Making Slack Less Overwhelming
This project got it’s start when browsing product review site G2.com’s Slack page. Lot’s of people were frustrated with the fact that over the past few years Slack has gotten really overwhelming.
Distracted and overstimulated employees can't do their work effectively. Failing to address this issue could cause businesses to sour on the platform altogether, leading to lost subscribers and revenue.
Slack Sr. Engineer (Consulted)
ML Engineer (Consulted)
Otter.AI
Adobe Suite
Google Workspace
Competitive Analysis
SWOT Analysis
Heuristic Evaluation
Problem Statement
User Interviews (x5 Users)
Affinity Map
Story Maps
Competitive Solution Analysis
Sketches
Wireframes
Flow Map
Usability Tests (x5 Users)
Final UI Screens
High Fidelity Prototype
Get caught up to speed on your channel activity up to x100 times faster. Goodbye information overwhelm!
.gif)
Searching for a problem to solve
One of the things I’ve always been fascinated by is how changing communication technologies shape culture. Back in the day, the invention of the printing press changed the world as did the radio and television in later centuries. What drew me to this problem in the first place was the potential magnitude of it as well as the opportunity to address the issue of healthy and productive use of technology in our everyday lives.
Slack has become a distraction machine
Avg. amount of time knowledge workers spend looking for internal info or tracking down people to help with specific tasks.
Avg. amount of messages sent by
Slack users at large companies. Power users are sending out more than x1000 per day.
Getting the Lay of the Business Landscape & Figuring Out Where My Knowledge Blind Spots Were
Again and again the theme that kept popping up in user comments was that they found Slack to be both really overwhelming as well as distracting them from their actual work with all of the messages, channels and threads there are to keep track of.
There was a business opportunity here in the fact that distracted and overwhelmed employees aren’t productive workers. There was a real threat that users could abandon the platform altogether, leading to lost revenue.
Next steps would be conducting a SWOT analysis, competitor analysis, industry trends, and heuristic evaluation. I specifically chose these methods of discovery because they were a quick way to get needed context on the business side of the problem as well as potential areas for improvement with the product as it currently exists.
Discovery Methods
“It quickly gets overwhelming within a busy workplace. Setting up channels for specific projects/topics or direct messages (DM's) for specific conversations helps organize concerns, but that can get unwieldy. They added favoriting channels and people, in essence you see those new messages first, but I find that no matter what I do, it becomes a chore to keep up with everything.”
- Jesse K. / comment G2.com
“Slack's fast-paced chat environment, while efficient, can sometimes lead to information overload. The constant influx of messages and notifications may become overwhelming for some users, requiring careful management of channels and reports.”
- Peter C. / comment G2.com
ChatGPT is announced for Slack. But what does that mean?
Great! I felt like I was on to a solid lead. Then the mixed emotions hit. An article pops up in one of my feeds mentioning that Chat GPT is being planned for integration with Slack and might be looking to address this very problem.
What would this look like though? My thinking moving forward was this. AI is worth taking a look at, but don’t assume its a solution to a problem I haven’t fully explored yet.
I wrapped up discovery with this guiding statement.
How might we help users of Slack feel more empowered and less overwhelmed when trying to retain important or actionable information?
Activities & Outputs
Lost in the woods: Is this problem too systemic?
To get a better handle on the problem, I interviewed x5 Slack users. The overall goal was to get specific and expand on what I'd learned in discovery and also uncover any other work habits, processes or coping mechanisms people were using.
In choosing my interviewees I wanted to get a broad range of people with different job title types. This would ensure that whatever solution came down the pipe later on it was catered to the collective needs of an entire organization.
I wound up interviewing the following.
- A start-up upper management exec
- A product manager at medium sized software company
- A freelance machine learning engineer
- A senior engineer at a medium sized software company
- A visual designer at a small agency
Here's what users had to say.
“From like a senior leadership level, I would have just liked kind of daily reports on each channel. I guess more like issues reporting or, anything in the channel that like directly related to the project, I wish there was a better way for Slack to kind of report back to me at the end of the day... things that I needed to know because of everything that was happening within Slack. I don't know if there's a great way to do it which is probably why they haven't come up with it yet."
- Erin T. / Interviewee
"l usually can't use Slack at the same time I’m actually doing my
work. So I'll just have to just silence it. Or don't even silence it. I just,
check in on Slack and I do my work, and I check in on Slack.
- Ben E. / Interviewee
"So you can always search your stuff but then it's like, oh, that
conversation happened recently but I can't remember what
keyword I should use to go look for it basically. Yeah, and it's hard to find."
- Nick R. / Interviewee
Looking for patterns in the interview data
Interviews wrapped up with an affinity map to gather all the data and pull together some core insights in one cohesive place. This would serve as an invaluable reference to keep coming back to to make sure whatever I'd be designing truly reflected what I'd learned from my users.
Well I learned a lot of specifics, but I still don't know which direction to go.
So here is where I was at, what I had learned and what I needed to do next.
-There is still no one obvious direction from here. I needed some outside expert help to break the stalemate.
- AI is something users are experimenting with. This ties into a huge business opportunity to keep up with the competition (like MS Teams).
- This project might go off the rails if the guiding problem statement doesn’t shift gears a bit. It had a couple of faulty assumptions built in.
How might we help users of Slack feel more empowered and less overwhelmed when trying to retain important or actionable information?
How might we help Slack users feel more empowered and less overwhelmed in their ability to interact with important information as part of a team?
Focus on the team aspect: This problem needs to be approached both from an individual AND from a team standpoint. There are many people involved that contribute to the issue, not just a single isolated individual.
Its not really about “saving” information: The previous statement had a wrong assumption that this problem was about “saving” information. Interviews didn’t support this.
Activities and Outputs
The tricky part: Trying to narrow project scope (but not too much)
This next part was tricky because I was trying to do opposing things at once. Narrowing the project focus based on what I’d learned so far with story maps and jobs to be done was objective one. Exploring ideas widely with sketching and a competitor solution analysis was objective two.
To strike a balance, I ended up doing two different story maps (instead of the usual one) based directly on insights I’d learned from user interviews. This would set the stage and allow me to explore more territory while sketching through ideas to ensure the right approach to this problem was being taken.
Story Maps: Leaving the scope open wide for exploration
Sketching like mad: Exploring all the angles
All said and done I looked at 7 different design directions.
1. Channels Smart Sorting
2. Channels AI Summary Inbox
3. Focus Mode AI Summary
4. AI Channel Summary with Action Items
5. Grouped Channel Dashboard
6. Interactive Channel Summary Timeline
7. Huddles Meetings Transcriptions
Closing in on a path & talking with a Sr. engineer friend at Slack
After doing my due diligence and exploring lots of different options, the decision was made to pursue an AI based solution. The reasoning was simple. It had the highest upside from both a potential business and user impact perspective.
I knew I needed more input though so I got in touch with an old college friend who was an engineer at Slack and showed him two quick ideas I’d wireframed up. I was trying to learn two things.
Q. How feasible are these ideas to actually build?
A. No major red flags raised
Q. Did he have any additional insight into how to make these concepts better?
A. Yes, maybe think about summarizing unreads and huddles transcripts (Slack’s version of Zoom notes)
Activities and Outputs
Catching the AI Wave with a little more help from my techie friends
I then ran x5 user tests. I was trying to find out three main things.
- Could users understand what the feature set was and what it was supposed to do?
- How useful would users find this feature set to their everyday workflows?
- Did users have any additional insights into how this feature set might fit their needs better?
User Test Insights
A Must Have
60% of participants were choosing to include action items as what they wanted to see returned in a channel summary.
It's a Team Sport
The lone exec. tester felt strongly that whatever info these summaries returned, it needed to be easily shareable with others on a team.
60% of interviewees had trouble understanding what the feature orientation pop up screens were trying to show.
40% of interviewees were not satisfied with where in the process you could rename summaries. It was also not obvious you even could rename summaries.
One user mentioned that when looking at the summary titles in the inbox there wasn’t much info there to help him. He wanted something to
A Nice Touch
60% of users mentioned that they appreciated the fact that the system alerted them to the fact that their channel summary was complete.
A technical deep dive with a machine learning engineer
I still had one burning question I had yet to answer that could make or break this feature set. Could AI summarize an entire channel or would a better approach be only trying to summarize Slack threads within a channel?
I needed to find out more so I spent an extra twenty minutes talking through some of the specifics of how these AI models work with one of my test users who happened to be a freelance machine learning engineer.
As a result of that conversation the decision was made to only summarize threads within a channel. This would be much easier and would require less time, effort and money to implement.
Activities and Outputs
The Home Stretch: Sticking to threads
AI Summarization of Channel Threads
Summarize up to x5 channels at a time (only the threads of each within a set time period) with the click of a button.
Action Items
Actionable items from the summary are extremely important to users, so they are pre-selected by default.
Summary Complete Alert
While you are waiting for your summary to finish, the left side panel lets you know its working. When it’s done you get a standard Slack pop up alert.
AI Summary Explainer Modal
Since this is a new feature, users need to quickly understand what it does and how to use it. A simple pop up does the trick.
Shareable Insights
Slack is about teamwork. Being able to share information from summaries with others helps everyone do their work more efficiently.
Summary Inbox Keywords
Keywords let you quickly scan your inbox for a past summary you may have created. You need to be able to have some context for what you are looking for.
Activities and Outputs
Giving people their sanity back: Calming the information storm.
Rather than having to manually browse through threads to get caught up to speed, you can now summarize those threads with the click of a button and save that info for later.
Reading on a computer screen is draining. By cutting down on user reading time in Slack, they’ll have the energy they need for their other day to day tasks!
If users know that putting their comments into the proper threads makes it easier to summarize them later they may be more inclined to keep things tidy in threads, in channel. More testing would be needed on this though.
If Slack makes employees more productive and not less, management will have little reason to look elsewhere for their internal communication needs.
“I immediately see how this would impact me, I might be gone for two or three days. And I've got 14 channels that have 50 messages, each all with threads in those in those channels. So having a tool that can capture the action items, summarize and then capture the action items that are specific to me, like, Dude... I could see raising VC capital for this 100%. Like you could raise money for this.”
- Andrew T.
Well would you look at that. Slack is actually building this thing!
Fast forward a few months. To my complete surprise I stumbled across this article on my Google feed one day. The most notable thing to me... It looks like I made the right decision after all.
Lessons Learned & Areas for Improvement
- Be careful with wording on usability tests: Unclear or overly complex tasks can mess up the objectivity of a test. A few of my questions could have been worded a bit more clearly.
- Do just enough research to move on: This is a difficult thing to master but is key to balancing speed and accuracy. I tried to do this throughout.
- Leave enough margin time for interviews: When scheduling interviews or user tests with people make sure I leave extra time in case someone is a no show or needs to reschedule. This happened a few times on this project but didn’t significantly mess up my timeline.
Future Roadmap
These are some things I’d look to continue to iterate on in the future with this feature.
Huddles are kind of Slack’s version of Zoom calls. Including text transcription summaries of these calls was something I had talked about with my engineer friend but didn’t make it into this first version of the feature.
This could be an interesting avenue to continue down because often times clarifying conversations with resolutions to work conflicts happen on video calls.
One of the things my machine learning engineer friend mentioned was that one of the avenues I might explore would be to design UI that would let users help improve the algorithm by letting them give feedback on how accurate or not accurate summary results were.
Were this project to continue this is something I could test with to improve effectiveness and accuracy of the feature.
Check out my other work!
