Most HR cost-cutting conversations start in the wrong place. Someone looks at the HR headcount, decides it is high relative to the company, and starts asking whether the department could run with one fewer person.

That approach almost always makes things worse, because it removes capacity without removing any work. The work simply redistributes onto the people who remain, quality drops, and within a year someone is quietly hired back.

The better question is not how many people are in HR. It is how much of what HR does every week is genuinely HR work.

The honest answer is: not much of it

Benchmarking work by APQC, which tracks how HR functions split their time, has consistently found that a large majority of HR time goes to administrative work rather than strategic work. Deloitte research commonly cited in the same context puts the figure at roughly 57% of HR time spent on administration. Other studies put it considerably higher.

Whatever the exact number in your business, the shape is the same everywhere we look. The people you hired to manage hiring, retention, capability and culture are spending the majority of their week on data entry, chasing signatures, reconciling timesheets, answering the same leave-balance question for the eleventh time, and fixing payroll mistakes.

Payroll processing on its own tends to be the single largest administrative consumer of HR leadership time. And payroll is also where the errors are most expensive, because a payroll error is never just a payroll error. It is the correction, the reconciliation, the conversation with an upset employee, and in regulated markets the compliance exposure that comes with it.

Where the cost actually sits

When we map an HR function, the cost almost never turns out to be where the client expected. It clusters in four places.

  • Data moving between systems by hand. Someone exports from the time-tracking tool and rekeys into payroll. Someone copies new-starter details from a form into three different platforms. Every one of these steps is a place errors enter, and every error costs multiples of what the original task cost.
  • Chasing. Chasing managers for approvals, chasing employees for documents, chasing finance for sign-off. This is invisible in every process document ever written and it is frequently the largest single time cost in the department.
  • Answering repeat questions. Leave balances, policy clarifications, payslip queries, letter requests. Individually trivial, collectively enormous, and almost entirely answerable from data the company already holds.
  • Rework. Everything that has to be done a second time because it was wrong, incomplete or done against an out-of-date version.

Notice that none of these are jobs. They are steps. That distinction is the whole basis of doing this without redundancies.

What we did in our own business

We took ITSOL from 43 people to 24 producing the same output. It is the proof point we lead with, and the mechanism matters more than the number.

We did not sit down and decide which roles to eliminate. We mapped every recurring process in the company and identified every step that existed purely because a human was moving information from one place to another, or checking whether someone else had done something. Then we removed those steps.

What was left, in every function including HR, was work that required judgement. Roles changed shape considerably. People ended up doing the part of their job they had originally been hired for and had not had time to do. The headcount came down through attrition and through not replacing administrative capacity we no longer needed, not through a redundancy round.

Across client engagements the pattern repeats. HR admin costs typically drop 60 to 70% within the first 90 days, and most clients get 60 to 70% of their target processes automated in that same window. It is not that a magic tool arrives. It is that a large fraction of the work turns out, on inspection, to be unnecessary.

The order to do this in

The reason most HR automation projects disappoint is sequencing. A system gets bought first, then the team tries to fit the existing mess into it. What you get is the same broken process with a login screen in front of it.

Do it in this order instead.

Map how the work really moves

Not the policy. The reality. Follow a single new starter, a single leave request and a single payroll run all the way through, and write down every hand-off, every wait and every place someone has to ask a question to proceed. This is uncomfortable and it is the highest-value step in the entire project.

Count the waiting, not the doing

People instinctively measure how long tasks take. The cost is usually in how long things sit. An onboarding pack that takes forty minutes to assemble but waits six days for an approval is a six-day problem, not a forty-minute one.

Fix the data entry points first

The highest-certainty wins are always the places where a human retypes something a system already knows. These are unglamorous, they never appear in a vendor demo, and they deliver the fastest measurable return.

Give every automation an owner

An automation with no named owner is an automation that will silently break and be abandoned. This single discipline separates the ones still running in year two from the ones quietly switched off.

What does not get automated

It is worth being clear about the limits, because overselling this is how trust gets lost.

Judgement calls do not automate well. Performance conversations, grievance handling, deciding whether an exception to policy is warranted, working out why a good employee has gone quiet. These need a person, and freeing up time for them is the actual point of the exercise.

Anything where the process genuinely changes every time is also a poor candidate. If there is no repeatable pattern, automating it produces a brittle system that costs more to maintain than it saves. Part of a good assessment is telling you which of your processes fall into this category and should be left alone.

Start by finding out what it is costing you

Most HR functions have never had the true cost of their administrative load calculated. Not estimated in a meeting, calculated. It is usually a larger number than anyone expects, and once it is on paper the business case tends to make itself.

That is what our AI Audit produces. Over 14 days we map how work moves through your business, cost the friction, and come back with the three to five automation opportunities worth doing, ranked by impact, effort and how ready your operation is for them. It is $1,500, or SAR 5,600, and you own the roadmap whether you build it with us or not.

If the honest answer turns out to be that your processes are already lean, we will tell you that too. It is a cheaper way to find out than a software subscription you spend a year regretting.