Starting point

An international software company with teams in several European countries had never deliberately built its HR administration. It had simply grown alongside the business. Employee master data lived in several spreadsheets that contradicted each other. Contracts and amendments were produced by hand from the last similar document. Leave balances were tracked in one file, payroll input in another, and somebody reconciled the two at month end. A reference letter took an hour apiece because every one started from scratch.

I was employed there permanently and responsible for HR operations. After the first few weeks I measured where the time actually went: around 30 hours a week of pure administration, spread across several people in the HR team. That was half the available capacity, locked in steps that required no decision at all.

An HRIS project was not on the table. There was no budget, no time and no sponsor for it. Whatever the route was, it had to work with the tools already installed on every laptop.

Approach

The first step was a process audit with a stopwatch. For two weeks, every recurring task was logged: what triggers it, which steps it has, how long each step takes and which of them involve a judgement call. The result was a list of roughly twenty routines, ranked by hours per week. The three largest items were data reconciliation, contract documents and reporting.

I automated those three with Excel VBA. The scattered files became one master data list that a macro checks against payroll and time-tracking exports and flags the differences, instead of somebody comparing line by line. Contracts, amendments and certificates have been generated from templates ever since, with the macro inserting the master data. The monthly report to management and finance runs from the same list at the push of a button.

Once the time saved by VBA was visible, the text work came next. Reference letters, standard replies to employee questions and job ads were drafted with AI: one prompt per document type, a fixed review step by a human, and no personal data in any system not cleared for it. The team tested every routine itself before it went into daily use.

The order mattered. Measure first, automate the biggest items, then the next ones. Start with the tool and you automate what is easy, not what costs time.

Outcome

After roughly four months, weekly administration stood at 5 hours instead of 30. The 25 hours gained went into recruiting and into working with line managers, the tasks the HR team was actually there for. Data quality rose noticeably because there was only one source, and payroll ran without corrections. A reference letter took about fifteen minutes instead of an hour, ten of them for the review.

None of it was expensive. Excel, VBA and an AI account with clear rules. The team has run the routines itself ever since, because it built them.

30 to 5hours of admin per week
25hours a week for value-adding work
1master data source instead of scattered files
0licence cost for a new system

What transfers to your company

  1. Measure first, then automateTwo weeks of audit with a stopwatch show where the hours really sit. Usually not where everyone assumes.
  2. The tools are already thereExcel and VBA are on every laptop. An HRIS can follow once the process is clean. Loading chaos into a system only relocates the chaos.
  3. AI within the rules, not beside themOne prompt per document type, a human review step, no personal data in unapproved systems. That turns a toy into a routine.
  4. The team builds it with youWhoever helped build and test a routine runs it afterwards. Whoever is handed one works around it.

Frequently asked questions

How long did the change take?

Around four months from the audit to the first month at 5 hours of admin. Two weeks of audit, then the VBA automations one after the other, and the AI routines last.

Why no HRIS?

There was neither budget nor a sponsor, and the process was not clean enough to be moved into a system. Automating with what was already there is what put the process in order. A system can be built on top of that.

Was this an interim assignment?

No. I was a permanent employee there, responsible for HR operations. The approach is the same one I use in interim mandates today, with more experience about which routines to tackle first.

Which AI tools were used?

A large language model under a company account, for reference letters, standard replies and job ads. The tool was not the point. The rule was: no personal data in a system not cleared for it, and every output gets reviewed.

Does this transfer to a company without Excel skills?

Yes. The audit and the order of work transfer, the tool does not have to. Where VBA is missing, the automation features of existing tools or a small script often do the job. The point is to start with the largest items.