Myna: Designing an AI Marketing Tool SMBs Actually Use
Myna began as a swipe-based approval, marketing app for restaurant owners, but real-world testing showed owners didn’t want content, they wanted guidance and proactive intelligence.
This case study shows how research reshaped Myna into a task-driven AI partner with measurable improvements in task completion, clarity and trust.
Overview
Myna is an AI-powered marketing assistant for small and medium businesses. It simplifies marketing tasks, reduces cognitive load, and drives engagement by guiding owners through high-value actions.
Problem Space
Restaurant owners juggle content creation, daily operations, and customer acquisition, leaving little time for marketing.
Most owners relied on costly agencies or complex tools, with 70% reporting difficulty managing marketing consistently.
Initial Product State
• Swipe-based app gamified marketing tasks, but were confusing, felt gimmicky, and shallow.
• Users struggled to understand the value of actions.
• No clarity or guidance.
Initial Outcomes
• <10% users returned; task completion was near zero.
• low adoption, low willingness to pay.
Solution
Pivoted from a swipe-based, approval app to an AI-assisted, task-focused experience combining chat, tasks, and insights.
My Role
Founding Designer - led research, UX/UI, brand, and product strategy.
Team
CEO, CTO, AI Engineer, 2 Developers, 2 Junior Designers
Project type & Timeline
Mobile app design B2B SaaS · AI for Restaurant Tech · 12 weeks
Project Outcome
65%
Increase in task completion rate
70%
Decrease in user confusion
50%
Decrease in AI-related complaints
STAGE 1
Initial Concept
Context
The Vision
The founder envisioned a Tinder-like swipe UI. Users swipe through cards to approve or dismiss suggested marketing actions for fast approvals. 'One gesture, zero friction.’
The Challenge
1
Misalignment and unclear value
I joined after the swipe concept was defined but before any user validation.
2
No UX research
Early decisions were assumption-driven.
3
Heavy speed-to-market pressure
Investors expected fast results.
Usability Breakdown
Usability tests & UX audits to validate assumptions and identify friction points
User testing revealed major confusion and low perceived value:
“What am I looking at?”
“It would make things faster… but I wouldn’t pay for it.”
“I can just open ChatGPT and get all this done for free.
Core issues identified (UX audit + tests):
1
No clear hierarchy
2
Weak guidance
3
Ambiguous cues
4
Users couldn’t identify the core action or navigate confidently
UX Audit and user test issues mapped on swipe cards
Business Impact
The experience failed to communicate value
1
<10% returning users
60% felt swipe UI was a gimmick
Owners wanted help running their business, not approving content
2
Near-zero task completion
No hierarchy or guidance
Users didn’t know what action mattered
3
Low willingness to pay
ChatGPT seen as a free substitute
Research
To validate assumptions and understand real needs, I led the research plan + analysis:
Interviews with 13 restaurant owners
Desk research on owner behavior, workflows, and pain points
Competitive benchmarking + SWOT to understand market gaps
Empathy & affinity mapping to synthesize themes
Quantitative & qualitative analysis of time spent, costs, and operational bottlenecks
Competitor Landscape Review
Market Positioning & SWOT Findings
Insight Synthesis: Affinity Mapping
Research Repository: Interviews & Analysis
Behavioral & Business Findings
What owners actually cared about:
Sample size:
13
“If you can bring me catering orders, I’ll pay you tomorrow.”
“If the app could make my everyday processes easy, I’d pay for it."
“ I want something that’s actively chasing opportunities for me.”
What Owners Valued
Operational + marketing support for daily tasks
Examples: catering leads, local events, competitor activity, weather impact, sales/labor metrics, trends, ingredient pricing, team motivation
Proactive guidance, not guesswork
Quantitative Insights
6–10 hrs/week spent responding to reviews (~20% admin time)
70% struggled with content creation + trend monitoring
CAC rising 15–25% annually
Competitive Findings
Revealed exsisting solutions and best practices, weakneses and opportunities for Myna.
Existing tools required owners to pull insights
Dense dashboards that shifted the burden of analysis onto already time-constrained owners
Clear opportunity for proactive, context-aware intelligence
Takeaway
Restaurant owners don’t want to pull information. They want intelligence pushed to them. The swipe-card concept, built on assumptions rather than validation, could never deliver the proactive guidance owners actually needed.
1
Key Findings
Usability issues (confusion, low hierarchy)
2
Behavioural Disconnect
Owners want guidance
3
Business Misalignment
low perceived value, low adoption, overlap with existing tools like ChatGPT
STAGE 2
Pivot to Task-Focused Experience
Ideation
How can I make Myna less of a tool… and more of a partner?
Hypothesis
"If we reduce cognitive load and reframe marketing as small, meaningful weekly tasks, owners will take consistent action and feel more in control."
To validate this, I aligned the team on:
1
Clear user value derived from research
2
Engineering feasibility with the CTO
3
A flexible workflow that could evolve with the company
Information Architecture (IA) and System Design.
I led task-flow architecture and product IA connecting tasks, analytics, AI, notifications, user actions, and outcomes to ensure alignment and catch issues early.
Clarifying:
How much structure owners wanted
When AI should take initiative vs. stay passive
Which tasks drove meaningful business outcomes (reviews, social, intelligence)
Task focused process flow on figjam
Iteration
I explored 3 early concepts
Internal reviews and critiques made the weekly task model the clear winner.
Enhanced swipe model
still shallow and content-centric
Smart suggestions feed
too noisy, didn’t reduce cognitive load
Weekly task system
clearest structure + sense of control
Converting the Concept Into a System
I collaborated with the CTO and AI engineer to translate the idea into a buildable framework:
AI capabilities and limits
Task-trigger logic
Required data signals
How task success should be measured
Solution
Designing a task focused experience
Clickable prototype
Old vs New
Old Concept (Swipe)
Pull-based (user has to come look at content)
Gamified gestures
Shallow actions (“approve content”)
No clarity on value
No guidance
New Concept (Task-focused)
Push-based (app highlights what matters)
Structured tasks & flows
High-value actions (reviews, metrics, leads)
Direct business outcomes
Proactive intelligence
Unexpected Roadblock
Tested with 10 restaurants in cohorts over 2 weeks; tracked task completion and feedback.
Findings:
Flow felt rewarding for most users
Owners naturally preferred chat, showing a clear path toward guided assistance.
Weekly tasks were easy to complete (finished in 1–2 days).
Some users found campaign activities overwhelming.
AI hallucinations and generic insights reduced trust.
See usability comments here
Quick Fix
Removed chat input, simplified to tap-based actions
Helped reduce hallucinations and stabilize the agent ahead of launch.
Before: with chat input
After: tap-based flow replaces chat
While detailed A/B testing or extended usability studies are ongoing, early testing and direct feedback from users helped inform rapid iterations post-pivot.
Impact
Outcomes (Weeks 1–4):
Early adoption metrics focus on user behavior and engagement rather than revenue, reflecting the product’s testing phase.
Metric
Baseline (Old System)
After Redesign (Week 1-4)
Change
How we measured
Task Completion (Adoption)
Less than 20% of recommended actions completed (6 pilot users, Usability test)
85% completed (13 pilot users)
+25% points
Analytics and task logs.
User Confusion (Clarity)
Avg. 3 navigation errors per session
< 2 errors per session.
Confusion: ↓70%
Usability testing and session recordings.
AI-Related Complaints (Trust)
Almost everyone had issues trusting the AI insights.
3/10 users complained about AI hallucinations and generic insights.
Complaints dropped post iteration: ↓50%
Direct user feedback and internal testing.
Continued Engagement (Retention)
N/A (new product pivot)
Drop after Week 1; users completed tasks within 1–2 days
-
Activity logs and follow-up interviews.
Session Duration (Activity)
-
< 3–4 min avg.
↓ (needs engagement loops)
Analytics tracking session duration
Execution
On a tight timeline, I focused on delivering high-quality key screens rapidly:
Delivered key screens + edge cases while guiding junior designers
Led junior designers through structured QA and accessibility tests scoring 90%+
Developers leveraged Claude for vibecoding
85% of final screens met quality standards
Detailed handoff in Figma and QA tracked in GitHub
Detailed dev handoff in Figma.
QA documented in Figma, issues tracked in GitHub.
Constraints & Tradeoffs
With a small engineering team and early-stage AI performance to stabilize, we focused on strengthening the core task-driven workflow before layering on complex features. This meant temporarily deprioritizing multi-channel automation, advanced analytics, and deeper engagement loops. These trade-offs allowed the product to mature in the right order, building a reliable, scalable experience while informing future design decisions.
Key Trade-offs:
Speed-to-Market vs. Research Validation
Early shipping led to assumption-driven swipe UI design.
Resulted in low adoption and confusion.
Insight: Validating assumptions early prevents costly pivots.
Core Features vs. Advanced AI Capabilities
Advanced AI features delayed to stabilize the MVP.
Resulted in low trust
Insight: Reliability and trust matter more than novelty, especially for SMBs.
Shallow Engagement vs. Measurable Business Outcomes
Gamification prioritized “fun” over real results.
Users engaged briefly but didn’t complete tasks or pay.
Insight: Design must solve actual problems; real value > engagement.
Investor “Wow Factor” vs. True User Needs
Demo-ready polish impressed investors but didn’t meet user needs.
Early adoption was low, highlighting misalignment.
Insight: Long-term adoption and trust outweigh initial “wow.”
Cohort Testing vs. Broad Rollout
Small test groups allowed rapid iteration but limited exposure.
Metrics were early indicators, not full-market validation.
Insight: Controlled cohorts enable faster learning with lower risk.
Wrapping it Up
Recap
Joined early, no UX research, high pressure to ship
Original swipe UI didn’t match real-world behavior
Conducted deep research → revealed need for proactive intelligence
Pivoted from swipe → task-focused experience (replaced product model, not just UI)
Early adoption metrics improved: task completion ↑65%, user confusion ↓70%, AI complaints ↓50%
My Learnings
This project humbled me in the best way possible.
01
Solve real problems first. Design polish cannot replace weak value.
02
Validate early. Observing actual behavior is more reliable than assumptions.
03
Build trust before engagement. Users return for usefulness, not novelty.
04
Prioritize reliability over flashy features. Stable, context-aware AI drives adoption.
05
Failure is a mirror, not a verdict. Every misstep taught more than success, improving the product and my user-centered approach.
06
Learning…
Next steps
Evolve from a task-based workflow to a multi-modal experience combining tasks, insights, content, and guidance
Expand qualitative validation once improvements roll out: With the temporary pause on new restaurant onboarding so the team can strengthen core features and focus on high-value AI outputs like social content and video generation, the next step is to reopen pilots and gather richer qualitative feedback on how the redesigned workflow supports owners in their day-to-day business outcomes.
Future Metrics: Adoption, retention, revenue, and referrals will be tracked as the product matures.
Validate which workflow users will actually pay for.
Improve AI accuracy and trust signals.
Tighten IA across chat, tasks, insights, and notifications.
Add retention loops tied to real restaurant activity.
Double down on one hero workflow that drives value.
