B2B // AI // UX // UI
Operational admin time
Saved time
of participants who said
they'd open it the next morning
BEFORE YOU SCROLL… A LITTLE CONTEXT.
// THE PROBLEM
Independent café managers are losing money daily to decisions made on gut feel, because the information they need to decide correctly is scattered across too many places to use in real time.
// THE COMPETITVE GAP
7shifts: Setup wizard required before any value. Free plan exists but multi-dashboard interface overwhelms the manager who needs one screen, not five.
MarketMan: 2–4 weeks of onboarding with a dedicated team. $239/month minimum. 12-month contract. This is structurally incompatible with a first-time tool adopter who cited budget as a barrier.
Restaurant365: Enterprise ERP.
Built for franchise operators with CFOs and ops directors.
Not SMBs.
→ No credible, mobile-first, zero-configuration, SMB-priced operations tool exists for the independent café manager. Not one competitor offers a unified AI-assisted production planning at any SMB price point.
// THE USER I DESIGNED FOR
I called them the FIREFIGHTER MANAGER.
Reactive, memory-reliant, tool-skeptical.
Responds to problems when they surface, not before.
Tried ChatGPT for scheduling, abandoned it because it was slower than paper.
Will adopt a tool only if it delivers visible value in the first session with near-zero setup cost.
// KEY FINDINGS
//01
The manager's operation runs entirely on their memory
//02
Existing tools fail before delivering value
//03
Invisible margins
//04
Stockouts during service & preventable waste
//05
Reactive decisions
//06
Scheduling takes up to an hour per week on paper
// THE CONCEPT
Concept 1 // The Morning Briefing
Push-delivered daily brief. 3–5 prioritized cards. One recommended action each. Zero configuration.
Concept 2 // The Control Room
Unified multi-panel dashboard. Everything visible simultaneously.
Concept 3 // The Shift Copilot
Context-triggered prompts at shift open, mid-shift, and close. Appears at natural decision points throughout the day.
I rejected the Control Room immediately.
A dashboard requires the manager to seek information. The Firefighter doesn't seek, they respond. This is the exact interaction model that fails them in every existing tool.
✓ I chose the Morning Briefing
For three specific reasons:
lowest time-to-first-value,
Highest RICE scores concentrated in its features,
Best fit for the primary archetype.
// THE DESIGN DECISIONS THAT DEFINED THE PRODUCT
//01
Hub-and-spoke > tab navigation
Tab navigation implies five equally important destinations. Delta has one: the briefing. Everything else is secondary.
A tab bar would flatten that and invite exploration they won't do.
//02
Confirm/override > data entry
They won't log data. So data capture happens as a byproduct of the decision itself. When she taps Confirm on a production card, that confirmation is the data point. The system learns from what she does, not from what she fills in.
//03
Confidence indicators on every recommendation.
"Based on your last 4 Thursdays" builds trust. "AI-powered" doesn't. She needs to know whether a recommendation comes from her data or an industry default — so she can decide how much weight to give it.
//03
Confidence indicators on every recommendation.
"Based on your last 4 Thursdays" builds trust. "AI-powered" doesn't. She needs to know whether a recommendation comes from her data or an industry default — so she can decide how much weight to give it.
The % of daily operational decisions made outside vs in-app.
NOT DAU, not session time.
//TARGET
>70%
of daily decisions made with in-app data within 60 days of onboarding.
//KILL THRESHOLD
<40%
confirm/override interaction rate on production cards within 30 days = managers seeing recommendations but not trusting them.
// THE IMPACT
of participants who said
they'd open it the next morning
// RETROSPECTIVE
WHAT WORKED
Starting from a real observed problem with a real user made every design decision easier to justify. The research wasn't a phase I completed, it was the thing the product is built on. The manager interview gave me the 7-source finding, the archetype, and the kill criteria. That's the whole product in three outputs.
WHAT I'D DO DIFFERENTLY
The research is the thing the product is built on. I interviewed 4 F&B managers and 4 staff. I should've took more time on the research by interviewing more users. Same for the testing.