01
Sources
Answers begin with approved records, not an ungrounded model response.
Operational agents
AI agents that work inside your rules.
An Airon agent does one job inside one workflow. It gathers the evidence, prepares the decision, and waits for the person who owns the call. Then it records what happened.
Five controls
The five controls in every agent we deliver.
02
Permissions
The agent sees and acts only inside its assigned role and information boundary.
03
Approval gates
Material decisions stop with the accountable person before any action leaves the workflow.
04
Exceptions
Uncertainty, conflicts and missing evidence are escalated instead of being smoothed over.
05
Audit record
Sources, decisions, actions and acknowledgements remain available for later review.
Provenance
Which document, who approved, what happened — on record.
When AI acts inside real work, the questions that matter come later, and they are specific. Which source document — which revision, section, clause? Which model drafted it, which agent ran it, which tools did it touch? Who approved the result? We build workflows so those questions have answers in the record. Provenance shows where an answer came from; your people still decide whether it's right.
Running today
Every workflow we deliver runs with the sources attached, approval gates on, and the audit record kept. Pick a number out of a draft and you can see the document behind it and who signed it off.
Designed in
The blueprint every deployment is built on: retrieval that follows who may see what, which classification applies and which revision is current — and records a reviewer can walk from an answer back to its source.
In development
Deeper lineage across models, agents and tools, so a result can be traced through every step that produced it.
Gauge · Estimating
A priced draft you can check line by line.
Gauge reads tender material, retrieves comparable delivered work and current rates, surfaces scope and margin risks, then prepares a draft for estimator review. Every number that matters keeps its source and approval history.
Marshal · Safety response
The response and the record move together.
Marshal assembles conditions, procedures, roles and prior events around a field signal. The controller sees the evidence and conflict surface, approves the action, and receives a reviewable incident record without reconstructing it later.
The engagement
Gauge and Marshal are worked examples.
A pilot builds this kind of agent into your workflow. What Gauge does for estimating applies to any evidence-grounded, signed-off document: valuations, tenders, contracts, procurement recommendations. What Marshal does for safety response applies to any signal that needs governed action: document control, project controls, HSEQ.
Map · Inference Audit
One workflow, mapped end to end, and a straight answer: where AI belongs, and where it doesn't. The map is yours either way.
Prove · Governed Pilot
Four to six weeks. A working agent in the real workflow, measured against a baseline agreed up front. At the end, a number — and the keep-or-stop call made on it.
Place · Private Deployment
The proven agent moves to production — on-premise, the airon appliance, or approved cloud where you allow it, decided by what its records require. It runs alongside the AI you already use.
Run · Managed Operations
After go-live we keep watch on quality, cost and model behaviour as the work changes. The audit record stays current, ready for your clients to inspect. When the numbers support it, the next workflow gets mapped.
The first conversation is about the work, not the technology. If an audit is the right next step, we scope one. If it is not, we say so.