.

The redesign retains the Slackbot name even though the underlying system is substantially different from the older algorithmic product. That naming decision gives Salesforce a familiar entry point while it tries to establish Slack as a working environment for both employees and AI agents.

Slackbot currently runs on Claude from Anthropic. Slack technology chief Parker Harris said the choice was partly related to compliance: Slack operates with FedRAMP Moderate certification for U.S. federal customers, and Anthropic was described as the only provider able to meet the relevant requirements when development began.

Salesforce said it plans to support additional model providers. Harris specifically cited Google Gemini and left open the possibility of using OpenAI models. He also said Salesforce does not train models on customer data, arguing that incorporating confidential conversations into model training would make it difficult to preserve the original access permissions.

Adoption claims come from Salesforce’s own testing

Salesforce has tested the new Slackbot internally across its workforce. Ryan Gavin, Slack’s chief marketing officer, said two-thirds of Salesforce employees had tried it and that 80% of those users continued using it regularly. The company also reported a 96% satisfaction rate for the feature and said employees estimated time savings ranging from two to 20 hours per week.

Those figures are vendor-reported internal results, not an independent evaluation. Salesforce also said organic sharing helped drive adoption, including an employee-created Canvas containing more than 250 prompts. UX researcher Kate Crotty attributed 73% of internal adoption to social sharing rather than management mandates.

Pilot customers supplied similarly positive accounts. Beast Industries said one employee saved at least 90 minutes per day, while Engine executive Mollie Bodensteiner estimated a saving of about 30 minutes daily from reduced context switching. Other named pilot participants included Slalom, reMarkable, Xero, Mercari and Engine.

These customer comments are useful signals about perceived value, but they do not establish consistent productivity gains across industries. The demonstrations also show a product that can retrieve and synthesize information today, while some more consequential actions—such as booking meetings and broader third-party tool calls—remain under development.

Salesforce is selling context against Microsoft and Google

Salesforce’s argument is that Slackbot has an advantage because it sits close to the conversations, documents and decisions that make up an employee’s work. Slack executives say users do not need to configure a separate knowledge system or move between applications to provide context.

That strategy directly challenges Microsoft Copilot, which is embedded across Teams and Microsoft 365, and Google Gemini, which is integrated into Workspace. Each company is trying to make AI more useful by connecting it to existing workplace data rather than asking customers to adopt an isolated chatbot.

Slack’s distribution may help Salesforce compete, but its data boundary is also a central product question. Slackbot is designed to use information a user is already authorized to view. That permission-aware approach could make enterprise security reviews easier, although the source material provides only customer and executive accounts rather than independent testing of access controls.

Salesforce is also presenting Slackbot as a possible hub for other AI agents. Slack already hosts or is attracting agents from companies including Anthropic, OpenAI, Google and Vercel. Harris said Slack could eventually act as a Model Context Protocol client, allowing Slackbot to connect to tools across the software ecosystem.

The company is cautious about the near-term scope of that ambition. Harris said organizations remain largely in a single-agent phase and warned against demonstrations featuring thousands of agents coordinating at once. For builders, that restraint matters: dependable retrieval, permissions and tool execution may be more valuable than a large but fragile multi-agent architecture.

Cost, access and deployment remain practical questions

Slackbot is included at no additional charge for Business+ and Enterprise+ customers. That pricing could lower the barrier for existing Slack users experimenting with enterprise AI, particularly where procurement teams prefer capabilities bundled into an established platform.

However, the broader Salesforce data-access model may affect the economics of connected workflows. VentureBeat cited warnings from Fivetran CEO George Fraser that changes to Salesforce API pricing could raise costs for companies that replicate Salesforce data into systems such as Snowflake or connect it to external AI services. Salesforce has characterized the pricing changes as standard industry practice.

The tension is important for enterprise buyers. A bundled assistant may be inexpensive at the user interface, while the cost of moving, querying or governing the underlying data can appear elsewhere in the stack. Buyers evaluating Slackbot will need to examine connector limits, API usage, retention, audit controls and the treatment of information outside Salesforce’s own products.

What to watch next

The first signal will be whether Slackbot’s reported internal engagement translates into sustained usage among paying customers. Product teams should also watch for the promised meeting-booking capability, mobile completion scheduled for March 3, and the addition of more model providers beyond Claude.

The next major test is tool breadth. Salesforce has discussed third-party tool calls and an MCP-based architecture, but has not provided specifics about support for competing CRM systems such as HubSpot or Microsoft Dynamics. Those integrations could determine whether Slackbot becomes a genuinely broad enterprise interface or primarily a Salesforce-centered assistant.

Model choice will be another competitive indicator. Adding Gemini or OpenAI could improve cost and performance flexibility, but it would also increase the importance of consistent safety, permissions and response behavior across providers.

Creati.ai perspective

Salesforce’s Slackbot launch is significant less because it introduces another workplace chatbot than because it makes Slack the proposed control surface for enterprise actions. The strongest near-term use cases are grounded in existing conversations and records: finding decisions, summarizing scattered evidence, preparing documents and reducing routine context switching.

The harder problem is execution. For Slackbot to become an enterprise “super agent,” Salesforce must prove that permissions remain intact, tool calls are dependable and the total cost stays predictable as customers connect more systems. The company’s vendor-reported adoption results are encouraging, but broader customer evidence and independent validation will matter more than internal enthusiasm as the competition with Microsoft Copilot and Google Gemini accelerates.