AI use-case expansion — seven new ideas

Last updated: 2026-06 · candidate expansion (UC15–UC21)

Seven candidate use cases — multi-channel inbound comms on a customer-data platform, a Staff-GPT internal assistant, and more. Concept stage; each has a mock-up you can open. These graduate into the built portfolio once validated by real demand.

This grouped page has been superseded. Each of these use cases now has its own detail page: UC15 · UC16 · UC17 · UC18 · UC19 · UC20 · UC21. This page is kept for reference.

15. Multi-channel guest comms

Front of house & bookings · Build (on the UC1 brain)

Interactive concept

See the mock-up for this use case

A static, on-brand design concept — illustrative data, not a live system.

Open the UC15 concept mock-up →

Show

A booking confirmation goes out by SMS / WhatsApp / email — not a no-reply. The guest replies "can we do 8 instead of 7?" and the agent modifies the booking and confirms. Pre-arrival it sends parking and an allergen note; post-visit it asks for a review and routes a complaint to a human. Break it: ask something off-scope and watch the handoff, with the quality layer logging the deflection.

Why

The central idea: AI agents for inbound communication across channels off a customer-data platform. Turning dead no-reply confirmations into a two-way thread is an instant, relatable operator win.

17. Staff ops assistant (Staff-GPT)

Team & ops · Build

Interactive concept

See the mock-up for this use case

A static, on-brand design concept — illustrative data, not a live system.

Open the UC17 concept mock-up →

Show

A new server asks "table 12 wants the pithivier without bone marrow — can we?" and the agent answers from the allergen and recipe knowledge base, citing the SOP. "How do I close the till?" gets a step-by-step. A manager asks "who's rostered Friday dinner?". It tracks what new hires ask and surfaces training gaps. Break it: ask something not in the SOPs and watch it decline rather than invent.

Why

Labour is SG F&B's deepest pain — roughly 10,000 unfilled roles, ~35% annual turnover, ~307 closures a month. Cuts onboarding time and tribal-knowledge loss. A "Staff Agent / check-GPT" concept, and a bridge to the training tier and a culinary-academy pedagogy play.

16. Dynamic markdown + specials board

Back office & finance · Build (ties UC3 + UC7)

Interactive concept

See the mock-up for this use case

A static, on-brand design concept — illustrative data, not a live system.

Open the UC16 concept mock-up →

Show

A 3pm dashboard: tonight is forecast 30% under, with 14 almond croissants and 6 lamb portions at risk. The agent proposes timed markdowns and bundles ("2 pastries SGD 7 after 5pm") and a digital specials board updates — in-store screen plus an Instagram story. The margin floor is respected. Break it: ask it to discount a loss-leader below cost and watch the guard hold.

Why

Food waste plus margin under Singapore cost pressure. Opens the cafe / bakery segment the fine-dining demonstrator otherwise misses, and reframes "discounting" as planned yield management. PDPA-light. Markdown only, to clear surplus and cut waste — not surge pricing.

18. Proactive yield-fill

Marketing & reputation · Build (orchestration over UC3 / UC6 / UC15)

Interactive concept

See the mock-up for this use case

A static, on-brand design concept — illustrative data, not a live system.

Open the UC18 concept mock-up →

Show

Monday, the agent flags Tuesday dinner trending 40% empty. It drafts a targeted offer to the top 50 regulars (from the UC6 CDP) — "chef's table, two glasses included, Tuesday only" — segmented by past spend and preferences, sends via the UC15 channels and tracks conversions. By Tuesday morning, 9 are booked. Break it: it tries to over-discount loyal full-price guests and the margin guard catches it.

Why

The out-of-box moment. Most AI demos predict; this one acts on the prediction to move revenue. It shows the compounding value of stacking use cases — sell the system, not the point tool.

20. Compliance / allergen-integrity guardian

Quality, trust & compliance · Build (with UC12)

Interactive concept

See the mock-up for this use case

A static, on-brand design concept — illustrative data, not a live system.

Open the UC20 concept mock-up →

Show

The agent continuously checks menu, recipe and allergen data for contradictions — a dish listed "gluten-free" whose recipe includes soy sauce gets flagged. It logs food-safety checks (temperature, cleaning) with timestamps and produces an audit trail. When a guest-facing agent (UC10 / UC13) answers an allergen question, the guardian verifies the answer against source. Break it: inject a contradictory menu edit and watch it caught before it reaches a guest.

Why

Bridges hospitality to the AI-quality compliance-risk narrative. Allergen errors are a real legal and safety risk; an assurance layer is something no generic AI vendor offers, and it reinforces why the quality underneath is hard to copy.

21. Floor-plan & covers optimiser

Back office & finance · Build (on UC1/UC2 booking data)

Interactive concept

See the mock-up for this use case

A static, on-brand design concept — illustrative data, not a live system.

Open the UC21 concept mock-up →

Show

Feed it the room dimensions and a year of bookings. It notices 58% of parties are twos while 85% of the floor is four-tops — when the book says full, 14 chairs are physically empty. It proposes three layouts — as built, max-covers, max-revenue — each scored on covers per service, $ per head and revenue per square metre, then turns the winner into a capital case: +SGD 289k a year supports a SGD 68k refit with a 4.3-month payback. Break it: ask it to pack the room to 60 seats and watch the SCDF fire-egress and BCA accessibility clearances hold.

Why

Most rooms are laid out once at fit-out, on instinct, and never re-tested against the booking mix actually walking in the door. Scoring layouts in dollars per head turns the set-up / install budget from a gut call into a calculated investment with a payback date — a natural conversation at fit-out, refurb or new-site moments, and it monetises data the booking agents already collect (including the parties turned away).

19. Members-club / boutique concierge

Guest experience & language · Build

Interactive concept

See the mock-up for this use case

A static, on-brand design concept — illustrative data, not a live system.

Open the UC19 concept mock-up →

Show

A member messages on WhatsApp: "table for 4 Saturday, the usual table, and can you sort a cake for Mei's birthday?" The concierge books across the group's venues, recalls "the usual", arranges the cake and briefs the floor team pre-arrival with a golden-profile card (preferences, allergies, dates, past orders). It handles bespoke / off-menu requests and identifies as AI when asked. Break it: a request outside policy escalates to a human host gracefully.

Why

The luxury / high-ACV segment — fewer, larger engagements with higher willingness to pay. New members clubs opening in Singapore make ready anchors. A "Guest Agent" concept taken upmarket.