Slack AI is now available for enterprise customers, though pricing details for the generative AI (genAI) features have not yet been disclosed.

Slack announced plans for its AI assistant app last year, focusing on three areas:

• AI powered search.This provides personalized answers to questions based on an organization's knowledge base. Slack AI helps users locate subject matter experts, or find information on anything from work projects to understanding unfamiliar acronyms.
• Channel recaps.This highlights key discussion points for a Slack user after a period away from the app, or for those who have recently joined a channel.
• Thread summaries. This feature recaps faster-moving discussions, provides thread summaries, and offers an overview of long conversations, with links to sources in each summary that enable users to check information where necessary.

These features are now available to customers on Slack's Enterprise Grid payment plan, Slack said Wednesday. Slack AI will also roll out to a wider range of customers "soon," a Slack spokesperson said, though no date was announced.

Slack is one of many digital work app vendors that have added genAI to their products in recent months, with rivals such as Microsoft and Google starting to roll out AI assistance based on large language models (LLMs) during 2023.

A key strength of Slack's AI tools centers on access to the large volumes of conversation data held in the platform, said Jackie Rocca, vice president of product at Slack. Early customer trials of Slack AI resulted in an average saving of 97 minutes each week, she said.

"Arguably the most valuable knowledge corpus an organization has lives within Slack," said Rocca. "It is built off of years of employees' expertise, learning, and insights. With Slack AI, we help employees unlock this intelligence - helping them make better, smarter, and faster decisions at work."

Amid significant industry buzz relating to genAI, businesses are keen to get their hands on the technology. Irwin Lazar, president and principal analyst at Metrigy, pointed to "a great deal of interest in generative AI virtual assistants [or] copilots for collaboration."

A recent Metrigy survey of around 400 companies found that 86% would purchase genAI features for at least some employees, while 80% would deploy to all staff if the virtual assistants were free.

Slack is well-placed to benefit from such business interest, said Lazar. "Slack is in a unique position, as for many of its customers it is the hub for ongoing communication and collaboration," he said. "Slack's ability to use genAI to surface conversation trends and summarize chats gives them a unique benefit versus those that are only able to capture and summarize conversations within meetings.

"I would expect to see substantial adoption by Slack customers."

No word on pricing

For now, it's unclear how much Slack plans to charge customers.

Slack AI will initially be available as a paid add-on for Enterprise Grid plans in US and UK English. A Slack spokesperson said enterprise customers will need to contact the vendor's sales team to discuss pricing for access. Slack did not provide any more details about pricing or usage limitations. Additional pricing information will be available "once Slack AI rolls out to non-invoice plans," a spokesperson said.

Lazar said it's likely Slack is waiting to see what kind of demand the feature will have before setting a price.

Digital work app vendors have taken a variety of approaches to pricing genAI features in their software, with some including them for free in paid accounts and others charging significant monthly payments.

An ongoing challenge is the compute resources required to provide genAI services to customers. Slack may be taking a cautious approach to ensure it doesn't lose money on delivering the feature, said Lazar.

Slack might also want to be careful to align with parent company Salesforce's own AI offerings, he said, potentially bundling both sets of AI features into a single package.  

Early days for genAI tools

As with any genAI tool that relies on LLMs, Slack AI users will inevitably encounter hallucinations from time to time.  

Rocca said Slack has taken steps to reduce the likelihood that incorrect information is generated. This includes citing sources in summary links to allow users to trace information back to specific messages, while an in-built "feedback mechanism" lets users rate each summary and share feedback directly with the Slack team. This should help improve the product over time.

"The great thing about Slack AI is that it is grounded on all of the conversational knowledge accumulated at your organization. It includes the discussions and decisions made by the experts at your organization," said Rocca.

Slack AI also uses a technique called retrieval augmented generation (RAG) to help reduce "susceptibility to hallucinations and increase response accuracy," she said. "This method augments prompts with relevant data at runtime instead of having the LLM rely only on data used during training."

Looking ahead, Slack is working on a new AI capability to let users create "digests" or summaries of information from specific channels they want to track but are less of a priority.

A native integration with Einstein Copilot - Salesforce's AI assistant for its customer relationship management platform - is in the works, too. This will enable sales teams to ask the genAI assistant about customer data in Slack without switching screens.

Alongside the native AI features, Slack partners have been building their own AI functionality in the collaboration app. They include Box, PagerDuty and Notion.