Salesforce announces Slack GPT as the platform pushes further into AI
A whole suite of AI features have been announced as part of Slack GPT and the platform’s push to automate workflows as much as possible.
Slack is the latest tech platform to push into the AI space, formally announcing Slack GPT and a suite of other AI integrations today in New York.
According to the company, Slack GPT will enable users to deploy generative AI app integrations, different language models, and tap into secure customer data insights from Saleforce’s Customer 360 and Data Cloud offerings.
We first heard plans of introducing Chat GPT to Slack during Salesforce’s TrailblazerDX conference back in March. It also announced its own GPT generative AI pilot at the time, EinsteinGPT.
Both apps will now be available in Slack with an aim to build AI natively into the platform for business use.
Slack will let a company use whichever generative AI it wants
During a call with SmartCompany, Slack’s VP of Product, Ali Rayl, said that the aim of these integrations was to help users save time.
“This is the very first piece of a much larger vision of how generative AI is going to add a brand new layer of productivity to our platform,” Rayl said.
AI-powered conversation summaries and writing assistance will be available directly in Slack just by typing a prompt — a feature reminiscent of Microsoft’s Copilot that was announced back in March.
Slack has also promised no-code workflows that embed AI actions with simple prompts at each step. They can also securely integrate a large language model (LLM) from companies like OpenAI and Anthropic. The latter recently received investment from Salesforce’s newly minted $369 million AI-focused fund.
Companies are also welcome to use the LLM of their choice, with Slack encouraging secure third-party apps and integrations.
In the future, Slack also aims to bring conversation summaries to Huddles and the recently-launched Canvas feature.
Privacy and training a concern for many companies
One of the biggest concerns with generative AI has been privacy and what data is being used to train them. Rayl confirmed with SmartCompany that both Anthropic’s Claude app and the version of ChatGPT available on Slack will not be used for AI training. Nor will the data be stored.
According to Rayl, this is the concern of most companies.
“It’s really funny to talk to customers about this market because, on one hand, they say ‘we are so excited. We want this innovation, we want this power. Let’s go.’
“And we’re like, ‘let’s talk about your data’ and they’re like, ‘not my data.’
“One of the most important roles that we have to play here is helping companies connect their data in a way that they trust.”
This is why Slack is allowing companies to use any LLM they want on the platform. While some businesses just want an easily accessible public LLM, others prefer to spend millions to build their own proprietary models.
“We don’t think there’s the one size fits all solution here in any way. We think that companies need to have a choice in which LLM they’re going to use for a lot of different reasons — data and privacy and security is one of them.”
Keeping users on Slack
The Summaries features, in particular, got me wondering if in the future Slack hoped to expand the idea of conversation summaries to other work platforms. For example, if you returned from holiday and wanted a summary of progress on account that could span across many applications — Slack, email, Zoom calls.
“I sure hope so,” Rayl said.
Bringing it back to Slack specifically, Rayl says that what’s interesting about working with LLMs is that they understand Slack and natural language “but the structure of the data is weird.”
Let’s say you needed information about a certain account, it might be spread across multiple channels.
“How do we get an LLM to think about this? The way I’d tackle that is probably more from search… for example ‘search for what’s happened in the last week about this’” Rayl said.
“One of the product challenges for us is figuring how we get these sort of AI-assisted moments to you when you need them, and surface those in the right way.”
Still, with so much integration with third parties and automation going into the future of the platform, it sounded like Slack is aiming to keep users on the platform as much as possible during their workdays.
“One thing we talk about a lot with our platform is the concept of thin work and thick work. A lot of the tools that we’re using at work are thick work — big and robust and they have so many buttons,” Rayl said.
Rayl gives the example of filing an expense report and the number of actions you need to take between logging in, going between tabs, and endless clicking.
“We look at the things that people do in the course of their day-to-day work and say ‘what are the things that we can automate with one or two buttons?’ so you don’t have to go into the other web page, you don’t have to do all the clicking.
“How can we reduce the amount of time that humans spend clicking around an application that is unnecessary if they can do that in Slack instead?”
