The Onboarding Steps You Can Automate, and the Two You Can't

By the time a new hire reaches their first Monday, most of their onboarding has already taken place. Accounts were provisioned when the offer was accepted, the equipment order went out the same afternoon, training modules assigned themselves, and the first-week tasks appeared in a channel with owners and deadlines attached.

That shift arrived at a useful moment. In March 2026, 22.6 percent of US workers teleworked or worked at home for pay. A substantial share of new starters are therefore people their employer will never meet in person, and the process has to hold together without anyone walking across an office to resolve a loose end.

Two steps didn't automate along with the rest, and most teams closed the gap by converting them into form fields. The new hire types an address and uploads a photograph of a document, and the task is marked complete. This moves at the same speed as everything around it, which is why it passes for automation, though a form field collects information rather than confirming it.

What automation genuinely resolved

The parts of onboarding that automated well were the parts that had always been a sequencing problem rather than a judgement problem.

Nobody was ever uncertain about whether a new starter needed a laptop. The failure was that the request sat unread in an inbox for four days, moved to the wrong person, and was then forgotten because the hiring manager was on leave. Workflow tooling addressed that by making the sequence explicit and attaching an owner and a deadline to every step. Workast's guide to automating your business with AI workflows covers the mechanics of setting it up.

The boundary of that achievement is worth stating plainly. Automation is reliable at ensuring a step occurs. It has nothing to say about whether the step produced a true answer, since a task can be marked complete by anyone who has typed something into the field it depends on.

For most of onboarding, that distinction does not matter. For two steps, it matters considerably.

The first is where the person actually lives

An address is easily mistaken for an administrative detail, the sort of information collected so that equipment arrives at the right door. For a remote hire it carries rather more weight, and it is the one piece of onboarding data that reaches an employer as a string somebody typed into a box.

This is the point at which proof of address verification does something a form field cannot. It reads the address from the uploaded document, compares it against the value the applicant entered, and examines the file for indications of tampering that a reviewer skimming a PDF on a phone would not detect. Utility bills, bank statements, tenancy agreements and government correspondence are all accepted, and some checks confirm an address against official records without requiring a document at all.

The reasons for confirming it have less to do with fraud than with the ordinary machinery of employment. The address determines which payroll taxes are owed and to which authority, and it decides which employment rules apply in matters such as notice periods, statutory leave and the minimum wage floors that differ by state as well as by country.

It also affects whether a contractor arrangement survives examination, because a contractor working from a jurisdiction the company never registered in becomes the finance team's problem rather than their own. And it governs where a laptop carrying encryption can lawfully be shipped.

What the check leaves behind matters more than the check itself. The record states that this address was confirmed, from this document, on this date, rather than noting that a file was attached and a task closed.

The second is that the person is who the document says

The other step that resists automation is confirming the identity of the individual being hired.

Most teams assume the interview settled the question. A candidate appeared on video several times, answered difficult questions well, and demonstrated a real command of the work. That establishes competence. It does not establish that the person added to payroll is the same person, nor that either is the person named in the contract.

The common version of this is undramatic. A contractor subcontracts the work without mentioning it, an agency substitutes the person assessed for the person available, or a candidate presents a relative's documents because their own circumstances are complicated. None of it requires a criminal, and all of it leaves an employer paying someone whose name it does not reliably know.

An identity check at onboarding resolves this in roughly the time an upload takes. The individual presents a government document and a live selfie, the two are matched, and the outcome is recorded against their file.

The argument for it has little to do with suspicion. When a payroll query, a tax notice or a client's own vendor audit eventually asks who exactly was paid, the answer needs to be a record rather than a recollection, and by that point the colleagues who would have remembered the circumstances have frequently moved on.

Collecting and confirming are different operations

The principle beneath both checks is straightforward. Collecting information and verifying it are distinct operations, and onboarding tooling is very good at the first of them.

It helps to look at how the current arrangement fails. A postcode is mistyped and the error goes unnoticed for nine months, until a payroll filing is rejected. A document is uploaded and never opened, because opening it belongs to nobody's task list and the checklist turned green the moment the file attached. Or the document is opened by someone who sees the correct name on a PDF and has no means of establishing whether the address line was altered before the file was saved.

None of these look like failures at the time. They look like completed onboarding.

For example, an agency that engages four contractors in a month and collects four uploaded utility bills holds four completed tasks and, in practice, no verified addresses, because nothing in the process has compared a document to a form field.

Placing the two checks in an automated flow

Neither check needs to sit outside an existing process, provided four points are settled deliberately.

The first is sequence. A check that runs after someone already holds a laptop, a payroll record and a Slack account is a report rather than a control, so both checks belong alongside contract signature.

The second is the treatment of failure. Most mismatches have mundane explanations: a recent move that the utility account has not caught up with, a middle name that appears on some documents and not others, or a bill held in a partner's name, none of which should halt a hire while somebody works out which applies. Route them to a named reviewer with a defined next step, since nothing teaches a hiring manager to circumvent a process faster than a hard rejection triggered by a typo.

The third is storage. The verification outcome belongs wherever the employment record lives, with its date attached, rather than in a direct message or a shared folder that nobody maintains.

The fourth is repetition. Annual re-verification of an entire workforce is largely wasted effort. A person reporting a change of address is the event that warrants a fresh check, and that event can be made a trigger like any other.

One further question deserves an answer before it becomes urgent, which is who owns these two steps. In smaller companies they tend to fall to whoever configured the onboarding workflow, usually an operations person with no particular mandate for them, and an owner named early costs nothing at a moment when the choice is obvious.

What automation does not cover

Onboarding automation has been a real improvement. Steps occur in order, nothing is lost when someone is away, and a new starter's first week reads as organised rather than improvised.

The gap it leaves is narrow and specific. Automating a workflow guarantees that a step took place. It establishes nothing about whether the answer was true, and for two steps in every onboarding that's the entire reason the information was collected in the first place.

Both checks take a new hire a few minutes and the team none at all once they are configured. Without them, what accumulates over a year of hiring is a set of completed tasks that nobody is able to stand behind at the moment somebody asks.