AI Infrastructure Still Needs Skilled Hands
A SEGRO report puts unfilled vacancies in critical emerging roles at 21% to 46%. The same page rounds that range. An ad is not a skills test.
Christian Kosiek•Published •10 min read
The next job associated with AI may not look like a model-training job. It may be an electrician tracing a fault, a mechanic keeping a line running, or a technician who can shut a system down safely. Those people are not writing foundation models. They are the reason a building, a warehouse or a data hall stays up.
A report commissioned by SEGRO, published on 9 November 2026 and informed by Randstad Advisory, is the clearest public account of that claim this week. The supplied draft pointed to a Times article with the same date. On 10 November 2026 that Times URL returned 403, so none of the figures below are taken from the Times wording. They are taken from the SEGRO release, which is the document that story describes.12
Read the selected findings, not only the headline
The release analyses more than 400,000 job advertisements across about 13,000 companies in eight European markets: the Czech Republic, France, Germany, Italy, the Netherlands, Poland, Spain and the United Kingdom. The focus is 60 emerging roles the report calls "new collar": frontline, operational and technician jobs that combine practical experience with automation, digital, engineering or data skills. The sectors named are manufacturing, transport and logistics, defense, life sciences, data centers, retail and construction.1
The selected findings are more precise than the opening lines:
| What the SEGRO page says | Figure | How to read it |
|---|---|---|
| Selected findings: vacancies in the critical emerging roles that remain unfilled | 21% to 46% | Not one Europe-wide vacancy rate |
| Opening bullet on the same page | "up to half" | A rounding of that range |
| Body text on the same page | "just under a quarter" to "almost a half" | The same range, in words |
| Selected findings: share of the existing blue-collar workforce with the named next-generation skills | 9% | Not a test of every tradesperson |
| Opening bullet | "only 10%" | A round number next to the 9% |
| Body text | "less than 10%" | Compatible with 9%, not a third study |
| Site-dependent vacancies unfilled, versus location-flexible roles | about 40%, versus about 13% | Six sector comparisons, as the page states them |
| New-collar roles with a direct entry pathway | 22% of the 60 roles | A reason apprenticeships still matter |
An advertisement shows what an employer wrote down. It does not show that the post was filled, that the requirements were necessary, or that a person already on the payroll could not do the work with a structured bridge. A skills share built from occupation categories is not an assessment of individuals. The page does not publish a single definition of "unfilled" or one observation window for every vacancy. The full report file was not available as a separate download on the page we opened.
SEGRO is a listed property company that owns, operates and develops warehouses, industrial buildings and data centers, with a portfolio of 10.9 million square metres. The research is not a statistical-office release. That does not make the advertisement analysis false. It does mean the commissioner has a stake in industrial locations and in the workers those locations need.1
Paul Dunne, SEGRO’s managing director for operations, digital and customer, said businesses already have many of the people who could make this transition, but they need to give them a route to do it. The page also says employers cannot solve the shortage only by competing harder for the same pool.1
We already argued that a new infrastructure job is not a transfer plan for the person whose task changed. This article is the European vacancy question, not a second tour of the U.S. electrician projections.
A pathway is a list of tasks
The report names routes rather than a vague instruction to "learn AI". An electrician may move toward high-voltage or data-center electrical work. A maintenance mechanic may move toward mechatronics or automation. A diesel mechanic may move toward electric-vehicle fleet work. A warehouse operative may move toward digitally enabled logistics. The page says only 22 percent of the 60 roles offer a direct entry pathway, which is why it points to vocational education, apprenticeships and time on the job.1
The transferable base, in the electrician example, is already specific: electrical safety, reading drawings, isolating a fault, planned maintenance, and a record someone else can trust. The gap may be critical-power redundancy, cooling dependencies, monitoring rules and a change procedure that does not allow improvisation. An employer who advertises only for a finished specialist will miss the electrician who needs that bridge.
A second illustration, not a placement the report verified: a warehouse mechanic whose site adds machine vision and autonomous conveyors. Bearings, motors, sensors and safe isolation still matter. New work is reading an alarm, checking whether the data are usable, and writing down when an automated decision should be overridden. Neither example is a promise that a short course produces a job.
Three practical barriers sit outside the percentages. The new site may not be where the person lives. A license or authorization the law already requires is not replaced by familiarity with a tool. And a course without paid time, a named assessor and a real vacancy at the end is a certificate, not a transition. Working beyond the training is not the same as mastery.
Construction peaks while a facility is being built. The operations team that stays can be much smaller. A headline that mixes temporary building work with permanent posts is not a local career guarantee. Separate the estimates.
The page gives one comparison that smaller firms should not skip, and should not over-read. As an indicator, 72 percent of employees in large companies received training in emerging technologies, versus 17 percent in small and medium-sized enterprises. The release does not attach a sample size or a year to that split. It is a reason to ask who can reach training, not a Europe-wide census of training hours.1
A tech-talent survey is a different workforce
On 29 October 2026 Randstad published a separate release, "Tech talent in transition". It is easy to drop next to the SEGRO figures because both carry the Randstad name. They are not the same population.
The tech release says 86 percent of organizations are growing their tech teams as AI reshapes demand. It also says 61 percent of employers report that finding talent with communication and stakeholder-management skills is harder than sourcing technical skills alone. A later paragraph on the same page restates 61 percent as human capabilities such as critical thinking and communication. Treat that as one claim with two wordings, not as two measurements.3
The primary survey was conducted in the first half of August 2026 in 11 markets, with 1,100 interviews, 100 in each market, among senior business and technology decision-makers. Sampling used Evalueserve. Secondary material from Randstad Market Intelligence covered more than 20 million job postings and about 100 million candidate profiles. That is a study of technology talent and of people who hire it. It is not a study of electricians, mechanics or warehouse technicians.3
The International Labour Organization publication of 13 August 2026, cited from its landing page only, says AI adoption changes how workers use cognitive, socioemotional and physical skills, and that jobs which develop and maintain AI systems are still a small niche even where they are growing quickly.4 A list of occupations next to an investment boom is not evidence that AI created each post, or that the post will last. An AI line in a job advertisement is not, by itself, an AI job.
What a small employer can try
Pick one role that is likely to change in the next year. Break it into tasks a supervisor can watch: diagnosing a failure, reading an alarm, inspecting equipment, recording a safety check, handing work to the next shift. Mark which tasks stay, which become software-assisted, and which need a standard the team does not yet use. "AI fluent" is a weak requirement when the real need is to validate a sensor or to respond to an alert.
Ask experienced people to show the skills they already use. A maintenance worker may isolate faults well and still have little practice with networked controls. Train the gap. Where a license is already required, do not describe an internal course as a substitute for it.
A workable pilot can combine paid shadowing, vendor instruction, one simulated fault and an assessed task on the real system. A technician explains a safe shutdown, identifies the alarm and documents the escalation. There is a named assessor and protected time. A certificate of attendance is weak evidence of that sequence. We have made the same distinction for apprentices whose tasks grew faster than their supervised judgment.
Then count outcomes other than attendance: people assessed, gaps closed, tasks performed safely without help, time to proficiency, and moves into open roles. Also count the transitions that failed, and why. Certificates without a safer task or a real opening are a reason to redesign the pilot.
If an employer in the EU uses an AI system to filter who is selected for the bridge, that purpose can fall under Annex III, point 4(a), on recruitment and selection. The Commission’s service desk says those high-risk rules apply from 2 December 2027.56 A supervisor watching a shutdown drill is not that system. This is not legal advice.
What a local partner can add
A vocational provider or a community organization can translate a jargon-heavy vacancy into a short bridge, share equipment a small firm cannot buy, and tell a worker when the vacancy is not real. The useful unit is a demonstrated task plus a local employer, not another general introduction to AI.
A regional partnership that compares recurring maintenance vacancies with the skills of experienced tradespeople, then funds a short assessed bridge, is a design. It is not an evaluated programme. Judge it by placements that last, by pay, and by whether people stay, not by enrolment.
How hilighter can help
The practical step is smaller than "retrain Europe". One emerging task, one person with adjacent skill, one supervised route, one record of what they can now do safely.
hilighter helps a team map skills by role, compare people with those requirements in Skills & Talent Insights, keep proof of capability in Skills Evidence & Recognition and set personalized growth paths. The AI Job Impact view is indicative. It is not the SEGRO vacancy range, and it does not promise a data-center job. A free plan covers up to three job roles, with unlimited employees.
Sources
- SEGRO, ‘New Collar’ worker shortage threatens to stall next industrial revolution, 9 November 2026. Commissioned report, research with Randstad Advisory and Randstad Enterprise. Read 10 November 2026.
- The Times, "Retrain blue-collar workers or miss out on the AI revolution", 9 November 2026, URL supplied in the draft: workers-retrain-ai-data-centre-skills. On 10 November 2026 the page returned 403. No figure in this article is taken from it.
- Randstad, 86% of employers are growing tech teams as AI reshapes demand, 29 October 2026. Survey of 1,100 senior decision-makers, first half of August 2026, 11 markets. A technology-talent study, not the SEGRO trades analysis. Read 10 November 2026.
- International Labour Organization and partners, Changing landscape of skills in the age of AI, 13 August 2026. Claims here are from the publication page. The PDF was not read line by line. Read 10 November 2026.
- European Commission, AI Act Service Desk, Annex III, point 4(a). Read 10 November 2026.
- European Commission, AI Act Service Desk, When does enforcement start?. Annex III high-risk rules apply from 2 December 2027. Read 10 November 2026.
Sources last reviewed: 10 November 2026. The 21% to 46% range and the 9% skills share are selected findings on a company press page, not a Eurostat release. The 72% and 17% training split has no sample on that page. The warehouse-mechanic scene is an illustration. This is not legal advice.