Salesforce has launched a rebuilt version of Slackbot that can search enterprise information, draft documents and carry out tasks for employees, turning Slack’s longtime notification assistant into a broader workplace AI agent. The release is available to Business+ and Enterprise+ customers at no extra charge, according to VentureBeat’s report from a Salesforce briefing.

The launch is a significant attempt to make Slack the operating layer for employees working with AI. Salesforce is positioning Slackbot against Microsoft Copilot in Teams and Microsoft 365, and Google Gemini across Workspace, while also using it to argue that AI will expand the value of its software rather than undermine the company’s core business.

Slackbot moves from reminders to enterprise actions

The previous Slackbot handled relatively simple algorithmic jobs, such as reminders and notifications. The new system was rebuilt around a large language model, an enterprise search engine and connections to external data sources, Salesforce executives told VentureBeat.

In its current form, Slackbot can work across Slack conversations, Salesforce records, Google Drive files and calendar information. It can summarize and compare material, create a Canvas document, and check participants’ availability. The product team said actual meeting booking would arrive a few weeks after launch. Image generation is not available at release, although Salesforce is considering it.

A product demonstration described by VentureBeat showed Slackbot analyzing customer feedback, comparing written information with an uploaded dashboard image, identifying Salesforce accounts that could be suitable for a pilot and assembling the findings in Slack Canvas. The workflow illustrates Salesforce’s larger ambition: an assistant that does not simply answer questions, but moves information toward a decision and prepares the next step.

Salesforce retained the Slackbot name even though the underlying product is substantially different. Parker Harris, Salesforce’s co-founder and Slack’s chief technology officer, described the new version as a “super agent” intended to become an entry point for work across the enterprise.

Claude powers the launch, but Salesforce plans a multi-model approach

The initial version runs on Anthropic’s Claude. Harris said the choice was partly shaped by compliance requirements: Slack’s commercial service operates under FedRAMP Moderate certification, and Anthropic was the only provider Salesforce could use compliantly when development began, according to the report.

That arrangement is expected to broaden. Harris said Slack plans to support additional model providers, including Google’s Gemini, and left open the possibility of using OpenAI models. The strategy reflects Salesforce’s view that foundation models are increasingly interchangeable infrastructure rather than the main source of product differentiation.

Data handling will be central to whether that strategy works for enterprise buyers. Harris said Salesforce does not train models on customer data. Slackbot is also designed to respect each user’s existing permissions, meaning it should retrieve only conversations, channels and connected information that the user is already allowed to access.

Those are executive statements rather than an independent security assessment. Still, permission-aware retrieval and a no-training-on-customer-data policy address two of the most immediate objections to deploying workplace AI: accidental disclosure and the use of confidential material to improve a vendor’s general-purpose models.

Adoption and productivity claims remain Salesforce-reported

Salesforce said it tested the new Slackbot across its approximately 80,000 employees. According to figures cited by VentureBeat, two-thirds of staff tried the product, 80% of those users continued using it regularly and satisfaction reached 96%. Employees reportedly said they saved between two and 20 hours per week.

Those figures have not been independently verified and should be treated as vendor-reported internal results. They also measure an early company deployment, where employees may have unusually high exposure to Salesforce products and strong incentives to experiment with a new internal tool.

Pilot customers offered similarly positive accounts. Beast Industries CIO Luis Madrigal said the implementation required a relatively quick security review because Slackbot followed existing access permissions. One employee at the company estimated a minimum saving of 90 minutes per day. Engine executive Mollie Bodensteiner said Slackbot saved her about 30 minutes daily by reducing context switching.

The accounts provide useful examples of potential value, particularly for employees searching across fragmented conversations and documents. They are not, however, a controlled productivity study. Buyers will need evidence on accuracy, error rates, administrative effort and sustained use across less tightly managed deployments.

Slack enters a crowded workplace AI contest

Salesforce’s competitive argument is based less on owning the best model than on controlling the place where work already happens. Slackbot is embedded inside Slack, where users exchange decisions, files and informal context that may not exist in a structured business application.

Slack Chief Product Officer Rob Seaman told VentureBeat that proximity is a major advantage because users do not have to leave the collaboration environment or configure a separate assistant. Salesforce executives also argue that Slackbot gains useful context from the work already taking place in Slack, rather than starting with a blank prompt window.

That positioning directly challenges Microsoft Copilot, which benefits from its presence across Teams, Outlook and Microsoft 365, and Google Gemini, which is distributed through Workspace. The products are competing for similar moments: locating information, preparing documents, coordinating meetings and turning scattered work into an actionable plan.

Salesforce also wants Slackbot to coordinate with other agents. Slack already hosts tools from Anthropic, OpenAI, Google and Vercel, including a preview of Claude Code for Slack. Harris said Slack could eventually operate as an MCP, or Model Context Protocol, client that connects Slackbot to tools across an organization.

The company is being cautious about the timeline. Harris said the industry remains largely in a single-agent phase and that broader coordination should be developed with customer outcomes in mind rather than presented as a large network of agents working together. That restraint matters: reliable tool use and clear authorization are harder to deliver than a convincing multi-agent demonstration.

What to watch next

The first signal will be rollout quality. Salesforce said the new Slackbot would reach eligible customers by the end of February, with mobile availability completing by March 3, according to VentureBeat. Buyers should watch whether the feature set arrives consistently across desktop and mobile and whether permissions behave as expected in complex organizations.

Model choice is another important indicator. Support for Gemini and potentially OpenAI would give Salesforce flexibility on price, performance and regional compliance, but it could also make behavior less predictable across tasks. Customers will want transparency about which model handles a request and how Salesforce manages changes between providers.

The next major product test is action-taking. Calendar lookup is available, while booking meetings is planned for a later update. More third-party tool calls, CRM integrations beyond Salesforce and support for systems such as HubSpot or Microsoft Dynamics will determine whether Slackbot becomes a genuine work hub or remains primarily a search and drafting assistant.

Pricing also deserves scrutiny. Slackbot is included with eligible plans, but Salesforce’s broader approach to API access and data products could affect the total cost of connecting enterprise information to AI workflows. A nominally free assistant may still create indirect expenses if customers must buy additional Salesforce data services or alter existing data pipelines.

Creati.ai perspective

Slackbot’s launch is important because it tests a practical thesis about enterprise AI: adoption may depend more on context, permissions and workflow integration than on access to a novel model. Salesforce is using Slack’s accumulated workplace data and daily usage patterns to make the assistant immediately useful, while keeping the underlying model layer open enough to add competitors.

The harder question is whether convenience can become dependable automation. Search and summarization are familiar use cases; safely updating systems, booking meetings and coordinating external agents require stronger controls and clearer accountability. Salesforce has made a credible product move, but the market will judge Slackbot on measurable accuracy, permission handling and the cost of turning chat-based answers into real business actions.