By Phil Wainewright

February 14, 2024

In the week that it celebrates ten years since it first launched, collaboration app Slack today enters a new phase of its journey as it begins the production roll-out of Slack AI. The ultimate goal, according to the press release issued today, is to become "the command center for work and the conversational interface for generative AI." But customers will have to pay a toll for that access, as Slack AI is available today only as a paid add-on for Slack Enterprise plans, only in US and UK English at present — and there's a waitlist too.

There are some new capabilities on the roadmap in addition to the previously-announced Slack AI features rolling out today. The known features include AI-powered search and summarization of channels and threads. New features coming soon include the ability to create channel digests that keep users up-to-date with what's going on in channels they interact with less frequently, and native integration with Einstein Copilot, the new conversational AI assistant for Salesforce CRM. Users will be able to query Einstein Copilot for answers directly from the Salesforce CRM system without having to leave Slack.

ISV partners are also unveiling AI functionality in their own Slack apps today. Examples available now include the ability to ask PagerDuty Copilot for help resolving incidents, to automatically summarize Notion documents in link previews, or to use Guru’s enterprise AI search to find answers from enterprise knowledge stores in apps such as Google Drive, Salesforce, SharePoint, Zendesk and Slack itself. In addition, customers who sign up for a new enterprise service from AI search startup Perplexity will soon be able to view insights pushed from its answers engine directly into Slack.

Saving time

Based on an internal analysis of customers' use of Slack AI during the pilot phase, the company found it could help users save around an hour-and-a-half — an average of 97 minutes — every week, by spending less time searching for information or reading through content. Zach Hyman, Co-CEO at early access customer SpotOn, comments:

During the pilot program, we experienced firsthand the major productivity gains that Slack AI could drive for our business. Slack AI has helped speed up our employees' work exponentially.

Slack has updated its statement on security and compliance (PDF) to coincide with the roll-out of Slack AI. The company runs the service on its own LLMs to ensure the integrity of customer data and to tailor its output to Slack use cases. In addition, any answers include source attribution so that users can verify the information or dig deeper. As Ali Rayl, then SVP of Product Management at Slack, told us back in September:

We want to make sure that everything that we show you, you can have confidence in because you can tie it to where it came from and who it came from.

My take

As I wrote last month, the advent of generative AI poses both a threat and an opportunity for the likes of Slack and other digital teamwork vendors:

Just a couple years ago ... these messaging, content management and workflow automation players stood on the brink of becoming the conversational interface through which all users connect to enterprise information and functionality. Now generative AI offers a new, even more flexible and accommodating means of accessing digital resources, relegating teamwork apps to a subsidiary role where they have to work even harder to prove their worth.

Slack is clearly determined to maintain its role as the conversational gateway to digital resources, co-opting generative AI to help it in that task. At diginomica, however, we're not keen on vendors charging a premium for generative AI functionality. In Slack's case this seems even more unwise — erecting a tollgate isn't the best way of maximizing your appeal.

Meanwhile several of the risks that I mentioned in my analysis last month remain, even though Slack seems to have the basics right when it comes to security and data privacy. The inclusion of source material so that users can vet the accuracy and relevance of the answers they're served is helpful, but it's still not a sufficient guard against out-of-date or incorrect information being surfaced. Without active curation of the content being searched, such as the additional human verification that a service like Guru's provides, there's too much risk that users will act on the wrong information, particularly in use cases such as product and support information, compliance processes or other contexts where being wrong can have serious consequences.

Enterprises face a difficult task as they navigate their way through all the clamor of generative AI offerings to settle on a viable strategy for their own organization. At diginomica we still believe that it's important to aim towards building a shared fabric that connects people and work across the enterprise — what we define as a Collaborative Canvas to channel digital teamwork. That job has become more complicated with the addition of generative AI into the mix.

Not only do enterprises have to evaluate vendors on how well they incorporate this new technology into their offerings. They also have to understand how it interacts with all of the raw information stored and generated across message threads, online meetings, recorded audio and video messages, document stores, illustrations and diagrams, slide decks, and applications. At the same time, the job of curating all of the enterprise content, conversations and workflows served up to generative AI becomes much more crucial. The ongoing hype about the potential productivity gains shouldn't be allowed to obscure the importance of addressing these difficult questions.