How to replace a job without firing anyone

In a trade where a vacancy takes six months to fill and most firms are already turning down work, removing administrative work from somebody's week does not produce a redundancy. There is nobody to make redundant. It produces capacity in a business that is refusing orders, which is a completely different situation from the one the automation debate usually assumes.

The premise, with numbers on it

This argument only holds if the labour shortage is real rather than a talking point, so it is worth establishing before anything else.

In the German trades, the competence centre for skilled labour at the IW in Cologne counted roughly 107,700 open positions in 2024 that could not be filled from the pool of appropriately qualified unemployed people. That is the careful measure. The trades association's own estimate of total vacancies runs to at least 200,000, and looser figures circulate above that.

The duration numbers are the ones that matter more. Open positions in the sector stay open for something over seven months on average, roughly twice as long as a decade ago, at around 173 days. In bottleneck trades such as heating, ventilation and plumbing it runs beyond 240 days, and in parts of construction close to 300. Meanwhile more than six in ten craft businesses report having to decline orders.

Those are the figures the argument rests on, so they are worth seeing together rather than in a paragraph, particularly the two duration rows.

Read the last two rows together and the argument is finished before it starts. A business is turning work away, and the person who would let it accept that work cannot be found inside nine months. In that situation an hour recovered is not a saving looking for a justification, it is capacity with a queue already formed behind it.

So the question in this sector is not whether to protect jobs from automation. It is what to do with the hours that come back, and most owners have not decided, which is how the hours quietly disappear into other work rather than into revenue.

Being honest about when this does not apply

The argument above is specific, and generalising it would be dishonest, so here is the limit.

Where there is genuine duplication, automation does displace people, and pretending otherwise is not credible. A consolidator that buys several companies each with their own finance, marketing and support functions can and does run the combined entity with a fraction of the staff. Bending Spoons took on roughly 1,830 people through three acquisitions and has said it expects a few hundred to remain. That is a real strategy with real results and a real human cost, and it is a different mechanism entirely.

The distinction is whether the work you are removing is work somebody is employed to do, or work that is currently being squeezed into the edges of somebody's day by a person whose actual job is something else. In a fifteen-person trade business it is almost always the second.

Nobody in a small service company is employed to chase unanswered quotes. It is being done badly by somebody whose real job is dispatch, in the gaps, when they remember.

Which is why removing it does not remove a role. There was no role. There was a person doing four jobs, one of them poorly, and now they are doing three properly.

What to actually do with the hours

This is the decision, and leaving it unmade is the most common failure. Freed hours do not automatically become revenue; they become whatever is nearest, and what is nearest is usually more of the same low-value work.

1. Name the destination before you start

   Decide in advance where the time goes: billable work, the reactivation list, faster quoting, or a specific person's development. Written down, before the change, or it will be absorbed invisibly and you will never know whether the project worked.

2. Move the person up, not sideways

   The office manager who no longer types job sheets should be doing something harder, not more of what remains. Customer relationships, supplier negotiation, running the chase list properly. Those are worth more and they are also more interesting, which matters for keeping them.

3. Give the technicians back their evenings

   In many of these businesses the quoting and paperwork happen after hours, unpaid or resented. Returning that time is not a productivity gain that shows up anywhere, and it is one of the few genuine retention levers available in a trade where people can leave on Monday and start elsewhere on Tuesday.

4. Take the work you were declining

   If six in ten businesses in your sector are turning down orders, capacity converts directly into revenue at your existing prices with no marketing spend. This is the highest-value destination and it requires having decided to say yes to things you have been saying no to.

Returning the evenings is undersold and I would put it near the top. Retention in a trade with a 173-day average vacancy is not a soft metric, it is the constraint. Somebody who stops doing paperwork at nine in the evening is meaningfully less likely to answer a recruiter, and replacing them would take most of a year.

Saying it to the staff

How this is communicated determines whether it works, because a team that believes a system is being installed to replace them will make sure it fails, and they will be right to.

• Say what is being removed and why, in specific terms, naming the task rather than describing a transformation
• Say plainly that headcount is not being reduced, and only say it if it is true
• Say what the freed time is for, specifically, because a vague promise reads as a euphemism
• Let the people doing the work design the change, since they know the exceptions and you do not
• Do not run it as a pilot in one corner without explaining why, because an unexplained pilot reads as a test of who is expendable
• Report back on what actually happened, including the parts that did not work

The promise about headcount deserves care. If you cannot honestly promise it, do not, because a promise walked back once is worth nothing afterwards. And if the truth is that a role genuinely will change substantially, saying so early is both fairer and more effective than allowing it to be discovered.

The version I actually believe

The comfortable framing of this subject is that technology creates as many jobs as it removes, which is a claim about economies over decades and is useless to somebody deciding what to do in one company this quarter. The honest version at the level of a single business is narrower.

In a trade with a hiring shortage, the work being removed is mostly work nobody was hired to do, so the effect is capacity rather than redundancy, and the constraint moves from paperwork to whether you can find another qualified person. In a business with genuine duplication, usually one assembled from several companies, the effect is a smaller headcount and that should be stated rather than dressed up.

Both are true, they apply to different situations, and anybody telling you only one of them is selling something. What decides which situation you are in is not the technology, it is whether the work you are automating had somebody's name on it. Where the recoverable slice actually sits, and how bounded it is, is in the product cannot be disrupted, the operations can. The five activities in question are in the five things every offline business still does by hand, and what to deploy against them, in what order, is in where AI belongs in a business that answers the phone.

Vacancy figures from the Kompetenzzentrum Fachkräftesicherung at the Institut der deutschen Wirtschaft (Cologne) for 2024, alongside the Zentralverband des Deutschen Handwerks estimate of total open positions; vacancy durations and the share of firms declining orders from published German trades-sector reporting. Bending Spoons headcount figures from its 2026 public filings.