PASS — infrastructure management for decision-making

Every estate looks well managed until the day someone has to explain it.

The technology refresh. The new contractor taking over. The outage that took the network down and that nobody can account for. Those are the three mornings on which an organisation discovers that years of infrastructure were managed without anyone holding the map — and that nothing in the estate ever said so.

Start a conversation

The part nobody puts on a website

When the network goes down, somebody usually knows why. And they usually do not say.

Not out of negligence. Because saying it in the wrong room costs a career, and everyone in that room knows it. The silence has two consequences and they compound: information stays a little bit hidden, just in case — and the root cause is never fixed, only the symptom, until the next time.

This is the half of the problem that no monitoring tool addresses, because it is not a monitoring problem. It is an incentive. Aviation solved it decades ago on the apron, by separating the report from the blame. It was never applied to the comms room.

PASS is not looking for someone to blame. It records what is there and when it was seen, so the organisation finds out without anyone having to report it.

What PASS is

Today the answer lives in somebody's head. PASS reads it continuously, without being asked, and leaves it available to any system.

What is connected, where it is, and what depends on it: today you find that out by asking somebody. It is like using a calculator — you ask, you get a number, and nothing is kept; the next question starts from scratch. Or it is in a sheet somebody typed once, which was already wrong the following week.

PASS gives you the cells. It discovers what is physically on the network and writes it down — devices, chassis, interfaces, addresses, VLANs, access points, transceivers, modules, cables — and every entry records where it came from and when it was seen. Then it builds the scaffold for the next level, and does it again.

Everything is read and written through APIs — no back doors, no direct database access, no exceptions. So the fact is available to any system that needs it, inside or outside PASS: your ticketing, your ERP, a third party's dashboard, or something that does not exist yet. Your team builds on top without asking us, and without us having to integrate it case by case.

10 routines that write discovered state into the configuration database, from chassis down to individual cables. measured 2026-08-30 · source: the CMDB adapter · method: counted per function and read in context Each written fact carries its origin and the moment it was seen. An absence you have measured is a different fact from one you have not looked for, and a system that cannot tell them apart will let you decide on the difference.

The decisions

Knowing something is down does not tell you what to do.

The question is which service narrows, whose asset it is, whether there is a window to touch it, and what breaks if you do. Observability is the input. It was never meant to be the output.

Before you touch anything: what goes dark, whose it is, and which shift will notice.

Signed by operations — the duty manager, not the network team.

A duty manager does not buy availability. He buys being able to say yes without gambling. Today his default answer to anything he does not understand is no — and that no is why an infrastructure project takes months.

What is plugged in, against what the contract says.

Signed by property and contracts — a budget of revenue, not of cost.

Every divergence is one of two things: a service you provide and do not bill, or a third party connected to your infrastructure who should not be. Both matter to you, and neither appears in any report today. This one is not read by IT — it is read by whoever owns the leases.

How much it actually held on the day of a peak.

Signed by operations, together with everyone who suffers the failure.

An availability percentage does not tell you what happens on the day the wave arrives. A declared threshold does. Measured capacity lets you write down how far you hold the service, have it signed by the parties who suffer the failure, and check after each event whether it held or you were lucky.

For the three people who read this

When you are the only one who knows how it is all wired, you cannot even take a holiday.

You cannot be promoted either, because nobody can replace you. And you carry the blame when it breaks. The first reaction to a system like this is that it will take your job. It takes the part of the job nobody should want: being the human index of a network.

It is better to be deciding the next step in adopting technology than fighting an unstable system. Being busy is not the same as being useful.

The ambition is not fewer engineers. It is engineers with enough time and enough information to think one level up.

And for the person who answers for all of it.

Today an IT function has no shortage of measurements, and almost none of them reach the table where budgets are decided. Not because they are wrong — because there is no route from them to that table. What travels is a change whose cost is known before it is made, a failure expressed as the service it narrowed, and a contract that is being served and not billed. PASS produces those, in the language of the people who sign.

AI-native

Artificial intelligence does not fix data that nobody has organised.

Applying intelligence to an estate nobody has inventoried does not produce worse answers. It produces answers delivered with the same confidence, and wrong. A correlation engine cannot distinguish A does not depend on B from nobody ever wrote down that A depends on B — the failure is silent, and it points the permissive way.

And the difficult part of an estate like this is that the dirt does not show. The dashboards are green, the services return 200, and there is no error anywhere.

You decide when your organisation adopts AIOps, and at what pace. What you cannot decide later is whether your data will support it — that was decided the day you started collecting it.

Pre-aggregating on ingest destroys high-cardinality identifiers irreversibly, and the rare failure — which in critical infrastructure is the one that matters — disappears for good. Widening the polling interval dilutes a fifteen-second outage into the average of the minute, so the alert never fires although the user certainly felt it. Both look like savings and are permanent blindness.

So we build the rungs nobody wants to pay for and without which the upper ones cannot work: an inventory with provenance, a topology that yields a blast radius, and the owner and criticality that turn a red light into a decision. That is what AI-native means here. It is a property of the foundations, not a layer applied on top.

Boundaries

What it does not do, said before you ask.

It is not a service desk, and it is never the system of record for an incident. It raises the ticket in the queue you already run.
It does not become a second source of truth for an estate you already record elsewhere. It reads and writes through that system's API. Two inventories that disagree are worse than one that is incomplete.
It does not diagnose for you. Statistics narrow the search; the diagnosis is signed by a person. Anyone selling automatic root cause should be asked what answer space it classifies against, and what its measured accuracy is.
It does not claim a certification we do not hold. The design is organised against the control families of ISO/IEC 27001 and NIST CSF 2.0, and we can show you which of our controls are procedures and which are technical. They are not the same thing.
It does not cover operational technology. Baggage handling controllers, SCADA, programmable logic. Different safety, certification and hard real-time constraints. If you need that, we will say so rather than stretch to reach it.

Contact

Talk to us at whatever level of detail you need.

If you are evaluating us as a supplier, say so in the first line and we will start with the security questions rather than the product ones.

Email: hello@airfexlabs.com
Security contact: /.well-known/security.txt