Ground truth at enterprise scale is finally possible. Whether it deserves to be trusted depends on how it is built.
The picture that’s been out of reach
Something has quietly become possible that wasn’t before: IT can now see what it actually provides — through the eyes of the people it supports.
Not what the catalog says it provides. Not what the vendors report or the dashboards summarize. What actually lands — the way work really flows, the way technology really behaves in use, the way thousands of people actually experience the services they rely on every day.
For as long as enterprises have measured themselves, that picture has been out of reach. What stood in for it was a set of approximations: hierarchies, taxonomies, catalogs, scores. Useful, hard-won, and always a compromise. The true picture has a name — ground truth: what is actually happening, as experienced by the people it’s happening to. This essay is about why ground truth is finally within reach at enterprise scale — and about the principle that decides whether it deserves to be trusted when it arrives.
Why every catalog is incomplete
Start with why the old picture was always an approximation. Enterprise reality is not hierarchical. Take an example every IT leader knows: a collaboration platform that is simultaneously a phone system, a meeting room, a file store, and a development platform. Where does it sit in the catalog? The honest answer is: in five places at once. But a hierarchy demands one box, so someone chooses — and the choice keeps one true dimension and sets the others aside. Multiply it by every service in the enterprise and the catalog becomes a structure of kept dimensions and quiet omissions, maintained with real effort, holding less of reality every quarter.
Two truths, one tree
There’s a deeper version of the same problem. Every service has two truths. There’s what it does for the people using it — the capability it delivers into their working day. And there’s who owns it — the team accountable for running it, funding it, fixing it. Both are legitimate. They are different shapes. A tree can encode one or the other. It cannot hold both. So every catalog quietly picks a side — usually the ownership side, because that’s who maintains it — and the people the services exist for end up navigating a map drawn from one of the two truths.
If you’ve worked in IT for any length of time, you’ve lived the large-scale version of this. The perpetual re-org. Centralize for efficiency and standards; a few years later, decentralize to get closer to the business; a few years after that, back again. Each swing is announced as a new strategic direction. It’s a cycle that has been running since the mainframe era, and the matrix structures invented to split the difference mostly just gave people two bosses. The usual explanations are politics or new leadership marking territory, and there’s some of that. But the deeper reason is the same one that limits the catalog: IT genuinely has two correct shapes — organized around what it delivers, or around what it runs — and a hierarchy forces you to pick one. Whichever truth you pick, the other one accumulates grievances until the pain justifies the next re-org. The pendulum was never indecision. It was a data structure problem wearing an org chart.
None of this was anyone’s failure. It was a limitation of the tools. Hierarchies were what we had for organizing complexity, so we organized reality into hierarchies and lived with the loss.
The clean-data myth
That limitation has just lifted. The prevailing wisdom about enterprise AI says: clean up your data first. Rationalize the taxonomy, fix the catalog, then bring in the machines. I’d suggest the more interesting truth runs the other way. Inference-based AI doesn’t need reality pre-sorted into a tree. It works on relationships. It handles many dimensions at once. It finds the structure that is actually present in messy, cross-cutting, contradictory evidence — rather than demanding that reality conform to a structure first. The messiness that was beyond the reach of twenty years of catalog projects is exactly the condition this technology was built for. The data never needed to be clean. We can finally make sense of it as it is.
What would make it trustworthy?
So the capability now exists for IT to build a true picture of what it provides, from the ground up. Which raises the question this essay exists to answer: what would make such a picture trustworthy?
Not the intentions of whoever builds it. Intentions are real, but they are not architecture. Companies change hands, incentives shift, people move on. A picture whose honesty depends on its operator staying honest is a promise, and enterprises should not run on promises. The standard worth building toward is different: independence that is structural — built into the measurement itself, so that trusting it doesn’t require trusting anyone.
Structural independence has two conditions.
Condition one: the anchor is ground truth
Every organization already contains the truest account of itself, held by the people doing the work. They see what the dashboards don’t reach and what never becomes a ticket. Their perspectives are plural, shaped by role and place and habit, and no single party prepares them before they arrive. They have nothing to hide and everything to gain from an honest picture. Gathered independently and at scale, that account is the ground truth of the enterprise — a foundation no ownership structure can reach. Every other source of evidence, from tickets to telemetry, can then be read safely against it. Owned data informs a picture. It cannot anchor one. Ground truth can — exactly because of where it comes from.
Condition two: the honest instrument
An anchored picture still passes through a tool, and a tool can flatter. So the instrument must show its own reliability, on its face. Confidence stated rather than implied. Denominators visible, not buried. Coverage gaps admitted as gaps. Imbalances in who responded, and how, flagged rather than smoothed away. Corroboration marked where it exists and absent where it doesn’t. Built this way, the instrument reveals any distortion in the picture — including distortion that might favor its maker. That’s the test of the second condition: the tool stays honest even if you stop trusting the people who built it.
Other disciplines crossed this threshold as they matured. Medicine pre-registers its trials. Engineering publishes its safety factors. Aviation investigates its own failures in public. In each case the field stopped asking to be believed and started showing its work — and trust deepened, not despite the exposure but because of it. Enterprise measurement is young by comparison. That threshold is in front of it now.
The standard
Put the two conditions together and you have a simple standard anyone can hold any picture to. Where does its foundation come from — could any party with an interest in the outcome have shaped it before it was seen? And does the instrument show me its reliability, or ask me to assume it?
That standard doesn’t yet exist at scale. Building toward it is why Voxxify exists. But the standard is bigger than any one company, and it should be. It belongs to every practitioner who has carried a catalog as far as hierarchy could take it, every leader who has presented a picture upward while quietly wondering what it missed, every builder of measurement tools willing to let their instrument be checked rather than believed. The technology for IT to see what it actually provides has arrived. What remains is the standard: independent ground truth, built so it deserves to be trusted.
