ORIN RESERVATIONS
Orin AI: Humanising Interaction in High-Pressure Environments.

ORIN is a leading hospitality management platform used by top-tier venues in London. Working under the mentorship of our Senior Product Designer, I took ownership of the end-to-end design for the AI Reply feature. My goal was to leverage generative AI as a high-utility tool that empowers restaurant managers to communicate faster without losing their authentic brand voice.

Role

Role

UX Designer

Duration

Duration

3 months

Industry

Industry

Hospitality

My Contribution

My Contribution

I worked closely with a Senior Designer to translate business goals into a mobile interface. I was responsible for the execution of field research, wireframing, high-fidelity UI, and usability testing.

Problem

Restaurant managers operate in a chaotic environment. Guest enquiries (e.g., "Do you have outdoor seating?") often take a backseat during busy shifts.

  • Writing Anxiety: Many managers (especially non-native speakers) feel "paralysed" by the fear of making grammatical errors.

  • Physical Friction: Managers juggle plates and menus, making "desk-focused" messaging tools impossible to use safely.

  • The Trust Gap: Early explorations showed that staff were wary of "robotic" automation that could damage guest relationships.

Objectives

Balancing Automation with Human Authenticity

The primary goal was to integrate AI into a high-pressure hospitality environment without it feeling like an interruption. Working with my Senior Designer, we defined three core objectives to guide the project:

Minimise Interaction Cost Restaurant managers are constantly on the move. We needed to reduce the time spent drafting a message from minutes to mere seconds, ensuring the tool could be used with one hand during a busy service.

Build a Trust-Based Workflow AI can be intimidating or unpredictable. Our objective was to ensure staff felt in 100% control of the output, eliminating "Send-Anxiety" and ensuring no message was sent without human approval.

Maintain Authentic Brand Voice Every venue has a unique "vibe." The objective was to avoid "robotic" or overly formal replies (The Black Tie Paradox) by allowing the AI to adapt its tone to match the specific personality of the restaurant.

Solutions

  1. Introduced Intent-Based Toggles ("Reply Positively" / "Reply Negatively") to act as cognitive shortcuts for staff mid-service.

  2. Developed the "Three-Sentence Hierarchy" for AI drafts to ensure messages are scannable in under 3 seconds.

  3. Designed the "Improve" feature logic to polish raw staff notes into professional, welcoming brand-aligned greetings.

  4. Established a "Human-in-the-Loop" editable draft system, ensuring staff maintain full control over every AI-generated response.

The Impact

The final design wasn’t just a UI update; it was a strategic overhaul that transformed guest communication from a chore into a 10-second reflex. By reducing cognitive load and removing linguistic barriers, the system enabled staff to navigate high-pressure shifts with greater speed and confidence.

Time Saving

90%

Response Efficiency: Average reply times dropped from 2 minutes to just 10 seconds, allowing staff to manage the "Pending" queue instantly between tables

User Satisfaction

8.5/10

Confidence Score: Venue managers reported a significant reduction in "Social Translation Fatigue" and increased trust in brand-aligned communication.

Data is based on final usability testing where participants performed tasks on the new "Verification Model" compared to manual composition, measuring average time-on-task and qualitative trust scores.

Discovery

I didn't guess the features; I used specific UX methodologies to uncover the friction points of hospitality staff.

Ethnographic Observation

  • What I Did: I spent 5 days shadowing staff in active London venues during peak "Friday Night Rushes." I tracked physical phone usage while managers were multitasking.

  • Finding: 80% of staff are "one-handed" users. They check messages while walking or holding items.

  • Design Outcome: I scrapped all pop-ups and top-aligned buttons. I placed all AI triggers in the "Thumb Zone" (bottom-right of the text box) for safe, one-handed operation mid-service.

Contextual Inquiry

  • What I Did: Conducted 50+ one-on-one interviews with native and non-native speakers to uncover the psychological barriers to AI adoption.

  • Finding: The "Black Tie" Paradox. Staff feared the AI would sound too formal or "robotic." One tester noted: "If it sounds this posh, guests will think they need to wear a black tie just for a burger."

  • Design Outcome: I designed Tone & Length Toggles. This ensures the AI matches the venue's specific vibe (Casual vs. Professional), making the communication feel authentic.

Competitor Audit

  • What I Did: Audited 5 major hospitality tools like SevenRooms, Resy, OpenTable, and analysed AI Tools like Intercom.

  • Finding: Most tools use "Performative AI" loud pop-ups and complex settings that interrupt the user flow.

  • Design Outcome: I pivoted to "Quiet UI." The buttons only appear when needed, staying invisible when the user knows what to type, respecting their natural workflow.

Insights

Insights

Raw research data only becomes actionable when it is structured. I used Affinity Mapping to synthesise over 200 observation notes, guest enquiries, and staff pain points gathered during the discovery phase. By grouping these insights into thematic clusters, I was able to identify the recurring patterns that directly guided my design priorities.

Raw research data only becomes actionable when it is structured. I used Affinity Mapping to synthesise over 200 observation notes, guest enquiries, and staff pain points gathered during the discovery phase. By grouping these insights into thematic clusters, I was able to identify the recurring patterns that directly guided my design priorities.

Affinity Diagram

To synthesise qualitative data from 50+ staff interviews and find recurring friction points.

Key Findings

1. The Trust Deficit

Finding: Staff expressed a high level of "Send-Anxiety." They were unwilling to use AI if they couldn't verify the information (e.g., allergy details or booking availability) before it reached the guest.

Impact: This led to the "Draft-First" logic, ensuring the AI never sends autonomously, but acts as a highly efficient ghostwriter.

2. One-Handed Ergonomics

Finding: Field research showed that 80% of managers handle enquiries while walking or multitasking on the floor. Standard pop-up AI assistants were too physically demanding to use.

Impact: We prioritised a "Thumb-Zone" UI, placing all AI triggers within the bottom 30% of the screen to facilitate seamless, one-handed service.

3. The "Black Tie" Paradox

Finding: Generic AI models default to an overly formal "robotic" tone. Staff felt this damaged the authentic, casual "vibe" of their local venues.

Impact: This necessitated Tone & Length Toggles, allowing the AI to mirror the specific personality of the venue, whether casual or professional.

Results

Ideating solutions

The analysis of my Affinity Map led to a series of high-impact linguistic and structural decisions. My goal was to move beyond "generic AI" and create a voice that feels native to high-end hospitality.

Echoing

Intent-Based

Intent-Based

The 3-Sentence Rule

The 3-Sentence Rule

Human in the loop

Human in the loop

Decision 1: Implementing "Echoing" as a Trust Metric

  • The Logic: Based on "Anchor Words" identified in research. Guests feel "ignored" if a response is too generic.

  • The Action: I established a rule: the AI must mirror the guest’s specific nouns (e.g., "high chair," "window table") in the first five words.

  • Impact: This reduces the guest's need to follow up, as they feel their specific needs have been accurately processed.

Decision 2: Transitioning from Free-Text to Intent-Based Toggles

  • The Logic: To combat "Social Translation Fatigue," I moved the primary interaction from a blank text box to binary choice buttons.

  • The Writing Action: I authored the microcopy for "Reply Positively 👍" and "Reply Negatively 👎" as cognitive shortcuts.

  • Impact: This reduced the staff's decision time by over 70%, as they no longer had to think about "how" to phrase a "no."

  • The Logic: To combat "Social Translation Fatigue," I moved the primary interaction from a blank text box to binary choice buttons.

  • The Writing Action: I authored the microcopy for "Reply Positively 👍" and "Reply Negatively 👎" as cognitive shortcuts.

  • Impact: This reduced the staff's decision time by over 70%, as they no longer had to think about "how" to phrase a "no."


Decision 3: The Scannability Hierarchy (The 3-Sentence Rule)

  • The Logic: Staff need to verify AI drafts in under 3 seconds. Long paragraphs were a friction point.

  • The Writing Action: I enforced a strict linguistic template:

    1. Greeting & Echo: (Personalised validation).

    2. Core Data: (Date/Time confirmation).

    3. The Hook: (A warm brand-aligned closing).

  • Impact: This ensures the draft is scannable even during the busiest dinner rushes.

  • The Logic: Staff need to verify AI drafts in under 3 seconds. Long paragraphs were a friction point.

  • The Writing Action: I enforced a strict linguistic template:

    1. Greeting & Echo: (Personalised validation).

    2. Core Data: (Date/Time confirmation).

    3. The Hook: (A warm brand-aligned closing).

  • Impact: This ensures the draft is scannable even during the busiest dinner rushes.


Decision 4: The "Improve" Feature Prompt Logic

  • The Logic: I recognised that some staff still prefer to type manually but lack the time for "polishing."

  • The Writing Action: I designed the underlying prompt for the "Improve" button. It acts as a "Hospitality Filter" that takes blunt inputs (e.g., "ok see u") and expands them into professional greetings ("We look forward to welcoming you on Saturday").

  • The Logic: I recognised that some staff still prefer to type manually but lack the time for "polishing."

  • The Writing Action: I designed the underlying prompt for the "Improve" button. It acts as a "Hospitality Filter" that takes blunt inputs (e.g., "ok see u") and expands them into professional greetings ("We look forward to welcoming you on Saturday").

Interaction Design

With the content strategy defined, I focused on the Interaction Design to ensure the AI assistance felt like a natural extension of the staff's existing workflow. I mapped the end-to-end journey to ensure the experience felt predictable, efficient, and free of unnecessary detours.

Mapping the Dialogue Journey (Flow Diagrams)

I began by mapping the end-to-end messaging journey, from the moment a notification arrives to the final confirmation of a booking.

  • The Goal: To reduce friction by identifying exactly where AI intervention would be most valuable.

  • Design Decision: I integrated the AI draft directly into the conversation thread, ensuring that the staff member never has to leave the context of the chat to generate a response.

Intent-Based Interaction (Prototyping)

To move beyond competitor analysis, I explored screen behaviour through hand-drawn sketches and low-fidelity explorations.

  • The Concept: I prototyped the "Intent Toggles" (Reply Positively/Negatively) as primary actions.

  • The Logic: By sketching multiple variations, I found that placing these buttons near the thumb-zone (bottom of the screen) allowed for rapid iteration and execution during a busy shift.

Visual Feedback & Transparency

It was critical to show the AI "working" without creating anxiety for the user.

  • Typing Animation: I designed the interaction so that the AI draft appears with a subtle "thinking" state, followed by the text appearing as if it's being typed.

  • The Verification Step: The draft is presented in a clearly distinct bubble, allowing the staff member to edit the text immediately before hitting "Send". This ensures they maintain 100% control over the final output.

Prototype

The implementation of orinAI Messaging fundamentally transformed the restaurant's operational speed and communication quality. By moving from a "composition" model to a "verification" model, we achieved the following measurable impacts:

  • 90% Reduction in Composition Time: By providing 90% completed drafts, we eliminated "writer’s block". Staff can now manage the "Pending" message queue in seconds between serving tables.

  • Reduced Cognitive Load: Shifting the task from "creative writing" to "verifying a draft" allowed staff to remain mentally present with on-site guests while managing digital enquiries effortlessly.

  • Enhanced Guest Confidence: By explicitly acknowledging special requests such as repeating "high chair" or "window table" the AI-driven responses led to fewer follow-up questions and higher guest trust.

  • Linguistic Consistency: Every guest now receives a consistently professional, warm, and welcoming response, maintaining the brand’s high hospitality standards regardless of shift intensity.

  • 100% Data Accuracy: Automated data injection removed human errors in critical booking details (dates, times, and party sizes), ensuring the guest and the restaurant are always on the same page.

Reflections

Context over Visuals: Designing for the "Service Floor" taught me that attention is a scarce resource. Respecting one-handed ergonomics and the Thumb-Zone was more important than flashy UI.

  • Trust via Control: Moving from "composition" to "verification" proved that users don't want AI to speak for them, but to draft with them. The human remains the final authority.

  • Linguistic Equality: I learned that AI can be a powerful tool for inclusivity, giving non-native staff the linguistic confidence to communicate professionally regardless of shift intensity.

  • Brevity is Hospitality: Stripping away "flowery" AI speech in favour of the 3-Sentence Rule proved that in a high-pressure kitchen, speed and clarity are the highest forms of service.