Headcount-accurate billing, voucher and subsidy logic built in, per-site performance on one screen, and a group close that takes days, not weeks — 16 institutional solutions, narrated by Iris.
Corporate cafeterias, hospitals, campuses: thousands of meals, subsidy rules, headcount billing, forty sites on forty spreadsheets.
Cost per meal — the number the contract was won on, measured per site.
Participation — how many of the eligible population actually turned up.
Compliance — nutrition standards and food safety, evidenced rather than asserted.
Contract margin — what each agreement earns against the model it was bid on.
Zentallio consolidates it — so forty sites stop being forty spreadsheets and become one board.
Headcount-accurate billing — what the client is invoiced matches who was actually fed, defensibly.
Voucher and subsidy logic built in, so who eats free, who pays partial and who pays full is a rule rather than a judgement.
Per-site performance on one screen, and a group close that takes days, not weeks.
Cost per meal, participation, compliance and contract margin — live, with Iris naming the site drifting from the model you bid.
A serving line where the price depends on who is standing at it — entitlement, subsidy and cost centre resolved at the moment of service, without slowing a lunch rush of thousands.
Identity resolves the price. Badge, card or app determines entitlement and subsidy tier before the tray reaches the till.
Every meal carries its cost centre. Department, contract and site attach at the line, so recharge never needs reconstructing.
Built for a lunch peak. Thousands of covers inside one hour means throughput is a design constraint, not a nice-to-have.
Cost per meal, participation, compliance and contract margin on one live board — with Iris naming the site drifting from the model you bid.
Per site and per contract. Forty sites resolve into one comparable view rather than forty differently-built spreadsheets.
Measured against the bid. The tender model is held in the system, so drift is measured against the promise rather than against last month.
Iris names the site. Not that cost per meal rose — which site, which driver, and how far from the bid model it now sits.
P&L, balance sheet, cash and close on a single balanced ledger across every site and every contract — a group close that is a review rather than a rebuild.
Consolidation is automatic. Site ledgers roll into a group position without a month of manual mapping.
Client billing posts itself. Invoices raised against contract terms flow into receivables already coded to the right account.
Ask Iris for any number. Any site, any contract, any period — answered in the conversation, not a week after close.
Demand-planned replenishment from central production out to every site — with tendered supplier agreements, contracted pricing and compliance carried through the whole chain.
Central kitchen and site demand as one plan. Production and distribution are planned together rather than each site ordering alone.
Contracted pricing enforced. A tendered rate is what gets paid, and off-contract purchasing is visible rather than absorbed.
Supplier compliance travels with the goods. Certification and audit status are attached to the supplier, not filed separately.
Digital checklists, shift routines and standard operating procedures across every site — completion tracked live, exceptions escalated, evidence generated as a by-product.
HACCP routines, signed off. Temperature checks and safety routines are completed in the system, producing audit evidence automatically.
One standard across the estate. A new site inherits the current procedure set rather than a copy of whatever the last one was doing.
Exceptions escalate themselves. The routine that didn't complete surfaces to a human; the rest stays quiet.
Rostering, attendance and labour percentage measured against forecast demand — in a sector where labour is usually the largest line in the bid model.
Rostered to forecast participation. Site cover is built from expected covers rather than from an establishment figure set at mobilisation.
Measured against the bid model. Labour is compared to what the tender assumed, so drift is visible while the contract can still be managed.
Compliance is a guardrail. Working-time rules and site-specific agreements are enforced in the roster, not audited afterwards.
Deterministic logic under the hood (the “Zen Rules” layer) — contracts, entitlements and safety evidence. In institutional food service these are not preferences: they are the agreement you signed and the inspection you will face. Four solutions run on this layer.
A department transfer left 40 staff mapped to two subsidy tiers at once. Meals are being served at the lower rate. Flagged before the billing run, not after the client queries it.
The agreement itself, held as executable rules rather than as a PDF in a folder — so what gets invoiced is what was signed, every cycle, without interpretation.
Contract terms are configured as rules — management fee, cost-plus, fixed price, minimum guarantee, indexation — and the billing run executes them.
Indexation and price reviews fire on schedule, so an uplift is never missed because nobody diarised it.
Each invoice is traceable to the meals, headcounts and terms behind it, which is what settles a client query in one call.
Contract expiry, renewal windows and break clauses are surfaced ahead of time rather than discovered at renewal.
Variations are versioned, so the terms in force on any past date can be reconstructed exactly.
Who eats free, who pays a subsidised rate and who pays full — determined by rule rather than by the judgement of whoever is on the till at the time.
Entitlement is derived from role, department, shift, site or scheme membership, and resolved automatically at the point of service.
Free-meal eligibility — free school meals, clinical diets, night-shift provision — is applied without the recipient having to identify themselves publicly.
Daily and periodic caps are enforced, so an allowance cannot be spent twice or drifted past unnoticed.
Conflicting entitlements resolve by a defined precedence rather than by whichever rule the system happened to check first.
Every entitlement decision is logged, which is what makes a subsidy claim or a client audit answerable.
Feeding hospitals and schools means the audit is not an event, it is a permanent condition. Evidence is generated as a by-product of doing the work rather than assembled before an inspection.
Critical control points are monitored continuously, with temperature and holding records logged automatically rather than on a clipboard.
A breach raises immediately with the batch and the site named, while the food is still recoverable.
Corrective actions are recorded against the incident that triggered them, which is the part most audits actually fail on.
The audit pack is generated on demand for any site and any period — a lookup rather than a fortnight of preparation.
Staff training and certification currency are held against the roster, so an unqualified person is never scheduled to a controlled task.
In this sector the integrations are the client's systems, not yours — HR and payroll for entitlement, the client's ERP for invoicing, and access control for headcount. All configured per contract.
Client HR and payroll feeds drive entitlement, so a leaver stops eating on subsidy the day they leave rather than a month later.
Invoices post into the client's procurement or ERP system in the format that client requires, rather than as a PDF someone rekeys.
Access control and badge systems supply an independent headcount signal to corroborate what the till saw.
Tax and reporting rules are configured per market, which is why a new country or public-sector framework takes days.
Every connection is health-checked continuously — a stale HR feed silently corrupting entitlement is exactly the failure this catches.
Forecasting, regression, optimisation and anomaly models under the hood (the “Zen Models” layer) — because almost every question in this sector reduces to one: how many people will actually eat, and what does that cost. Twelve solutions run on this layer.
The baseline model for a site — how many of an eligible population actually turn up, learned per institution, because a hospital, a campus and a corporate office behave nothing alike.
Participation rate is modelled per site rather than assumed from an estate average that fits nobody.
Institution-specific rhythms are learned — term dates, ward rotas, shift patterns, hybrid-working days, exam weeks.
A new site inherits a forecast from comparable institutions until it has enough history of its own.
Forecast accuracy is tracked per site, so a persistently wrong site is corrected rather than quietly over-catered.
The same forecast feeds production, procurement and rostering, so the three cannot disagree.
The same-day correction. Booked is not served — and the gap between them, applied late enough to be accurate and early enough to matter, is where over-production is prevented.
Expected no-show is predicted per site and per service, so the production number is adjusted rather than the booking taken at face value.
Same-day signals — badge-ins, weather, calendar events, ward occupancy — revise the number while the kitchen can still act on it.
The adjustment is applied at a cut-off tuned per site, late enough to be accurate and early enough to change the cook.
Systematic over-booking by a department is surfaced as a pattern, which is a conversation with the client rather than a kitchen problem.
Actual versus booked is retained per service, which is the evidence behind any billing discussion about headcount.
The production side of headcount — turning an expected number of diners into how much of each dish to actually cook, and closing the loop with what came back uneaten.
Forecast covers become batch quantities per dish, accounting for take-up split across a cycle menu rather than one flat number.
Batch cooking is staged through service, so the last hour is not served from food cooked for the first.
Plate waste and counter waste are recorded separately, because over-production and over-portioning need different fixes.
Waste by dish and service flows back into the forecast, correcting a repeatedly over-produced dish rather than repeating it.
Waste is costed and shown against the contract's food-cost model, which is what makes it a commercial conversation.
The commercial side of the same number. What the client is invoiced for depends on how headcount is defined in the contract — and that definition is where most disputes in this sector actually start.
The billable headcount is derived exactly as the contract defines it — served, booked, eligible or attended — not as whichever number is easiest to produce.
Minimum guarantees and volume bands are applied automatically, including the shortfall charge when a site under-delivers.
Every billed figure is traceable to the transactions behind it, which turns a client dispute into a lookup.
An independent signal — badge or access data — corroborates the till count where the contract requires it.
Variance between forecast, served and billed headcount is monitored, since a persistent gap is either a pricing problem or a data one.
The invoice run itself — consolidated across sites, split by cost centre, and checked for the anomalies that would otherwise turn into a query, a credit note and a delayed payment.
Charges consolidate across every site on an account, then split by department and cost centre as the client requires.
Anomaly detection reviews the run before it goes out — a site billing double last month's figure is caught internally, not by the client.
Credit notes and adjustments carry a reason code and an audit trail rather than appearing as unexplained corrections.
Ageing and disputes are tracked per account, so a query held up by one line item is visible rather than silently unpaid.
Each invoice arrives with its own backing detail, which is what removes the fortnight of email that usually follows.
The transaction where entitlement becomes a price. Subsidy applied, voucher redeemed, employer share and employee share separated — in the second it takes to serve a tray.
Subsidy is applied at the point of service, with the employer share and the diner share separated on the same transaction.
Vouchers and allowances redeem by badge, QR or app, and a spent allowance simply cannot be spent again.
Classification handles the ambiguous cases — a contractor, a visitor, a transferred employee — rather than defaulting everyone to full price or free.
The employer-funded portion accrues per department, so recharge is computable from the first day of a period.
Anomalous redemption patterns are flagged, since a shared badge is the most common leak in a subsidised scheme.
Hospitality and event orders captured properly — the part of an institutional contract with the best margin and the worst record-keeping, usually run on email and a shared calendar.
Requests are captured with room, headcount, timing, dietary needs, service style and the cost centre to bill.
Confirmed events enter the production and staffing plan rather than sitting in a mailbox until the morning of.
Late changes and cancellations follow contract cut-offs, so a room cancelled two hours out is charged as the agreement says.
Event demand is forecast from history, so recurring board meetings and term-time patterns are anticipated rather than reacted to.
Event margin is reported separately from core catering, because the two behave completely differently.
Cycle menus that satisfy a nutrition standard, a budget and a client specification at the same time — validated before publication rather than defended after an inspection.
Multi-week cycle menus are built with variety rules, so the same dish does not reappear in a way diners notice and resent.
Nutrition standards are validated at menu level before publication — school standards, hospital diet codes, corporate wellness targets.
Cost per meal is computed for the cycle as a whole, so a menu that breaches the contract's food-cost model is caught in planning.
Therapeutic and texture-modified diets are held as first-class variants rather than as an exception handled off-system.
Compliance evidence for the published cycle is generated automatically, ready for a client or regulator to review.
Making the same standard happen in forty kitchens — what is produced centrally, what is finished on site, and where execution is quietly diverging from the specification.
The make-or-buy split is optimised per dish and per site, balancing central production capacity against distribution cost.
Central production is scheduled against aggregated site demand rather than against each site ordering independently.
Anomaly detection surfaces sites drifting from the standard — portion, cost or method — before it shows up in an audit or a client complaint.
A new site inherits the current standard on day one rather than after a mobilisation cycle of local improvisation.
Local exceptions require sign-off and are reviewed, so genuine variation is recorded and drift is not.
Buying at the price you tendered, from suppliers who still hold the certifications they were awarded on — with off-contract spend visible rather than absorbed into food cost.
Tendered prices are held as the expected cost, and any invoice above the agreed rate is flagged rather than paid quietly.
Off-contract and maverick purchasing is surfaced per site, which is usually where a food-cost model first breaks.
Supplier certifications, insurance and audit status are tracked with expiry dates, and a lapse blocks new orders.
Volume commitments and rebate thresholds are projected, so a rebate is earned deliberately rather than missed by a margin.
Price movements on tendered lines are forecast, so a contract renegotiation is prepared for rather than reacted to.
Your own view of the estate — forty sites made comparable, with the outliers surfaced rather than buried in an average that describes none of them.
Every site reports on the same definitions, so cost per meal at one site genuinely means the same as at another.
Anomaly detection surfaces the handful of sites that need attention rather than producing a forty-page pack nobody reads.
Like-for-like comparison accounts for site type and size, so a hospital is not benchmarked against a corporate office.
Roll-ups follow the operating structure — region, sector, contract — without a separate reporting build for each.
The same numbers feed the group close, so operational reporting and finance never disagree about what happened.
The client-facing half — the SLA pack, the satisfaction data and the evidence that goes into a retender. Contracts in this sector are lost at review meetings, not at the serving counter.
The client's own KPIs and SLA definitions are configured per contract, so the pack answers their questions rather than yours.
SLA breaches are visible as they accrue, with time to remedy before they become a penalty or a review-meeting ambush.
Diner satisfaction is captured at the site and segmented by population, so a problem is located rather than generalised.
The review pack is generated rather than assembled, which is days of account-management time returned every month.
Retender evidence accumulates continuously, so a renewal is argued from a record rather than from the last quarter's memory.
Not a screenshot — the actual agent. No solution in this sector is filed under L3, because none needs to be: Iris queries all sixteen, across every site and every contract, in plain language.
Sixteen solutions. Every card opens a live guided demo.
You configure a sector playbook, not a custom project. Iris applies it herself — agentically, from day one.
Onboarding is agentic. Iris connects the tills, the client's HR feed and the procurement system directly — no manual data mapping.
Mobilisation is configuration. A newly won site inherits the estate standard and its contract terms on day one, rather than improvising for a quarter.
A new framework takes days. Tax, reporting and public-sector rules are configured per market rather than coded per customer.
The first line of support is agentic — Iris resolves most of it herself. Our engineers pick up from there.
Layer 1 — Iris, 24/7. Configuration questions, anomalies and routine issues resolved directly, instantly.
Layer 2 — our engineers. Anything Iris can't close escalates automatically to a Zentallio engineer.
No blank tickets. Every escalation arrives with Iris's own diagnosis — engineers start from an answer.
“Corporate cafeterias, hospitals, campuses: thousands of meals, subsidy rules, headcount billing, forty sites on forty spreadsheets. Scale here has historically meant blindness.”
Zentallio consolidates it: headcount-accurate billing, voucher and subsidy logic built in, per-site performance on one screen, and a group close that takes days, not weeks. Underneath sits the same platform every sector runs — one data spine from the till to the ledger, three layers of intelligence, and an agent that narrates every screen, flags what needs a human, and acts on the rest.
Headcount billed on a number nobody can trace, so every invoice is a negotiation.
Subsidy applied by judgement at the till, and an entitlement rule nobody has read since mobilisation.
One site drifting from the bid model for two quarters, invisible inside an estate average.
Contract terms, entitlement and HACCP evidence are executed as rules, not interpreted per site.
Headcount is forecast, corrected same-day, costed and billed — four models on one number.
“Which site is drifting from the model we bid?” — answered across every site and contract in one question.
Every Zentallio sector runs the same three-layer intelligence, tuned to how that format actually loses margin.