The newest member of my team answers enquiries at three in the morning, in four languages, and has never once complained about the shift. She is an AI. She runs my front desk, and building her showed me a category of work that hides inside every role on every team. It is exactly the work AI should get first, and it is usually not the work teams hand over.
I have started calling the way to find it the front desk test. The leadership teams I work with in Switzerland apply it to their own functions within an hour, usually with some discomfort, because the test keeps pointing at work everyone had silently accepted as the job.
Every Role Has a Front Desk
Watch any team for two weeks and a layer becomes visible that nobody was hired for and everybody does. The status update assembled from the same four sources every Monday. The first draft that always starts from the same skeleton. The question from another department that has been answered thirty times this year. The reading of an incoming request just far enough to know where it belongs. The follow-up that never happens because the day is full.
This is front desk work: essential, recurring, and not the reason anyone on your team chose their profession.
From the inside it is nearly invisible. Every instance carries a real sender and a real deadline, so nobody experiences it as repetition; it simply feels like load. In workshops with leadership teams, the moment that changes the conversation is deliberately unglamorous: each person lists what they actually did over the last ten working days. The repeating layer is always larger than anyone guessed. And it concentrates in the most reliable people, because front desk work flows towards whoever handles it well. That is how your strongest people end up spending their attention on the work that needs them least.
The Front Desk Test
The teams that get AI adoption right almost never start from a tool. The decision that matters is not made in a steering committee; it is made by one tired team member on a Tuesday afternoon, deadline approaching, choosing between the tool and the way they worked last year. What tips that choice is whether the tool has already proven itself on a task exactly like this one.
So start from a task, and use the test to choose it. Three questions, asked of any recurring piece of work.
First: does it repeat in the same shape? Not the same words, the same shape. Enquiry in, standard answer out. Data in, report out. If the last five instances would look interchangeable in a stack, that is a yes.
Second: can you write down what good looks like? If two experienced people can each describe a good version in a few sentences, and their descriptions match, then AI can be briefed against a standard and, more importantly, checked against one. If nobody can articulate the standard, stop. The task only looks routine because one person has been carrying its judgment silently, and that judgment is not something to automate away unexamined.
Third: does anyone on the receiving end need it to be a person? Some work is repetitive on paper and human in substance. A contract renewal call with a nervous client repeats every year, and the trust in the room is the actual product. If the other side needs to feel a person, keep the person.
Repeats, has a describable standard, needs no relationship: that is front desk work, and AI should get it. Then one discipline from every successful rollout I have seen: hand over one task completely rather than five tasks halfway. A team that watches AI take one Monday task, fully and reliably, believes something no announcement can make it believe. Adoption spreads sideways. One colleague visibly getting two hours back does more than a leadership town hall, so give that story a slot in the Friday meeting and let it travel.
There is a reason to start with front desk work beyond the obvious one. Adoption moves through curiosity, then confidence, then courage, in that order. Front desk tasks are where confidence gets built: small, repeated wins on familiar work. March a team straight to courage by mandating AI on critical work and you skip two stages; what comes back is compliance theatre or quiet sabotage. And give the practice working hours. If trying AI on the Monday task has to happen after hours, the team hears what the priority really is.
What the Test Protects
The test has a fourth question, and it is the one leaders skip: who was learning by doing this work?
Some repeating work is not only output. It is formation. The junior analyst assembling the Monday numbers is also, invisibly, learning what the numbers mean and when one smells wrong. Strip every repetition out of a role and you can remove the apprenticeship along with the load. The report still gets written; the judgment that used to grow in the writing quietly stops growing.
So the front desk test ends with a leadership decision, not an automation decision. Hand the work to AI, and then decide on purpose where each person's judgment gets built now. In some teams that means juniors correct AI output against a standard a senior has written down. In others the reclaimed hours are pointed, explicitly, at the client work and the judgment calls people were actually hired for. What it must never mean is that the work disappears and nobody decides anything.
Expect the test to surface fear, because it draws a category called work AI should get, and everyone privately checks whether they are standing inside it. Say the honest sentence out loud: this layer goes, the roles change shape, and the intention is to move people up into the layer only they can do. Teams can work with a hard answer. What they cannot work with is silence.
My own front desk twin replaced nobody; there was nobody to replace. What she changed is where my attention goes, and that is the honest promise of this test for a team. Not fewer people. More of the week spent at the layer that actually needs them. You can meet her at www.thechangerepublic.com/ai-twin and see what a front desk sounds like when the human it frees is somewhere more useful. Then go and find your team's version of her job.
Change management consulting for AI adoption at The Change Republic, working with leadership teams in Zürich and across Switzerland to choose the work AI gets and build the practice that makes it stick: www.thechangerepublic.com/change-management
Tünde Lukacs is an executive coach and founder of The Change Republic. She works with leadership teams across Switzerland and Europe on leadership development, team performance, and the people side of AI adoption.