For restaurant owners, operators and GMs
Catch every leak before the next service
Ask Tiersel out loud from the office, the pass or your phone. It reads last night's sales, the register's voids and comps, the schedule and the kitchen rail, then prepares the fix: a second PIN on a server's voids, a text to cover a call-out, an 86 on every screen. Nothing changes until a manager approves it.
Live demo on synthetic data · Built on Jev by TypeSafe AI
An animated example types requests as if they were spoken and shows the card Tiersel builds for each one.
For the Owner or operator of an independent or multi-unit restaurant group, with GMs and kitchen managers as daily users
You find out about the leak a week after it happened.
The voids that cluster on one server's Friday nights, the drawer that comes up short on the same closes, the vendor who adds a few cents a pound every delivery: it is all in the register and the invoices, and nobody has time to read them. Meanwhile the GM spends the afternoon texting around for a call-out, the salmon is still orderable online after the kitchen ran out, and the pickup shelf fills with food going cold. Labor already takes about a third of sales, so every hour and every comp shows up in the margin. The facts to fix each of these exist before the next service. Pulling them together by hand is the part that never happens.
6 agent workflows
Built for the work your team already does.
Each workflow starts from a sentence. The agent calls your systems, shows its evidence, and waits for a person before anything consequential happens. Every card below is the real, working interface.
Workflow 01 · Owner or operator
Loss watch on the register
01 · Trigger
The owner asks
From the office or a phone, the owner asks whether anyone is stealing, or a nightly check finds a new pattern.
02 · Tool call
Read the register
Pulls every check, void, comp, refund and no-sale for 14 days with the manager PIN on each, plus drawer counts and close-out times.
03 · Tool call
Compare like with like
Lines each person up against everyone who worked the same dayparts, and reads invoice unit costs by vendor for price creep.
04 · Reasoning
Rank the cases
Ranks void spikes, after-close cash rings, short drawers, comp patterns and vendor increases by dollar impact, each with its evidence file.
05 · Human approval
Owner approves a control
The owner opens the evidence and approves a controlled change, such as a second manager PIN on one server's voids for 14 days.
Nothing happens until a person approves06 · Outcome
Caught the same week
The pattern is on the table with the checks attached, and the control and its approval are in the audit trail.
What changes
The owner sees who is voiding at a multiple of the team rate, with the checks and PINs behind it, and a proportionate control ready to approve.
Workflow 02 · Owner or GM
Owner brief after close
01 · Trigger
The morning after
The owner asks how the night went, or the brief arrives after close.
02 · Tool call
Close the books
Reads gross to net sales, tax, tips, covers, check average and the dine-in, pickup and delivery split from the register.
03 · Tool call
Labor and the menu
Reads time clock hours against sales and ranks dishes by sales and food cost, and servers by net sales and check average.
04 · Reasoning
Look ahead
Projects the next 14 days from your own weekday history and checks next week's schedule against it.
05 · Human approval
Approve the changes
Suggested changes, like cutting one server on a slow lunch or raising a prep par after an 86, are prepared for the owner to approve.
Nothing happens until a person approves06 · Outcome
One page, then act
The owner reads the night in a minute and the approved changes go to the GM to publish.
What changes
Every morning starts with net sales, labor percent and next week's staffing checked against the forecast, and the fixes ready to approve.
Workflow 03 · GM or shift manager
Cover a call-out
01 · Trigger
Someone calls out
A server texts that she can't make her shift. The manager says so out loud.
02 · Tool call
Read tonight
Reads tonight's schedule by station and the hourly cover forecast to see where the gap hurts.
03 · Tool call
Check who is eligible
Checks availability, time off, station skills and hours this week for everyone who could take it.
04 · Reasoning
Rank the picks
Ranks them by availability, overtime risk and reliability, and shows labor percent with each choice.
05 · Human approval
Manager approves the text
A WhatsApp or SMS message to the top picks is drafted. It sends only when the manager approves.
Nothing happens until a person approves06 · Outcome
First yes takes it
The first person to reply gets the shift and the schedule updates, with nobody else contacted.
What changes
A call-out gets covered from one sentence, with overtime and labor percent visible before anyone is asked.
Workflow 04 · Kitchen manager or GM
86 an item everywhere
01 · Trigger
The kitchen runs out
The kitchen manager says to 86 the salmon, or asks what tonight's special should be.
02 · Tool call
Check stock and usage
Reads what is on hand against par, every dish and modifier that uses it, and the open tickets and online orders that include it.
03 · Tool call
Find what is expiring
Reads ingredients that expire within three days and the recipe cost of dishes that could use them.
04 · Reasoning
Build the fix
Suggests substitutes guests can be offered, and a special with its food cost, price and expected sales.
05 · Human approval
Approve per channel
The manager chooses which channels to mark unavailable, from the register to the menu boards, and approves the special.
Nothing happens until a person approves06 · Outcome
Every screen agrees
Guests stop ordering what is gone, and stock that was headed for the bin goes out as a special.
What changes
An 86 reaches every channel at once, and expiring stock becomes a costed special instead of waste.
Workflow 05 · Expo or kitchen manager
Unstick the kitchen
01 · Trigger
Tickets start dying
Expo asks what is holding up the kitchen, or which tickets are late.
02 · Tool call
Read the rail
Reads every open ticket with its fire time, promised time and station from the kitchen display.
03 · Tool call
Measure each station
Compares each station's recent times with its target and counts what is all day.
04 · Reasoning
Name the bottleneck
Finds the station that is behind, the tables past their promised time, and who on the line could help.
05 · Human approval
Manager PIN for comps
A comp for a late table or a void is prepared with the reason. It needs a manager PIN and an approval.
Nothing happens until a person approves06 · Outcome
Recover the table
The late table is looked after before the guest asks, and the comp is logged with who approved it.
What changes
Expo sees the bottleneck and the late tables in one view, and every comp or void carries a reason and a manager's PIN.
Workflow 06 · Front of house or counter lead
Keep the pickup shelf moving
01 · Trigger
The counter gets busy
Someone at the counter asks which pickup orders are waiting.
02 · Tool call
Read every order
Reads pickup and curbside orders with status, quoted time and kitchen progress.
03 · Tool call
See who has arrived
Reads curbside check-ins with the car and spot, and how long ready orders have sat on the shelf.
04 · Reasoning
Spot the misses
Finds orders that will miss their quote and food sitting long enough to lose quality.
05 · Human approval
Approve the texts
New-time texts to late guests, and a longer quote for new orders, are prepared for the lead to approve.
Nothing happens until a person approves06 · Outcome
Guests hear first
Guests get a new time before they arrive to a cold order, and the change is logged.
What changes
Late pickups get a text before the guest walks in, and nothing sits on the shelf unnoticed.
How it plugs in
An agent on your systems, one component per workflow.
Requests come in by voice, text or system event. The router picks the workflow, the agent calls your systems, and the answer lands as a live card where your team works. Nothing is written back until a person approves it.
In
Tiersel
Your systems
Out
Workflow blueprint · Owner or operator
Loss watch on the register
- Starts when
- The owner asks. From the office or a phone, the owner asks whether anyone is stealing, or a nightly check finds a new pattern.
- Reads
- Read the register. Pulls every check, void, comp, refund and no-sale for 14 days with the manager PIN on each, plus drawer counts and close-out times.
- Compare like with like. Lines each person up against everyone who worked the same dayparts, and reads invoice unit costs by vendor for price creep.
- Decides
- Rank the cases. Ranks void spikes, after-close cash rings, short drawers, comp patterns and vendor increases by dollar impact, each with its evidence file.
- Waits for
- Owner approves a control. The owner opens the evidence and approves a controlled change, such as a second manager PIN on one server's voids for 14 days.
- Writes back
- Caught the same week. The pattern is on the table with the checks attached, and the control and its approval are in the audit trail.
This workflow is one component
export const losswatch = defineSkill({id: "losswatch", category: "restaurant",// when this workflow, and not a neighbour, is meantcriterion: "A restaurant owner or GM asks whether staff are stealing or where money is leaking: void spikes, after-close cash rings, short drawers, comp patterns, vendor price creep. Not home security, a stolen package or a personal card fraud alert.",examples: ["is anyone stealing from me", "who is voiding the most this week"],match: (text) => score, // intent score, runs on every wordparse: (text, ctx) => record, // entities + tool calls into your systemsreply: (record) => "…",Card: LossWatchCard, // the live UI, with its approval step});
The id, criterion, examples and reply are read live from the skill that renders the card above. Adding a workflow for your team means adding one of these: its intent, the calls it makes, what it says, and the interface your people approve in.
Fits your stack, answers to your people
Reads your systems. Acts only with approval.
What the agent connects to
Point of sale and payments
Online ordering and delivery
Kitchen display
Scheduling and time clock
Inventory, recipes and purchasing
Staff messaging
Product names are examples of systems in this category. Each deployment is connected to your systems through their APIs as part of the build.
A manager approves every change
Voids, comps, 86s, specials, schedule changes and texts to staff or guests are prepared by the agent and carried out only after a named person approves. Voids and comps also need a manager PIN.
Every step is logged
Each read, the evidence it returned and each approval or rejection are recorded, so you can see who approved what and why.
Synthetic data in this demo
Every restaurant, employee, guest, check and number on this page is invented. Nothing here touches a real register.
Each role sees its own work
When deployed, access follows the role: a server can ask about their section, a kitchen manager about stock and the rail, and only owners see loss cases and the full books. Your data stays in your environment.
The cost of doing it by hand
Put your own numbers in.
Move the sliders to match your operation. Nothing is sent anywhere unless you choose to email it.
Each year
384
staff hours back, about 0.2 full-time people
$16,512
of staff time, at your hourly cost
Email me a pilot plan for these numbersCounts only manager time spent pulling reports, covering call-outs, updating menus by channel and reviewing voids and comps. It does not count money recovered from loss or waste, which depends on your own numbers. The hourly default starts from the BLS median wage for food service managers, $33.36 an hour in May 2025, plus about 30 percent for payroll taxes and benefits.
Hours = shifts a month × 12 × minutes each × share, ÷ 60. A full-time person is 2,080 hours a year. Default hourly cost from U.S. Bureau of Labor Statistics, Occupational Outlook Handbook, food service managers, 2025.
Questions buyers ask
Straight answers.
No. Tiersel is a voice layer over the systems you already run. When deployed, it reads from your register, scheduling, kitchen display and ordering systems through the access you grant. The demo on this page uses synthetic data and connects to nothing.
No. It prepares the void, comp, 86, special or schedule change and shows why. Nothing happens until a manager approves, and voids and comps also need a manager PIN. Each approval is logged.
No. A case says a pattern stands out against people who worked the same dayparts, and shows the checks behind it. You read the evidence and decide whether it is a conversation, a control or nothing. The suggested controls are temporary and reversible.
When deployed, it stays in your environment and each request reads only what the question needs. Staff see what their role allows. In this demo there is no real data at all.
It is built for moments when hands are busy: a GM at the pass, expo on the rail, a lead at the pickup counter. Each answer is one sentence spoken back plus a card to tap, so nobody has to dig through reports mid-service.
Try the live demo on this page now. A pilot starts at one location, read-only at first, so you can compare its answers with your own reports before anything writes back.
See it on your own workflow.
Watch the restaurants workflows run end to end, or speak to the agent yourself. Then tell us which one to build on your data.
Scope a pilot with hello@sasid.ai- 01
Pick one workflow
A short call to choose the workflow that costs you the most time, and who approves its actions.
- 02
See it on your data
We build that workflow against your sample or de-identified data so your team judges the real thing.
- 03
Run it with people in charge
Your staff use it on live work. The agent drafts and prepares; a person approves every consequential step.
- 04
Decide from results
You keep what measurably saves time and drop what does not.
Other operations
