Savings alone shouldn’t decide who you lay off because of AI this year

At the beginning of this year, one European media company laid off just under a fifth of its in-house production team, and management justified the decision by rolling out AI. Three months later, it was hiring the same people back at higher salaries. They kept their severance because there was no way to take it back, and no one can return the quarter of a year when production ran without them.

The calculation management used to justify the layoffs in January probably projected salary savings years into the future. What it did not show was which competencies would leave the company with the people being let go. The savings lasted a quarter, and no one had included what replaced them in April in the original calculation.

Someone who has been told by a company that AI can do their job also comes back at a higher price and, in my experience, with less loyalty than when they left. Sometimes a company has to cut costs and layoffs are the only option. What I think is a bad idea is making that decision based on cost alone, because the savings end sooner than the consequences the decision sets in motion.

In this article, we will look at why no one reopens a competency model once it has been approved and why AI makes that more costly than it used to be. And above all: what management needs in front of it to see not only the savings, but also what it will pay for elsewhere.

A competency model gets built once, and no one reopens it when the company changes

When management decides on layoffs, hiring or what people will be evaluated against, the company then tracks whether the decision was carried out. In fifteen years in HR, I have never heard anyone in a meeting ask whether the savings that justified the layoffs are still savings a year later. Or whether the company is still evaluating what it needed at the time the evaluation criteria were set.

Competency models survive like this the longest. In companies where I have seen them, they were created as a project, presented to management and then used for evaluation from that point on. The company reorganized in the meantime and the content of some roles shifted, but the model still describes the state of the company when the project was created, because no one set a date for reopening it when it was approved.

The company then keeps evaluating and hiring for work that no one does anymore after the reorganization. The consequences may only show up years later, as a bad hire or a promotion that does not fit. By then, the blame falls on the manager or the candidate, because no one remembers the competency model the decision was based on.

The steam engine gave companies generations to adapt. AI gives them years

The steam engine redrew the division of labour between people and machines, and companies had generations to adapt because the technology reached one workshop after another over decades. AI is redrawing the same division, but companies have only a few years. Which work machines take over will be determined by technology, its price and time. What management can decide is how quickly its people can adapt when the work changes.

The methodology behind competency models was not built for this pace. It emerged in the 1980s and 1990s, when roles often stayed unchanged for years, so a model could be built once and reviewed every few years.

As long as the content of work changed just as slowly, companies could get away with leaving a poorly defined role untouched for years. An outdated job description would nudge hiring and evaluation slightly off course, and that deviation never appeared as a separate line item.

Would you make a layoff decision today using a document created two reorganizations ago?

Today, companies lay people off, hire and promote several times a year, each time using job descriptions and evaluation criteria created for earlier decisions. Hiring from a two-year-old job description brings someone into work that has already shifted.

Evaluating against pre-reorganization criteria promotes people who are good at what the company needed then. Each decision can be defended in a meeting, but together they slow the company down at the very moment it may need to redeploy people within a month.

What the media company was missing in January was a competency map

A competency map is a record of what people in the company can do today, and it becomes outdated every day no one records a change. A competency model describes what people should be able to do and remains valid from approval until the next review. For decisions about layoffs or reshaping a department, the map is useful because it reflects the competencies the company actually has in its people.

TIP: I explain the difference between a map and a model in the Unmapped newsletter.

In January, the map would have shown the media company from the introduction two things: which part of production the machine could do faster and more cheaply than the people doing it, and which of those people could move into the work that the machine would create instead – managing it, checking its outputs and doing the people-focused work there had not been time for before. 

Management might then have laid off fewer people, or different people, and above all would have known what was leaving the company with each person.

Why Sloneek connects competencies, feedback and goals

At Sloneek, we have built it so competency evaluation, 360° feedback and goals all run in one system for the same people. When management needs to reshape a department within a month, it opens a record that managers and colleagues have been updating throughout the year.

It can decide just as quickly as the media company mentioned above, but it also knows who it is laying off, who it might need to hire back three months later, and who should be offered a different role instead of a termination.

Before you start cutting, find out what you would lose

When management decides who to replace with AI, it is not enough to know how much that person’s work costs today. It also needs to know which competencies, experience and options for redeployment would leave the company with that person. Otherwise, a saving on paper may simply move the cost a few months down the road.

Two questions to end with. Which job description or evaluation criterion in your company has been in place for more than two years without anyone reopening it? And what would you use to decide if you had to reshape a substantial part of one department within a month, perhaps because of AI?