OrderNest

The agent crew

Thirteen agents. One job each. Rupees stated before they act.

Every flow in a restaurant ends in a decision somebody makes alone. We can put an agent on each because OrderNest already holds the one thing they need: a priced-up, balanced record of what actually happened.

How an agent works

From the books to your phone, and back into the books.

01Reads the ledgerNot a dashboard export: journal lines tied to a dish, a shift and an outlet.
02Finds the driftVariance against recipe, budget, last week or the forecast, above a threshold you set.
03States it in rupeesWhat it is worth, what caused it, what to do. One message, no chart to interpret.
04You approveOne tap on WhatsApp. Silence is not consent. Low-risk, reversible actions can run on their own.
05JournaledWhat changed, who approved it, what the agent saw. Your CA can audit every action.

The crew

Pick an agent to see what it sends.

Badges: Live shipped · Early access running with early-access outlets · Coming 2027 on the 2027 roadmap.

ReadsLedger, sales, variance, labour, supplier prices
Decides9am WhatsApp: top three drivers in rupees per outlet, one action each
ModeInforms
NeedsNothing new

Example messages use illustrative numbers from the product; they are not customer data.

Guardrails

An agent you can hand your books to.

These apply to all thirteen. They are product rules, not a policy page.

  1. States the rupee impact and the driver before any action.
  2. One-tap approve or reject on WhatsApp. Silence is not consent.
  3. Every action journaled: what changed, who approved, what the agent saw.
  4. Stays silent below a threshold you set per outlet.
  5. Reversible from the same message: price, roster or purchase order.
  6. Never moves money, payroll or supplier payments without explicit approval, even in act mode.

Why this works here

The agents are only as good as the record underneath.

Nory, the closest comparison abroad, has to import purchases, invoices and payroll from other systems before its agents can act. OrderNest books them. That is why the first four agents needed no new data to build.

Recipe coverage matters: an outlet with costed recipes for 90% of its menu gets useful agents in week one. Onboarding includes recipe costing for that reason.

What each agent reads
AgentReadsStatus
Daily BriefingExisting ledger dataEarly access
Ask OrderNestExisting ledger dataEarly access
Stock WatchExisting ledger dataEarly access
Close AssistantExisting ledger dataEarly access
Bill ReaderExisting ledger dataLive
Aggregator AuditorPayout statement importComing 2027
RepricingMenu push to aggregatorsComing 2027
Profit-aware OffersCoupon and loyalty redemption at billingComing 2027
ForecastingForecast spineComing 2027
RosteringForecastingComing 2027
OrderingForecasting, Bill Reader pricesComing 2027
Voice HostReservations moduleComing 2027
ReviewsFeedback and NPSComing 2027

Running a restaurant is hard enough

See what your books would have said last week.