How we designed Slack AI

 

Editor’s note: Jackie Rocca is a VP of product at Slack and leads the development of Slack AI. We take our commitment to protecting customer data seriously. Learn how we built Slack to be secure and private.

We’re on the frontier of a productivity revolution.

Over the past year, several advancements in AI have reached mainstream adoption, transforming the tech landscape. One notable trend is the rise of large language models (LLMs) and their user interfaces, such as OpenAI’s ChatGPT and Google’s Bard, which have game-changing abilities to understand and generate human language. This has birthed a plethora of applications from chatbots to automated content creation.

With the meteoric rise of these AI tools, there’s a lot of hype and inflated expectations around what AI can actually do. Our product teams have kept boots on the ground and viewed things through a slightly different lens: how can we apply these new technologies to unlock possibilities in our platform and help customers achieve even more productivity through Slack today?

Our guiding design principles

As we began the design process for Slack AI, two priorities became clear. Customers have developed so much valuable context in Slack but are just scratching the surface of what they can learn from their organisation’s collective knowledge. We want to help people find that information immediately. And secondly, we know that work moves fast – so we want to help people quickly see what’s most important.

While we’re extremely excited about the potential of AI, we’re also cognisant that it can raise important questions and challenges that need to be addressed thoughtfully and clearly. It was vital to us to keep three things in mind:

1. Safety: Is it secure? AI is irrelevant if we can’t keep our customers’ data safe.
2. Relevance: Is it useful? How does it add helpful context that maximises the Slack experience?
3. Usability: Have we made our AI easy and delightful to use?

Based on these priorities, we developed a set of principles that guided our product development.

Don’t force humans to be prompt engineers. Our primary goal was to provide immediate value – without forcing you to learn brand-new skills in order to take advantage of these tools. Slack AI will help when you need it, guiding you with suggestions instead of leaving you staring at open-ended text boxes.

• Make it clear when AI is involved. Flagging AI-generated content is crucial. As we venture into this new landscape, we’ll be clear when a piece of information is generated by AI, encouraging a culture of trust and understanding about the AI’s role in assisting you.
• Let users interact with AI privately. You can interact with AI in a one-on-one space by default, and it’s always clear when only you can see something. And when you are ready, share your AI-assisted insights with your teammates, building a collaborative experience one step at a time.
• Stay flexible. While our AI is designed to surface when you need and recede when you don’t, we know you may want to tweak your results beyond the default outputs. Want a summary for a specific date range? Prefer concise responses? We’ve got you covered.
• Offer useful context. We’re dedicated to creating relevant experiences and offering you summaries of missed messages auto-magically. Our goal is to delight you with AI assistance that is intuitive and accessible – without you having to command it explicitly.

During a time when so many companies are trying to squeeze AI into all corners of their product, we’re taking a different approach. We’re honing our debut features to perfection in a way that will powerfully enhance users’ day-to-day experience. You’ll look back and wonder how you ever worked without it.

Transparency builds trust

At Slack and Salesforce, our parent company, trust is our most important value. In fact, the top concern about AI we hear from customers is around security. That’s why we adhere to a set of product values that guide our decision making, all focused on maintaining your trust above all else.

Your data is not our product

In a world where some companies use customer data as an asset to hawk, we prioritise user safety and data privacy above all. Our new AI product features are designed inside Slack’s secure infrastructure from the ground up to ensure that your data remains yours – and only yours. We don’t sell or rent your data. We’ve built robust mechanisms to improve Slack for you and for all Slack customers while protecting the privacy and security of your information. Because we firmly believe that your trust is not for sale.

You control access to your data

We’re diligent about preventing any proprietary information from mixing between customers. Slack’s entire ecosystem adheres to a high standard of security and compliance. Our Enterprise Key Management (EKM) empowers customers to control their encryption keys – even giving you the autonomy to revoke access whenever you choose. You don’t just have to take our word for it; our compliance with recognised standards such as HIPAA, FINRA, FedRAMP and SOC 2, all of which have been audited by third-party industry experts.

We give you the means to verify the results

We understand that to have trust in a piece of technology, you need to be able to count on its reliability at all times. Slack AI won’t replace your employees’ judgment. Instead, we’re designing our AI product features to be transparent, so you can dig into what’s important and independently verify the results.

What’s coming next from Slack

This winter, we’re starting the pilot for Slack AI. With every message sent, canvas shared and clip recorded, Slack has become a treasure trove of knowledge across an organisation. To advance the full potential of that collective knowledge, we’re starting with some basic but mighty features that promise to unlock massive daily value for users:

1. Channel recaps. Users can generate key highlights on any channel, helping them cut straight to what’s most important. Imagine being able to catch up on multiple conversations you missed during PTO or instantly extract key themes from a project channel you’ve just joined. With new insights quickly and easily accessible, teams can make informed, customer-centric decisions.
2. Thread summaries. Users can get up to speed on any conversation in one click. Say the customer support team is swarming on a case and adds subject-matter experts to the long thread. With thread summaries, the experts can quickly onboard and jump in to help resolve the issue.
3. Intelligent search results. Users can ask a question and get a clear, concise answer based on relevant Slack messages. For example, when learning about a new project, search answers help users quickly get a grasp on the project’s goal, the key stakeholders involved and more – all packaged in a visually separate design next to the traditional search results users are accustomed to seeing.

And we have a lot more in store, with plans to build out even more intelligent features to help you do your best work. Want to be the first to know when these AI tools are available for your workspace? Explore them in detail in our new e-book, and sign up for the Slack AI pilot today.