Information overload
Search, filters, lifecycle states, impact labels, confidence percentages, evidence, and multiple actions all competed for attention.
AI Product Design · Restaurant Operations
From an AI inbox to an operational copilot.
I designed a new AI-powered experience that tells restaurant operators what needs attention, explains why, and helps them take the next step.
AI product experience
Less interface. More direction.
The final concept moved away from asking operators to manage AI insights and toward letting the AI prioritize the day, explain each issue, and lead directly into action.
01 / The challenge
The first concept surfaced too much at once.
The initial prototype explored rich operational analysis, confidence scores, filters, lifecycle tabs, recommended actions, and ROI estimates. It was powerful, but the operator still had to scan, interpret, prioritize, and manage the system.
AI behaving like software
The draft made operators manage an AI insight tool instead of allowing the AI to behave like a proactive operational partner.
The key question changed from “How should operators manage AI insights?” to “What should the AI help them do next?”
Interaction language
Questions reveal depth only when the operator needs it.
Short, contextual prompts replaced generic dashboard controls. They make the next step feel conversational while keeping the home experience calm.
Opportunity cue communicates value before asking for action.
02 / Concept iteration
Compare two approaches to the same problem.
Toggle between Draft and Final to compare the same concept at full width, then use the open link for an unrestricted view.
Draft
Open full prototype
AI insight inbox
A data-rich command center organized around insights, filters, status, evidence, and recommended actions.
Final
Open full prototype
AI operations assistant
A calm home experience that prioritizes four issues and uses direct actions to move naturally into resolution.
03 / What improved
The final design changes the user’s job.
Instead of managing a system of recommendations, the operator arrives to a short, personalized briefing and acts through focused entry points.
Dimension
Draft
Final
Why it improved
Entry point
Inbox containing a set of AI insights
Personalized daily briefing
Creates immediate relevance and removes the feeling of another tool to maintain.
Prioritization
Search, filters, tabs, and impact labels
A short list that needs attention today
The AI performs the prioritization instead of outsourcing it back to the operator.
Information density
Metrics, confidence, evidence, status, and actions
One headline, context line, and next step
Progressive disclosure improves scanning while preserving deeper analysis when requested.
Interaction model
Manage, dismiss, customize, and track insights
Ask, investigate, submit, or see how
Actions are expressed in the language of the operator’s goal rather than the system’s workflow.
AI role
Insight generator inside a dashboard
Operational assistant with follow-through
Proactive prompts and contextual actions make the experience continuous rather than transactional.
Scope
Primarily optimization recommendations
Revenue, staffing, devices, and growth
The concept expands from analytics into a unified restaurant operations surface.
04 / Design decisions
Four decisions made the AI feel more useful.
01
Lead with the day, not the database
“There are four things that need your attention today” gives the page a clear purpose and frames the AI around urgency.
02
Replace taxonomy with intent
Instead of asking users to navigate statuses and impact filters, the final design offers contextual actions in the language of their goals.
03
Use conversation as progressive disclosure
Detailed analysis no longer occupies the starting screen. Operators enter a focused conversation only when a topic needs investigation.
04
Unify insight and execution
The assistant can identify a decline, explain an opportunity, prepare a support request, and celebrate growth from one operational surface.
The final concept is simpler because the AI does more of the organizing.
The final iteration reduces cognitive overhead, broadens the concept from analytics to operations, and creates a natural path from awareness to investigation and action.