OpenAI and Salesforce announced today that ChatGPT is coming to Slack. The solution is still in beta with limited access, but the announcement states:
Built by OpenAI on the Slack platform, the app integrates ChatGPT’s powerful AI technology to deliver instant conversation summaries, research tools, and writing assistance directly in Slack…
With the ChatGPT app for Slack, customers can:
Get up to speed faster on channels or threads: with AI-powered conversation summaries help users quickly catch up on what’s happening.
Instantly find answers on any project or topic: with AI-powered research tools, users can learn and build expertise faster right from Slack — whether they're researching best practices, prospecting a new account, and more.
Draft messages in seconds to communicate with customers and teams: with AI-powered writing assistance, users can spend less time crafting replies, status updates, and meeting notes — and more time putting the plan in action.
Summarization
I had the chance yesterday to see a closed demo for the media, and it is immediately obvious how this could add value to Slack users. The thread and channel summary features are particularly intriguing. If you have used Slack (or Discord or Microsoft Teams) you will have encountered the problem of entering a long thread or being invited to a channel with dozens or hundreds of messages and trying to get up to speed. It can be a tedious process.
The ChatGPT summary of these threads can definitely save time. One of the best features of ChatGPT is summarization. However, there would be no practical way to cut and paste a long Slack conversation, so having this feature baked into the messaging platform actually opens up this use case as a possibility.
What is less clear is the context window. The context window is the character limit that the model can ingest and take into account for its task. In this case, if you have more than about 2,000 words in a Slack chat, then it will likely default to the last part of the conversation. This means that summarizing an entire channel that has been active may not be feasible. You could miss some important context. With that said, most threads will be shorter than this context window. That may turn out to be the best use of the feature.
Research and Writing Assistance
The research assistance is similar to another feature that was demonstrated by Salesforce for its Einstein GPT for Slack. If you need some quick information about a topic where ChatGPT might be useful, you can just do it within Slack. There is no need to go to another application. This could have the added benefit of more easily sharing ChatGPT results within a channel or DM group. Slack will like this as it keeps users in the application even more.
There is, however, no mention of accounting for hallucinations or lack of a real-time internet connection. That makes it unclear how useful that feature will be for day-to-day business users.
Writing assistance is another feature that is easily accessible by clicking on another tab and accessing another application. It may be helpful and save time cutting and pasting from ChatGPT or Jasper AI. The larger benefit, I suspect, would be for real-time collaboration by team members, wordsmithing a message, or marketing copy. Bringing ChatGPT interactions into a shared space more easily could be an interesting change from the individual interactions we see today.
Slack Over Teams
There is another point to note here. Microsoft is OpenAI’s largest investor, and it appears the company has effective control over the company. Microsoft has already said GPT-3 technology will come to Teams, but where is the announcement? Its biggest rival in the enterprise team messaging space is Slack, and OpenAI is now pushing that integration.
In fact, OpenAI built this feature because it uses Slack internally. And it doesn’t sound like a company that is anticipating a Teams implementation anytime soon.
“I love Slack Connect so much. We were very early adopters, and, alongside Huddles, it’s been the lifeblood of our customer communications. I think it’s the coolest workspace communication advancement ever. Injecting our own technology into this product has supercharged our ability to connect with and delight our customers. Life without it with our customers is unimaginable.” –
Zack Kass, Head of GTM for OpenAI
With that said, Microsoft produced a video way back in January that walked users through using Power Automate to create a workflow that integrates ChatGPT into a Teams channel.
This is not quite as easy as simply adding a ChatGPT Slack app to your workspace, but it does appear to work and is available today. For Slack GPT, you will have to wait a bit until OpenAI lets you into the beta. However, if the waiting period for ChatGPT Plus is any indicator, you should expect to have access very soon.
Granted, there are some questions that have yet to be answered. Who will pay for the use of ChatGPT in Slack? The Power Automate example with teams expects you to add a personal ChatGPT account (granted a free account), but you would think that a corporate account option will soon be available. And for Slack, will it be available to free users or only paid users? I suspect the latter, but maybe it will be available to all during the OpenAI run beta period.
By the way, this feature is being run by OpenAI. That is where you sign up for the beta waitlist. It is heavily supported by Slack, but it is not their product as it is positioned today.
Features Over Product
As we’ve said here in the past, there will be new products and even product categories built on top of large language models (LLM). However, I am looking closely at the adoption of LLM-based features in existing products. A new technology can become pervasively adopted far faster if it becomes embedded in the tools we already use as opposed to requiring everyone to embrace new tools and habits.
LLMs are intriguing because it is both the foundation of new products and viewed as an enhancement of existing products. That offers the technology many chances to succeed. This is particularly true when it directly addresses a problem in an existing product, like the need for summarization when joining a new conversation in Slack.