When a recruiter leaves, their pipeline leaves with them

It loses almost nothing that is in the ATS, and almost everything that is not. The records, the applications and the history of every exchange stay in the software. What leaves with the person are the commitments in flight (the candidate they promised an answer by Friday, the client to call back after the holidays) and the context of each account, which was never written down because they never needed it to be. Both cost money, and the most solid measurement we have puts a figure on it: sales to the customers of a B2B salesperson who leaves fall by 13 to 17% over the following year.

That figure comes from a study published in 2017 in the Journal of Marketing, run on the data of a Fortune 500 company. It is not about recruitment, and that needs saying before applying it. But the mechanism it measures, a commercial relationship that rests on a person more than on a firm, is exactly the one at work in an agency or an IT consultancy, where the client knows their business manager by first name and does not always remember the name of the company employing them.

What exactly leaves with a recruiter?

Three layers, and only one of them is protected by the tools you already have.

The first is the database: candidates, clients, open roles, interview notes. It lives in the ATS or the CRM, and a departure does not touch it, provided it was actually entered there, which remains the most frequently broken rule in the trade. We have written elsewhere about why the spreadsheet next to the ATS always comes back, and a departure is the moment you find out what it held.

The second layer is commitments. An active recruiter carries, at any given time, several dozen dated promises: call a candidate back after their technical interview, chase an IT director who was due to sign off a budget at month end, tell a consultant what the client decided. None of them is an object in the software. They live in the recruiter’s head, inbox and calendar, and every one of them has a deadline that keeps running after the person has gone.

The third is context, and it is the most expensive. This client turns down candidates from the big consultancies, this buyer wants two CVs per role and never three, this manager picks up the phone and never answers email. None of it is wrong, none of it is written, and the successor will learn it again by making the mistakes their predecessor had stopped making.

How much does a departure cost?

More than hiring the replacement, and most of the cost sits where nobody measures it.

The Journal of Marketing study measures a fall of 13 to 17% in annual sales to the affected customers in the year following the transition. It adds two findings that are more useful than the headline number. A new hire limits the loss less well than an existing employee who is handed the account. And a successor who resembles the person who left, having handled similar customers, does better than a stronger salesperson brought in from a different book. According to the authors, the firm they studied would have avoided about 40% of its loss by choosing more carefully who took over each account.

Applied to our trade, the order of magnitude is quick to work out. Median staff turnover in French IT consultancies is 18%, and it rises sharply in firms of 15 to 50 consultants, as we set out when looking at how long placements take. A thirty-person firm therefore loses more than five people a year, some of whom carry a client book. Each of those departures creates orphaned accounts, and an orphaned account is one that nobody calls: it is the mechanism we describe for the clients who make your revenue without anyone calling them, produced in a single day.

Cobalt argues, in its piece on why systems will beat individual talent in IT consultancies, that business manager turnover makes an organisation that relies on their memory alone unsustainable. We share the diagnosis, and this article takes its narrowest part: what to do in the three months between a resignation and the last day.

Why is the notice period not enough?

Because it is almost entirely spent finishing things, and almost never spent passing them on.

On paper, the window is comfortable. A manager covered by the Syntec collective agreement owes three months’ notice on resignation, which leaves plenty of time to hand over every account. In practice, the person leaving wants to close their open deals, their manager wants them closed too, and the handover shrinks to one meeting in the final week, around a spreadsheet filled in the night before.

What gets passed on in that meeting is whatever the person thinks to mention, which means what is most present in their mind and usually already written down. The quiet commitments, such as the follow-up promised three weeks ago to a candidate who has not chased it yet, do not come up, because they only announce themselves when they fall due. They reappear a month later as a candidate who has stopped replying or a client who has given the role to someone else.

There is a concession that handover guides tend to leave out. Some departures come with no usable notice at all: a negotiated waiver, a move to a competitor where you would rather the person stopped touching the accounts, sick leave. In those cases the window is zero, and anything not written down before the resignation is lost. That is the strongest argument for dealing with the question before it arises.

What should be handed over, and in what order?

First whatever has a date and faces outward, because that is what costs you first when it is forgotten.

The order we recommend follows the cost of forgetting rather than apparent importance. First, the dated commitments made to people outside the firm: every candidate waiting for an answer, every client to call back, with the date that was promised. A candidate left without news drops out fast, and we have shown that silence drives candidates away well before bad news does. This list takes an hour to draw up if someone asks for it explicitly, and it never gets drawn up if nobody does.

Second, the live assignments and the consultants for whom the person was the main contact. A consultant on assignment who loses their business manager often loses the only link they had with their employer, which puts them exactly in the position of the consultant nobody calls anymore. They should hear the name of their new contact from a person, not discover it when their assignment comes up for renewal.

Third, the context of each account, in a few sentences per client: what they turn down, who really decides, how they like to be reached, what went wrong last time. And for the takeover, the study cited above offers a simple rule that runs against instinct: give the account to someone who already knows similar clients rather than to the best salesperson available.

What changes when an agent has worked with the person who leaves?

It keeps what it learned, because what it learned belongs to the company and not to the person who taught it.

This is one of the two points where the digital colleague analogy breaks down, and here it breaks in the right direction. When a business manager has told the agent, over the months, that a given client wants two profiles per role, that a given candidate will not travel, or that budget approval goes through procurement, those facts are recorded in the company’s memory. Their successor asks the same agent the day after the departure and gets the answer, with no handover meeting and no spreadsheet filled in the night before.

The limit needs stating precisely, because it is real. An agent only knows what it has been told or what has passed through it, and it learns nothing by osmosis. A team that never gave it any work gets nothing back from it when someone leaves. And that memory has to be able to forget, otherwise the successor inherits the out-of-date facts as well, which we cover in what an agent has to forget.

A second boundary matters just as much. What the agent remembers about the company goes to the successor, but what it remembers about the person who left (their preferences, how they work, what they told it about themselves) goes to nobody, neither the successor nor an administrator. A company’s memory and a person’s memory are two different things, and a departure is the moment when confusing them would become visible.

And the tasks the departing person had scheduled?

They stop when that person leaves, and that is a cost we chose to pay.

A scheduled task has nobody asking for it at the moment it runs. It acts with the permissions of the person who created it, and when that person leaves the company, keeping it running would mean running permissions that no longer belong to anyone. So we decided it stops, knowing what that costs: the weekly list of candidates to chase, useful to the whole team, ends on the day its author leaves. We explain that choice at more length in the runs nobody is watching.

The practical consequence belongs in the leaving procedure. An inventory of the person’s scheduled tasks is taken alongside the return of their laptop, during the notice period rather than on the last day, and every useful task is explicitly taken over by someone who agrees to own it. That transfer is a human act, and that is exactly what gives it value: somebody decides the task should carry on and becomes answerable for it.

Where to start before the next departure?

With a test that takes ten minutes: if your busiest business manager left tomorrow, could someone list, within an hour, everything they have promised to people outside the firm over the next three weeks?

If the answer is no, the problem is not the departure, it is already here, and it costs you every time that person goes on holiday. The fix is not one more handover form that nobody will fill in until they need it. It is a place where commitments are written down at the moment they are made, and where an account’s context builds up as the work is done rather than in an end-of-contract meeting.

It is one of the things we have Balt do: note the follow-up at the moment it is promised, and remind whoever owns it when it falls due. It does not replace the handover conversation, but it means that conversation is about judgement rather than memory. The next question mirrors this one and comes up the day the successor arrives: how you bring someone in, human or agent, so they do not start from zero.