When AI leaves the chat window it spends real money and makes real promises. The field's words, plainly, and the engine Burq brings to it.
- Physical-world AI
- AI whose outputs move atoms: orders, vehicles, inventory, promises to real customers. Unlike a wrong sentence, a wrong decision here costs money and breaks a promise, so the standard of proof is higher than fluency.
- Decisioning
- Choosing among priced alternatives under real constraints: which provider, which vehicle, which window, at what cost, against which promise. The hard part of physical-world AI, and the part that deserves a record.
- Orchestration
- Moving work between systems along a defined flow. Necessary plumbing, frequently sold as intelligence. Orchestration executes decisions; it does not make them, and it leaves no account of why.
- Telemetry
- What happened to an order: completed, on time, how long, how late. Every dashboard in the category reports it, and Burq includes it. The record adds what was known, what was weighed, who decided, and whether the call was right.
- Agentic AI
- The industry's word for software that pursues objectives across multiple steps and tools. The word describes effort, not trustworthiness: nothing about being agentic makes a system's decisions inspectable, and inspectability is what earns authority.
- Human-in-the-loop
- A human confirms decisions before they execute. Often used as a permanent disclaimer; better used as a stage. In Burq it is Review mode: one mode on a ladder, with a threshold for moving past it and a record either way.
- Shadow deployment
- The engineering practice of running a new system beside the incumbent with no authority, comparing outputs before any switch. Standard discipline in aviation and machine learning. Burq's Shadow mode is this practice made a product, per segment, on your live orders.
- Auditability
- The property that a system's decisions can be replayed afterward: what it chose, what it beat, what it knew, and which rules governed. A record written at decision time, not a report assembled after the fact.
- Ground truth
- What actually happened, as recorded: the order arrived in its window or it did not, the promise held or it broke. The only honest grader of a decision, and the reason grading must separate judgment from luck.
- Counterfactual
- The path not taken, kept with its price and its reason. Most systems discard theirs the moment a choice is made. Burq preserves them as alternatives weighed, which is what makes a decision arguable later.
- Quote
- What a provider says it will charge and what it says will happen: a price, a window, sometimes an ETA. The right interface for what it carries, and silent on whether the provider will take the job at this hour, how its window holds under load, or what the alternative cost the last time. Burq corrects every quote against the record before it competes.
- Burq Intelligenceburq
- The engine that reads orders, weighs alternatives, and makes graded decisions against the Decision Graph. It is the same engine in every mode; what changes between modes is authority, not care.
- Promise
- The commitment made to a customer at checkout: the window, the price, the handling. Burq decides promise-first, and a policy that would break promises cannot be saved.
- Shadow runburq
- A period in which Burq Intelligence decides beside your team on live orders with zero authority, its call written next to yours, order for order. The delta between the two records is the case for granting more.
- Matchburq
- The measure of agreement between Burq's call and your team's: how often the two come out the same on the same order, counted only where both decided under the same constraints. Two calls can differ by name and match in effect: the same promise held, a price inside your tolerance. Every divergence is priced, with its reason.
- Productive disagreementburq
- A divergence between Burq's call and your team's where, with both paths graded on what was knowable, the grade favors Burq's call. Matching your team is how Burq earns authority. Productive disagreement is why granting it was worth anything.
- Self-QAburq
- Burq grades its own work. Every closed order is revisited with hindsight: was the winner the best call on the board? A better call that existed is written down with its dollar figure. Your team's wins are written in your favor.
- Decision qualityburq
- One call graded on several dimensions, not one score: what it cost, whether the promise held, how reliable the provider was, how fast it arrived, how the customer experienced it. A cheap decision that broke a promise is not a good decision.
- Baselineburq
- What the same orders would have cost the way you used to dispatch them. Every saving on the statement is measured against it, order by order.
- Statementburq
- Savings realized against your baseline, itemized by reason and by order, exportable any time. What reconciliation recovers from provider invoices lands on it too.
- Replay
- Past orders run through Burq Intelligence to see what it would have chosen: a drafted rule against your last 30 days, a fleet change on paper before you commit. Replay orients. It does not grant authority, because history cannot say what the road not taken would have done. Authority is earned on live orders.
- agents
- Lowercase on purpose. Short-lived workers Burq Intelligence spawns per job to carry a decision out, then retires. Plumbing, never the product. Burq does not sell agents; it sells decisions you can read.