Nº 001 · Platform
One layer above every system you run.
Opaxa reads your systems, holds your operating context, and turns both into decisions for each location, inside the rules you set.
Nº 002 · How a Decision Is Made
One decision, end to end.
Every recommendation runs the same loop: bring in everything the operation runs on, work out what it means for margin, recommend with the reasoning attached, act at the level you have granted, then measure the result and learn from it. The next recommendation starts where this one ended.
Bring it all in
Understand it
Weigh the margin
Recommend
Decide
Do the work
Measure it
Learn from it
- Bring inPoint of sale Schedule Flight board Weather Lease terms Last eight Tuesdays
- UnderstandA 6:40 am flight was added at the nearest gates, and the lease requires this location to open at 5:00 am.
- WeighWith one cashier, the line passes 12 minutes before that departure and roughly 9% of morning guests leave without buying. A second cashier keeps labor at 24% of sales, under the 26% target.
- RecommendAdd a second cashier from 5:00 am to 8:30 am.
- DecideSuggest The general manager approves it the night before.
- ActThe shift goes to the next person on the seniority list, who accepts. It is on tomorrow's schedule.
- MeasureThe line peaked at 5 minutes. Morning sales finished 11% above the original forecast, and labor landed at 24% of sales.
- LearnAdded flights at the nearest gates now move this location's opening plan.
Nº 003 · Signals and Context
What your systems know, and what they never will.
Everything your operation records.
Sales by item and hour. Labor, time and payroll. Purchasing, invoices, receiving and inventory. Outside the building: flight schedules, event calendars, occupancy, traffic and weather. Opaxa reads them together, so the decision about tomorrow's labor already knows about tomorrow's demand and this week's food cost. New sources are connected per deployment.
Everything that makes it yours.
Recipes and plate costs. Vendor contracts and pricing. Leases and their requirements. Labor rules. Equipment manuals. And every correction your managers make. Context is the part no general model has and no competitor can copy. It compounds with every day the agents run.
Nº 004 · Agents
One agent for each kind of decision.
Each agent owns one kind of decision. All of them share the same signals and context, so the labor decision is made knowing the food cost decision, and the markdown knows the forecast.
Demand
Forecasts every location by the hour from sales, schedules and outside signals, and measures itself against actual results every day.
Concessions
A schedule change added two departures to the 6:40 am bank. Opening-hour demand revised up 18% at the three nearest locations.
Labor
Recommends staffing against the forecast, naming the shift and the hour.
Restaurants
Twelve locations ran 2.1 points over their labor target last week, concentrated between 2:00 and 4:00 pm. Three schedules rebuilt around it.
Shift Coverage
Fills call-outs by offering the shift to eligible staff in the right order.
Concessions
The opening cashier called out at 4:10 am. The shift was offered in seniority order and accepted by the second eligible employee.
Labor Compliance
Checks draft schedules against labor contracts and local scheduling law before they post.
Restaurants
Draft schedules at two locations miss the city's advance-posting requirement. Corrected before publication.
Menu and Cost
Costs every item to the recipe and flags items whose margin moved.
Concessions
The street-pricing cap holds the burger at a fixed price while beef rose 9%. A portion adjustment restores the cost target.
Inventory and Ordering
Sets pars and drafts every order from the forecast.
Restaurants
Tomorrow's orders drafted to the forecast across 18 locations, with pars adjusted for Saturday's weather.
Invoice Audit
Checks every invoice for price against contract, quantity against what was received, duplicates, surcharges and unapproved vendors.
Restaurants
The same invoice was submitted twice through two approvers. The duplicate was held before payment.
Vendor Credits
Drafts the credit request for every variance and follows it until the credit posts.
Concessions
Three cases short on Tuesday's delivery. Credit requested with the receiving record attached and tracked until it posted.
Pricing and Mix
Weighs what each item sells against what it earns, and recommends what to reprice, feature or remove.
Restaurants
The top-selling item carries a 38% food cost against a 30% target. A $0.50 increase closes most of the gap with minimal volume risk.
Markdown and Sell-Through
Finds what is not moving and recommends whether to discount, bundle or transfer it before it becomes waste or dead stock.
Retail
A seasonal line has sold through 31% at the halfway point against a 60% plan. A 20% markdown is recommended at eight stores, with remaining units consolidated from three.
Promotions
Measures what each promotion actually earned after discount and cannibalization, and recommends what to extend or end.
Restaurants
Happy hour lifted traffic 14% but reduced margin on core items. Net contribution is negative; a revision is proposed.
Briefing
Delivers each manager yesterday's results, today's outlook and the decisions waiting.
Concessions
Yesterday landed 3% above forecast. Two staffing decisions and one invoice variance are waiting.
Cash and Loss
Flags cash, void, refund and discount patterns by shift and register, with the transactions attached.
Convenience
A cash variance recurred on the same shift at one site three times this month. Transactions routed to the district manager.
Payroll Audit
Catches payroll errors before payroll runs: missed punches, tip pool math, miscoded overtime and rate mismatches.
Hotels
Banquet overtime was coded at the standard rate for two pay periods. Correction staged before payroll runs.
Equipment
Answers equipment questions from your own manuals, cites the page, and says plainly when to call for service.
Restaurants
A walk-in cooler ran above temperature overnight. The manual's diagnostic points to the door gasket; service call drafted.
Contracts and Leases
Tracks every obligation, breakpoint and renewal in your leases and contracts.
Concessions
One location is tracking to cross its percentage rent breakpoint in November. The rent impact is added to the forecast.
New agents deploy continuously. The roster grows with every operation we run.
Nº 005 · Governance
Suggest, Confirm, Auto.
Suggest
The agent recommends. A person decides. Every agent starts here.
Confirm
The agent prepares the action. A person approves it.
Auto
The agent acts inside the limits you set and reports what it did.
Every agent starts in Suggest mode. It moves forward only when it has been measurably right on your data, and only when you promote it. When a manager overrides a recommendation, that correction carries more weight than anything else the agent learns from. Every decision is logged with the data and reasoning behind it, and limits are set by dollar amount, vendor, location and approver.
Nº 006 · The Daily Briefing
One page, every morning, before the first hour.
Each manager gets one briefing: how yesterday landed, what today looks like, and the decisions waiting for a yes. Leadership gets the same view across every location.
Daily Briefing ·
Yesterday landed 3% above forecast.
Nº 007 · Deployment
Deployed carefully. One operator at a time.
Onboarding agents map your menus, recipes, vendors, par levels and rules into Opaxa, so each new location starts from everything the last one taught it. The first location takes the most work. Every location after it takes less.
01
Discover
We map how your operation actually runs: systems, locations, rules, and the decisions that cost the most when they are wrong.
WeMap systems and data. Choose the first decisions.
YouAn operations lead and time with the people who make the calls today.
02
Connect
We connect your data and load your operating context.
WeBuild and test each connection. Structure your documents.
YouData access or exports, and the documents that define how you operate.
03
Calibrate
Agents run in Suggest mode on your real days. Nothing is written to your systems.
WeTune each agent until it is measurably right. Report accuracy openly.
YouManagers who review and correct recommendations.
04
Promote
When the record supports it, you decide what moves to Confirm or Auto.
WeShow the record behind every promotion.
YouThe decision.
Every deployment includes work that cannot be automated away: your contracts, your leases, your recipes, your rules. That work is what makes the agents right on your operation instead of roughly right on everyone's, and it is why we deploy with a limited number of operators at a time.
Nº 008 · Trust
Your data stays yours.
Never Used for Training
Operating data is never used to train any model, ours or anyone else's.
Encrypted and Protected
Your data is encrypted and protected at every step, and it is only used to run your operation.
Every Decision on the Record
Every agent decision is logged with the data and reasoning behind it. Agents act only at the level you have granted.
Built on Frontier Models
Opaxa's agents are built on frontier models. As the models improve, every agent improves with them.
Nº 009 · Contact