As high summer gave way to early autumn, the rhythm of the office never once slowed down.
Sitting beside me is Daisy, our HR administration partner. Eight hours a day, her hands barely leave the keyboard: for the company’s 400-plus colleagues, every onboarding, departure, unpaid leave and unit transfer has to be checked, every notice sent, every list distributed.
She is quiet, careful and reliable — which makes it easy to forget that this day-after-day repetition can wear a person’s enthusiasm down, bit by bit.
That kind of busyness has a structural cause. When an organization grows fast, people come and go far more often, and all the administrative chores of every arrival and exit land on the same pair of shoulders.
So, through scenario interviews, we laid the work out task by task — and found that the most repetitive of all was those onboarding and departure notification emails.
Every day she had to pull a list from the HR system, check it column by column, apply the formatting, and send it out one message at a time. The same queries, the same fields, the same recipients had followed her like a shadow for years. An engineer had once written her a piece of Python automation, but after he left, it sat broken for more than half a year; and the payroll cases outside the HR system ate up plenty of her time too.
After the interviews, we set about handing this over to automation: letting the system enter the database on its own each day to read the HR data, match the lists, format and send. To make adoption painless, we also built a web app as the bridge between her and the system — monitor, preview, rehearse, send for real; every step visible and clickable, without touching a single line of code.
The most important stroke in the whole design was “human in the loop” — in plain words, keeping a person inside the process at all times.
The gatekeeping level of every automated flow can be tuned: fully automatic, or paused before sending until a person has confirmed it by hand. Every rehearsal and every send leaves a record the system keeps, traceable at any time. The machine handles repetition, the human handles judgment, and trust grows slowly in the space between.
After the platform went live, my hope is that the time it frees up can go back to the places that truly need warmth: face-to-face care, careful communication, understanding each colleague’s situation.
This process convinced me that the most moving thing about applying AI is that it can take over the repetition that wears people down, so that we can give our most precious time to the people beside us.
And even as we embrace the tools, we will always remember: efficiency goes to the machine; judgment and warmth belong, always, to people.