The short version
Restaurants don't usually die of one wound. They die of accumulation: the booking call that rang out on a Friday, the deposit that was never paid, the supplier increase that quietly repriced the menu downward for a whole quarter. Each one is small. Fifty-two weeks of each one is not.
Resk (formerly Elyx) is an AI operator for restaurants, built in Ireland. It answers the calls, protects the covers, chases the money that already belongs to you, watches the margin, and hands the owner one short brief every morning with the moves already prepared. The name is the mission compressed into four letters: restaurant rescue.
The numbers behind that sentence so far: over €400,000 in revenue recovered that would otherwise have been lost to margin leak, missed calls and unpaid payments. Over 1,000 admin hours handed back to owners by the Operator. Over 5,000 minutes of live calls answered by our agents, and thousands more in leaks surfaced before they slipped through the cracks.
This is the story of why we're building it, why it has to exist now, and what happens when it works.
Chapter I: Why now
Four things converged, and any one of them alone wouldn't have been enough.
1. The cost stack stopped forgiving mistakes
Ask anyone who traded through the last few years. RTÉ Prime Time called it a perfect storm: food costs, energy, wages, interest, all rising at once. When margins were fatter, a rung-out phone or an unchased deposit was an annoyance. Now the macro pressure is heavy enough that the leaks you control decide whether it kills you or merely hurts. The margin for error is gone. That's the real change.
2. A window just opened, and it has a shape
On 1 July 2026, three days before this was published, VAT on food and catering services in Ireland dropped from 13.5% back to 9% under Budget 2026. The Restaurants Association of Ireland campaigned hardest for that cut and welcomed it with a warning attached: the labour cost increases that arrived in January would eat much of the benefit. Policy handed the sector breathing room, temporarily, and the people who fought for it are telling you not to spend it twice. Breathing room is the cheapest time to fix an operation. That window is open right now.
3. AI can finally work a restaurant's actual reality
A restaurant does not run on clean, structured data. It runs on a phone ringing in a loud room, a guest with an accent and a train delay, a supplier invoice that arrives as a PDF or a photo on WhatsApp, a stock note scribbled during prep, a no-show risk buried in a diary. This mess is precisely what broke every previous attempt to automate hospitality operations. Three years ago software genuinely could not handle it. It can now. A voice agent can hear what a caller wants, check the diary against house rules, and hold the table. A model can read the invoice photo and connect the price change to the four dishes it touches. The operational reality that kept restaurants un-automatable is exactly the surface modern AI handles natively.
4. The guest rewrote the rules
Guests book later and cancel with less guilt. Nobody under sixty leaves a voicemail. A caller whose call rings out at 7:40pm books somewhere else from the same phone in under a minute. Deposits went from confrontational to normal. QR ordering went from gimmick to expected. The demand is still there. It just stopped waiting.
Nobody is building for the independent
Here's the part that decided it for us. The big chains have ops teams, group buying power and internal systems. The delivery platforms build brilliant technology, but they build it to own the guest relationship, the data and the commission. Nobody has any incentive to arm the independent restaurant, the one that anchors a town square or a village street, with the same operating capability. Systems that record are everywhere: the POS records sales, the booking system records bookings, the accounts record, months later, what it all meant. All of them assume a human will read the records and act. That human is on the pass, or fixing the dishwasher, or doing the eleventh hour of an eleven-hour day.
The independent restaurant needed an operator. So that's what we built.
Chapter II: The state of things
When a restaurant anchors a street, other business follows. Foot traffic. Evening trade for the shops either side. A reason the town centre is still lit at 9pm. Restaurants are load-bearing institutions in Irish towns, and the RAI's April 2024 report on closures put formal numbers on what it costs a community when one goes dark.
The survival numbers deserve honest handling, because folklore has done real damage here. The line that 90% of restaurants fail in their first year is a myth, and Ohio State University research went and found the real figure: roughly 59% fail within their first three years, around 61% for independents. Two things are true at once. The doom number people quote at owners is wrong, and the corrected number is still brutal. Six in ten gone inside three years is not an industry with room to leak.
Sort the research-backed closure drivers into two piles and the strategy writes itself. Macro pile: rent, energy, wages, VAT, interest rates. You don't get a vote. Operational pile: missed demand, unpaid revenue, margin drift, stock pressure, labour mismatch, cashflow blindness, and the owner burnout that comes from being the only person watching all of it. That pile is yours, and it's the pile that decides which side of the 59% you land on.
The gap between how much these rooms matter and the quality of the tools they run on is the opportunity. Not because owners bought bad software. Because everything they bought was built to record, and nobody was responsible for acting.
Chapter III: The engine
Resk is one platform where restaurants previously ran five or more disconnected systems: the phone, the booking diary, the payments tab, the ordering site, the POS, the stock spreadsheet, and the owner's memory holding it all together. Three systems in one, and the third is the one that matters.
The system of record
Every booking, call, order, payment, stock count, supplier price and guest preference in one place. POS and till hardware, branded ordering across web, QR, NFC and kiosk, the diary, the guest book. One source of operational reality instead of fragments.
The system of action
This is the Operator, and it's the reason the company exists. It doesn't produce another report. It notices, drafts and chases. The voice agent answers the call that lands mid-rush, hears intent, checks your rules and books the table. The payment watch spots the deposit that was requested and never paid, and the link that failed silently at 11pm, and prepares the chase. The margin watch connects the supplier invoice to the dishes it touches and frames the decision: reprice, renegotiate or re-spec. Everything lands in one morning brief with the actions ready to approve. The owner stays in charge. The owner stops being the integration layer.
The system of transaction
The money itself: deposits, Tap to Pay, checkout links, failed-payment recovery, and the operating evidence that feeds Resk Capital when stock, equipment or cashflow blocks the next move. Recovery isn't a feature bolted onto payments. It's the point.
Why this is hard to build
The hard part has nothing to do with writing code. It's knowing which code to write, and that knowledge only comes from standing in the room. A phone answered mid-service in a loud Irish kitchen is a harder voice problem than a call centre. A booking is a promise with rules attached, not a row in a table. Deposits touch real money and real relationships; chase wrong and you've insulted a regular. Supplier data arrives in whatever format the supplier felt like producing. And every mistake costs a table, a guest, or the owner's trust, which is the one currency you don't get back. This is why the Operator drafts and waits for approval rather than acting behind your back. Autonomy is earned one approved action at a time.
Chapter IV: What happens when it works
The honest answer is: nothing dramatic, which is the point. The phone stops being a coin flip. The Saturday eight-top either shows up or their deposit means the empty table wasn't free. The supplier increase gets caught the week it happens instead of the quarter after. The morning brief takes three minutes with coffee instead of an evening of tab-hopping.
We measure it venue by venue, and we publish the measures we ask every trial to hold us to: how many missed calls became bookings, which deposits and payment links were chased before they went cold, and whether the owner can see what the day needs without opening another dashboard.
Across live venues so far, that adds up to more than €400,000 in recovered revenue: money that had already been earned and was on its way out through margin leak, rung-out calls and payments nobody chased. The voice agent has answered over 5,000 minutes of live calls, during services when nobody had a free hand. The Operator has given owners back over 1,000 hours of admin, the hours that used to live between the payments tab, the diary and the spreadsheet. And beyond what was recovered outright, it has surfaced thousands more in leaks that would otherwise have gone through the cracks unnoticed, which is often the part owners say changed how they run the week.
Chapter V: The economics
We hold a simple line on pricing conversations: if the recovered money doesn't cover the software, you shouldn't buy the software.
That's why everything starts with the free leak audit. Five quick answers. One clear number: what the leaks you control are probably costing you. If the number is small, you've lost three minutes and gained a compliment on running a tight ship. If it isn't, a 30-day trial exists to prove the number in your venue, on your phone line, with your guests, against measures you write down before day one.
And the claim stays narrow on purpose. Resk does not fix rent. It does not fix energy, the VAT rate or the wage floor, and any vendor implying its software does is lying to you at a vulnerable moment. Resk catches the money you already earned and were losing anyway.
Chapter VI: What comes next
One address, first of all. Everything now lives at resk.ai; elyx.ie and getelyx.ai keep working and will redirect in time. A boring migration, deliberately. Restaurants have enough moving parts without their software supplier turning a rename into an event.
Deeper autonomy, earned gradually: more of the chase, the reorder and the reprice moving from "drafted for approval" to "done, with a receipt", at whatever pace each owner sets. Resk Capital growing into the funding path for operators whose numbers deserve better terms than their bank offers. Ireland first, properly, before anywhere else.
Chapter VII: A founder's note
We renamed the company this summer. Elyx was the name we picked back when we thought we were building restaurant software. Somewhere between the first rescued booking call and the €400,000 we've now recovered for owners, the job revealed itself as something sharper: revenue recovery. Once that was the work, the mission belonged in the name. Resk is short for restaurant rescue, and when someone asks what it means, the answer is the pitch.
Rescue, in our experience, is rarely one heroic intervention. It's catching the small failures early enough that the owner still has options. The 7:40 call that becomes a booking instead of a story you never heard. The deposit that gets paid because someone, or something, remembered to ask twice. A restaurant that makes it through year three because a hundred small leaks got closed in year one.
That's the company. That's the whole pitch.
Chapter VIII: Work with us
If you run a restaurant: start with the number. The free leak audit is at resk.ai/leak-audit. Five quick answers. One clear number. If it annoys you, book a demo and hold us to the trial measures above.
If you build things: we're hiring people who want their code to matter in a room with fire and knives in it. Get in touch through resk.ai.
And if you've recommended us under the old name, thank you. It's easier to say now. Four letters. Resk.