Your HR system already knows

This issue is about a distinction that took me longer than it should have to see clearly: the difference between an integration that keeps two systems agreeing, and one that actually causes something to happen. Almost every integration in the IT stack is the first kind. The gap between them is where a new hire’s first day quietly goes wrong.

Remote employee onboarding hardware goes wrong not because your HR system is weak, but because most integrations sync records and very few cause actions. An HR system holds the start date, country, role and equipment tier weeks in advance — the best data in the company about who needs a laptop and when. The record syncs, two systems agree, and then somebody has to notice, open another system and retype the name. Closing that gap means something has to be listening: a start date should cause device deployment in the right country, not just appear in a second system. HRIS integration came up in 11.9% of 3,977 customer conversations, almost always as a system buyers already had.

The signal: a solution buyers already have

HRIS integration came up in 418 of 3,977 customer conversations (11.9%), counting only cases where the customer raised it before we did. That places it firmly in the second tier of buyer concerns — well behind the middle-of-lifecycle themes, well ahead of most things vendors like to talk about.

What makes it interesting is not the frequency. It is that buyers raise it as a solution they already have. The HR platform is in place, it works, and the reasonable expectation is that connecting it to device management closes the loop.

The same pattern shows up across the whole stack. The most-mentioned names in our corpus are not competitors — they are systems our customers already run. Endpoint management appears constantly: Intune in 479 conversations, Apple Business Manager in 274, Jamf in 190. Identity platforms are assumed rather than discussed. HR platforms, 418.

Three mature layers, all working. And a fourth — the physical one — with no equivalent system at all.

The problem: most integrations sync records

Start at the beginning, because the beginning is genuinely good.

Three weeks before someone joins, an HR system holds their start date, the country they will work from, their role, often their manager, their office, and the equipment tier attached to their band. It is accurate. It is entered by someone whose job depends on getting it right. It is, without much competition, the best data in the company about who needs a laptop and when.

Then the trigger fires, and here is what actually happens next. The record syncs. The person now exists in a second system too. Two systems agree about who has joined — which is exactly what they were built to do, and it is genuinely useful.

And that is where it stops.

Somebody reads the start date, decides it is time to act, opens another system and retypes the name. That step — a human noticing — is the one almost everyone assumes is automated, and it is usually the point at which the chain becomes a person’s memory.

The distinction worth holding onto is this:

Most integrations sync records. Very few cause actions.

Syncing a record means two systems stay in agreement about who exists. Causing an action means something happens in the world because a field changed — sourcing begins, a specification is chosen, a device is configured and sent, and somebody is told if any of that is not going to work in time.

Vendors use the same word for both. Buyers reasonably assume the second and receive the first.

Once you see that distinction, several familiar frustrations resolve into one cause.

The start date that arrives too late

Nobody missed it. It synced on the day it was entered, three weeks out. But nothing downstream was listening for it, so the work began when a person happened to look — usually much closer to the date than anyone would have chosen.

The integration that is scoped narrower than you think

Connections are built against a specific product, and a platform’s regional variants are frequently separate products with separate connections. An integration that works perfectly for one country’s version of an HR system may simply not exist for another’s. This is nobody’s failure; it is how integration catalogues get built. It is also rarely checked before it matters.

The offboarding that fires everywhere except physically

The leaver date propagates cleanly. Accounts are disabled, licences reclaimed, access revoked — all of that is software acting on software, and it works. The one item that requires a person to travel to a place and collect an object is the one item that does not move.

The two-trigger problem

Some companies route device provisioning from the HR system and some from a ticketing system, and a surprising number do both, because both plausibly own it. Two authoritative triggers is worse than one manual process, because each looks like the system of record and neither is.

The common thread is not that these tools are weak. It is that the whole stack is excellent at knowing and has no layer that does. Identity systems know who someone is. Endpoint management knows what a device is permitted to do. HR knows who is joining and when. All three hold a truthful record. None of them can make a physical object appear in a country.

The operator takeaway: five moves, in order

1. Ask what happens on the other side of the connection. Not “do you integrate with our HRIS” — that answer is almost always yes. Ask what a new record causes. If the honest answer is “it appears in our system,” that is a sync, and someone on your team is still the trigger.
2. Find the human step and name its owner. There is one in almost every provisioning chain. It is not a failure to have one, but it is a failure for nobody to own it. Write down who notices, and what happens when they are on holiday.
3. Check the integration against the version you actually run. Regional variants of HR platforms are often separate products. Confirm it against your instance, in your country, before the design depends on it.
4. Pick one trigger deliberately. Whichever system owns the start date for provisioning, make it the only one. A second authoritative trigger creates drift that looks like a system of record.
5. Measure from the trigger, not from the order. The useful number is the time between the start date being entered and the employee having a working machine — not the time from when someone got round to placing an order. The first number contains the problem. The second hides it.

Five questions worth answering in writing:

• Which system owns the start date for device provisioning?
• What does that system cause when a start date is entered, as distinct from what it records?
• Who is the first human in the chain, and who covers them?
• Does our integration exist for the regional version we actually run?
• How long is it, typically, between a start date being entered and a machine being ordered?

One GroWrk lens

Modern IT runs on three layers that all got automated, and one that didn’t. Identity governs who someone is. Endpoint management governs what a device is allowed to do. HR governs who is joining and when. The physical layer — procure, deploy, store, retrieve, repair, redeploy, retire — is the one still running on email and whoever is in the right country.

ITAM records the asset. MDM controls the asset. GroWrk executes the physical lifecycle.

Which is the whole reason we exist underneath rather than alongside those systems. A start date should be able to cause sourcing in the right country at the right specification, and an offboarding should be able to cause a physical collection — without a person in the middle remembering to begin.

I would be careful about overclaiming here, because this is genuinely hard and nobody has finished it. An action that crosses into the physical world can fail for reasons no integration can prevent: stock, a courier, an address, a person who is not home. What a system can do is start earlier, escalate on its own, and tell you the truth about what will not be ready — which is most of the value, and is still rarer than it should be.

One stat

418. The number of customer conversations, out of 3,977, in which the buyer raised HRIS integration before we did — 11.9% of the corpus, and almost always as a system they already have rather than one they are looking for.

(GroWrk Call Intelligence — analysis of 3,977 customer calls, April 2024 – August 2026. Buyer-led mentions only; conversations can raise several topics.)

The data was never the problem. Your HR system knows who is starting, where, and when, and it has known for weeks.

The question is whether anything in your stack is listening — or whether the plan is still that somebody notices.

If you want to see what it looks like when a start date causes the work instead of just recording it, we would be glad to show you.

Book a demo →