Where the two hours after the shift actually go
Ask a hotel duty manager or a restaurant general manager what the end of a shift actually looks like, and it rarely matches the job description. The floor closes, the last cover leaves, and a fourteen-hour day that started before breakfast service does not end — it changes shape. The laptop opens. The property management system gets exported for tonight's occupancy and tomorrow's arrivals. The point of sale gets pulled for covers, revenue by daypart, and voids that need explaining. The booking platform gets checked for the next seven days of pickup. None of those three systems talks to the other two, so somebody — usually the person who has been on their feet since the morning briefing — copies the numbers into one document by hand, because ownership expects the pack in their inbox before the next shift starts.
That is one task out of several that quietly land on the same person at the same hour. Tomorrow's covers need translating into a supplier order before the kitchen opens, which means checking par levels against what actually got used tonight, not what the delivery note from three days ago assumed. The rota for next week needs checking against who actually clocked in this week, because a schedule and a timesheet drift apart within days if nobody reconciles them, and payroll is due whether or not that reconciliation happened. A guest who left contact details at check-in is sitting in the property management system, unreachable by the marketing platform that would otherwise send them a reason to come back, because nobody moved the record and nobody has time to. And if this is one property in a small group, every one of those numbers packs is arriving from a different manager, in a different format, some time after ownership actually needed them, because building a companywide picture out of several separate spreadsheets is itself a task nobody has two extra hours for.
None of this is a staffing failure. It's what happens when the systems a hospitality business runs on — the PMS, the POS, the booking platform, the rota tool, the supplier portal — were never built to talk to each other, so a person becomes the connection between them, every single day, on top of an already full shift.
What the system does, every single day
Workflow automation for hospitality does not run the front desk or answer the phone — that is what voice AI built for hotels and restaurants does. This is the layer underneath: the fixed pipelines that move numbers, orders, and records between your systems on a schedule, without anyone needing to remember to run them.
- The daily numbers pack. Occupancy and ADR from the PMS, covers and revenue by daypart from the POS, and pickup for the coming week from the booking platform arrive as one report, assembled overnight and waiting before the first shift starts — not three exports someone stitches together by hand. Automated report generation covers the mechanics behind pulling numbers out of several systems on a schedule.
- Supplier ordering built from actual covers. Tomorrow's forecast covers get checked against current stock and par levels for the ingredients that matter, and a purchase order gets drafted — or sent outright, where a chef has approved that step — before the kitchen opens, instead of a weekly order built on habit rather than what the booking calendar actually shows. Automated inventory alerts covers the stock-level side of this pipeline in more depth.
- Rota and payroll preparation. Scheduled hours from the rota tool get reconciled against actual clock-in and clock-out data, exceptions get flagged — a no-show, a shift that ran long, a swap nobody logged — and a payroll-ready record gets prepared before the cutoff, instead of someone cross-checking two systems by hand the night before it's due. Automated shift scheduling covers this pairing of rota and attendance data in more depth.
- Guest data reaching the marketing system, with consent respected. A guest's contact details and stay history move from the PMS into the marketing or CRM platform as a matter of course, tagged with whatever consent they actually gave at check-in or booking — so a returning-guest campaign has real, current data behind it instead of a list someone last exported for a mailer a year ago.
- Allergen and menu data kept in sync everywhere it appears. A dish description, its price, and its allergen information update on the printed menu, the website, and every delivery platform the same day a recipe changes, instead of drifting apart until a guest orders something the menu no longer matches.
- Property-level numbers rolling up without an inbox full of spreadsheets. For a group running more than one site, each property's occupancy, covers, and labor numbers feed a single rolled-up view automatically, so ownership sees the group picture without every general manager emailing their own version of the same report.
Three pipelines, start to finish
The clearest way to see what this looks like is to walk through three of these pipelines the way the system actually runs them.
The numbers pack that's already in the inbox by the time the shift starts
A busy full-service property closes out the night audit. As soon as the PMS finalizes the day, the system pulls occupancy, ADR, and tomorrow's arrivals straight from it. It does the same with the POS for covers, revenue by outlet, and any voids over a set threshold that need a note attached, and with the booking platform for the coming week's pickup against the same period last month. All three get merged into a single report, laid out the way the general manager actually reads it — not a raw export — and it's sitting in her inbox before she's back on property the next morning. The task that used to cost the outgoing duty manager the better part of an hour at the end of an already long shift now costs nobody any time at all, and the report is more consistent than a hand-built one because it pulls the same fields from the same systems the same way every night.
An order that follows tomorrow's covers, not last week's habit
A restaurant's POS shows a strong forecast for Saturday based on reservations already on the books through the booking platform — well above a typical Saturday. The system checks that forecast against par levels for the ingredients a busy Saturday actually depends on, cross-references it with what's currently in stock, and flags the gap. A draft purchase order goes to the head chef for a quick check rather than a from-scratch build, because the system already did the arithmetic against a real forecast instead of the standing weekly order that assumes every Saturday looks the same. He adjusts one line where a dish has been swapped off the specials board, approves it, and it's with the supplier that afternoon — not two days later, when someone finally gets a free half hour to think about Saturday's stock.
A rota reconciled against reality before payroll is due
A hotel's rota tool shows who was scheduled for the week. The time-and-attendance system shows who actually clocked in and out. On their own, those two records drift apart within days — a shift covered informally, a late clock-in nobody flagged, a no-show that never got marked. The system reconciles the two automatically each day, flags the exceptions that need a manager's eyes rather than burying them in a spreadsheet, and by the time payroll is due, the hours going into it already match what actually happened on the floor, not what the rota said should have happened. The manager who used to spend an evening before every payroll run cross-checking two systems now spends a few minutes clearing flagged exceptions instead.
Guest data and marketing: moving with consent, not around it
A guest who books directly, checks in at the desk, or joins a loyalty scheme has usually agreed to some level of contact — a confirmation, a receipt, sometimes marketing. What breaks down is not the guest's willingness, it's the plumbing: that consent and the guest record it belongs to often live in the PMS, while the tool that would actually send a returning-guest offer or a post-stay survey lives somewhere else entirely, and nobody has built the connection between them. The workflow reads the consent flag along with the record — never the record without it — and only what a guest actually agreed to reaches the marketing platform. A guest who opted into a loyalty scheme but not a marketing list stays exactly that: reachable for the scheme, not for the campaign. Getting this pipeline right matters as much as building it at all, since a marketing system populated by data moved without real consent behind it is a liability wearing the shape of a database.
Allergen and menu data: this is a safety requirement, not tidiness
A menu that says one thing on the paper in front of a guest, another on the website, and a third on a delivery platform is not a branding inconsistency. It's a way for a guest with a genuine allergy to order something the kitchen has already changed without the printed description catching up, and that failure mode has nothing to do with how careful the kitchen actually is on the night. A recipe change — a swapped supplier, a substituted ingredient, a dish pulled from the menu — needs to update the dish description, the price, and the allergen information everywhere a guest might read it, on the same day the change happens in the kitchen, not whenever someone remembers to update the website weeks later. The workflow treats the kitchen's record as the source of truth and pushes every downstream copy — the printed menu, the website, each delivery platform the business lists on — from that one place, so there is exactly one version of the truth instead of three that quietly stop agreeing with each other.
Rolling up numbers across a group without a single emailed spreadsheet
A single property can build its own numbers pack easily enough once the pipeline exists. A group running several properties has the same problem multiplied: every general manager building a version of the same report, in a slightly different format, landing in an owner's inbox at a different hour, and someone at head office reassembling all of it into one picture before a Monday meeting. The same pipeline that assembles one property's numbers can feed a rolled-up view across every site instead — occupancy, covers, and labor cost sitting in one place, broken out by property or combined, without a single site having to email anything. A general manager still sees their own numbers first. Ownership sees the group's, built from the same source data, on the same schedule, without a reconciliation step in between.
Connecting to the PMS, POS, and booking platform you already run
Most modern property management systems, point-of-sale platforms, and booking systems expose an API or a scheduled data feed, and that is usually enough to build these pipelines without disrupting the booking or guest record your team already relies on. Where a system is older or more closed — which is not unusual in this industry, especially at independent properties running software chosen a decade ago — there is generally still a workable way in: a scheduled export, a webhook, or a shared file a supplier or payroll provider already expects. We scope that honestly on the first call, not partway through a build. Calfy does not claim a formal partnership with any specific PMS, POS, or booking platform; every pipeline is built to connect to whatever you already run.
Access is scoped narrowly on purpose — read-only wherever reading is enough, and write access limited to the specific record a pipeline is authorized to update, like a rota's payroll export or a menu's allergen field. None of this depends on replacing software your team already knows. The goal is that the numbers pack, the supplier order, and the rota sit in the systems your team already checks, not a parallel dashboard that becomes a second source of truth nobody fully trusts. For the wider set of systems Calfy builds across this industry, including the guest-facing side, see AI for hotels and restaurants.