Enhance Slack AI Interaction through AI empowered Design
Enhance Slack AI Interaction through AI empowered Design
Overview
Overview
The launch of first pilot of Slack AI has received positive customer feedback about different features, but they do not remember to use them frequently to continue paying for it. Therefore, there needs a new, radical way for Slack users to find and engage with AI features.
The launch of first pilot of Slack AI has received positive customer feedback about different features, but they do not remember to use them frequently to continue paying for it. Therefore, there needs a new, radical way for Slack users to find and engage with AI features.
*: This is a passion project inspired by a design challenge, and the video featured below is credited to Slack.
*: This is a passion project inspired by a design challenge, and the video featured below is credited to Slack.
Problem Statement
Problem Statement
While Slack AI offers valuable features, low discoverability hinders user adoption, leading to product-market fit (PMF) challenges. With Deepseek, I developed user behavior hypotheses, defined key metrics for data collection and validation, deconstructed the core problem into targeted, solvable design challenges.
While Slack AI offers valuable features, low discoverability hinders user adoption, leading to product-market fit (PMF) challenges. With Deepseek, I developed user behavior hypotheses, defined key metrics for data collection and validation, deconstructed the core problem into targeted, solvable design challenges.
Business Challenge
Business Challenge
The Sean Ellis Test score is currently 32% ( which is lower than the goal of 40% for PMF)
The Sean Ellis Test score is currently 32% ( which is lower than the goal of 40% for PMF)
Design Challenge
Design Challenge
Users perceive difficulty in finding to AI entries.
Users hardly interact with AI features.
Users don’t know how to act on the AI generated results.
Users doubt the AI generated results’ accuracy.
Users perceived limited value from the AI features.
Users perceive difficulty in finding to AI entries.
Users hardly interact with AI features.
Users don’t know how to act on the AI generated results.
Users doubt the AI generated results’ accuracy.
Users perceived limited value from the AI features.
Solution Prototype
Solution Prototype
Make visible AI entry points
Make visible AI entry points
Highlight AI capability through icon animation.
Add extra AI entry buttons in threads for quick access.
Highlight AI capability through icon animation.
Add extra AI entry buttons in threads for quick access.
Shorten action paths
Shorten action paths
Highlight ‘Suggested Actions’ in AI outputs.
Enable quick redirection to document, chats, etc.
Highlight ‘Suggested Actions’ in AI outputs.
Enable quick redirection to document, chats, etc.
Reduced interaction drop-off
Reduced interaction drop-off
Offer ‘Notify me when ready’ for loading long summaries.
Enable user to go back to unacted AI-output.
Offer ‘Notify me when ready’ for loading long summaries.
Enable user to go back to unacted AI-output.
* The solution designed here targets only for prioritized design challenges to simulate the real-life impact-feasibility trade-offs.
* The solution designed here targets only for prioritized design challenges to simulate the real-life impact-feasibility trade-offs.
Success Metrics
Success Metrics
Main design goal: Make AI features visible and actionable.
Main design goal: Make AI features visible and actionable.
Design Goals
Design Goals
Design Goals
Metrics
Metrics
Metrics
Baseline*
Baseline*
Baseline*
Target
Target
Target
How to measure
How to measure
How to measure
# 1
Improve Discoverability
# 1
Improve Discoverability
# 1
Improve Discoverability
CTR on AI Entry Points
CTR on AI Entry Points
CTR on AI Entry Points
Time-to-First-AI-Use (New Users)
Time-to-First-AI-Use (New Users)
Time-to-First-AI-Use (New Users)
Awareness
Awareness
Awareness
15%
15%
15%
Never (60%)
Never (60%)
Never (60%)
‘Didn’t know this AI feature’ (70%)
‘Didn’t know this AI feature’ (70%)
‘Didn’t know this AI feature’ (70%)
30%
30%
30%
< 30% never-used
< 30% never-used
< 30% never-used
< 30% unaware
< 30% unaware
< 30% unaware
Track clicks on AI buttons ot tips
Track clicks on AI buttons ot tips
Track clicks on AI buttons ot tips
Onboarding analytics
Onboarding analytics
Onboarding analytics
Post-rollout survey
Post-rollout survey
Post-rollout survey
# 2
Reduce Drop-off
# 2
Reduce Drop-off
# 2
Reduce Drop-off
Drop-off During AI Summary Prcessing
Drop-off During AI Summary Prcessing
Drop-off During AI Summary Prcessing
Finish rate of AI Summary
Finish rate of AI Summary
Finish rate of AI Summary
Action-rate on AI suggested content
Action-rate on AI suggested content
Action-rate on AI suggested content
Perceived Actionability
Perceived Actionability
Perceived Actionability
40%
40%
40%
60%
60%
60%
25%
25%
25%
‘Don’t know how to act on it’ (2/5)
‘Don’t know how to act on it’ (2/5)
‘Don’t know how to act on it’ (2/5)
< 20%
< 20%
< 20%
85%
85%
85%
50%
50%
50%
‘Easy to act on AI suggestions’ (4/5)
‘Easy to act on AI suggestions’ (4/5)
‘Easy to act on AI suggestions’ (4/5)
Analytics during AI summary loading
Analytics during AI summary loading
Analytics during AI summary loading
Analytics about when user exit AI summary
Analytics about when user exit AI summary
Analytics about when user exit AI summary
CTA clicks on AI suggested actions
CTA clicks on AI suggested actions
CTA clicks on AI suggested actions
User interviews
User interviews
User interviews
Main business goal: Prove AI features drive retention and revenue.
Main business goal: Prove AI features drive retention and revenue.
Business Goals
Business Goals
Business Goals
Metrics
Metrics
Metrics
Baseline*
Baseline*
Baseline*
Target
Target
Target
How to measure
How to measure
How to measure
Improve Feature Adoption
Improve Feature Adoption
Improve Feature Adoption
% of MAUs using AI features
% of MAUs using AI features
% of MAUs using AI features
Interaction Rate of AI
Interaction Rate of AI
Interaction Rate of AI
Sean Ellis PMF Score
Sean Ellis PMF Score
Sean Ellis PMF Score
Conversion Rate after trial
Conversion Rate after trial
Conversion Rate after trial
Support Tickets Related to AI
Support Tickets Related to AI
Support Tickets Related to AI
5%
5%
5%
1-2 times/week
1-2 times/week
1-2 times/week
32%
32%
32%
Low
Low
Low
High
High
High
15%
15%
15%
3-4 times/week
3-4 times/week
3-4 times/week
40%
40%
40%
+ 20% AI subscription users
+ 20% AI subscription users
+ 20% AI subscription users
50% less
50% less
50% less
Analytics
Analytics
Analytics
Analytics on AI CTAs
Analytics on AI CTAs
Analytics on AI CTAs
Post-rollout survey
Post-rollout survey
Post-rollout survey
Subscription Analytics
Subscription Analytics
Subscription Analytics
Analytics
Analytics
Analytics
* The baseline score is based on the hypothesis of the real-life data.
* The baseline score is based on the hypothesis of the real-life data.
Process & Refletions
Process & Refletions
This project explores how AI can enhance collaborative design processes. Using Deepseek, we simulated a team-based workflow for for early-stage assumption validation and problem framing. The sketching and prototyping phases were executed independently, allowing me to focus on hands-on design execution.
This experience highlighted AI’s potential as a thought partner in early-stage design—particularly for refining problem statements and validating assumptions. Moving forward, I aim to strategically integrate AI tools to augment (not replace) critical thinking, ensuring they complement human creativity in tackling complex challenges.
This project explores how AI can enhance collaborative design processes. Using Deepseek, we simulated a team-based workflow for for early-stage assumption validation and problem framing. The sketching and prototyping phases were executed independently, allowing me to focus on hands-on design execution.
This experience highlighted AI’s potential as a thought partner in early-stage design—particularly for refining problem statements and validating assumptions. Moving forward, I aim to strategically integrate AI tools to augment (not replace) critical thinking, ensuring they complement human creativity in tackling complex challenges.
* The ‘User Testing & Implementation’ stage is not covered in this case study’s process.
* The ‘User Testing & Implementation’ stage is not covered in this case study’s process.
* The ‘User Testing & Implementation’ stage is not covered in this case study’s process.
My other work
My other work
