Pharma Bloodbath Part XXIII: AI was blamed for 120,136 job cuts this year. Now it’s applying for the jobs that are left, in your name

Dear MoreThanCareer.de community,

It’s 3:14 a.m. in Offenbach. A laptop in front of an empty chair applies for a Clinical Trial Assistant role, its 41st application of the night. At 3:16, the employer’s screening tool sorts it into “no”, while the recruiter who owns that tool sleeps two towns away.

I made that scene up. Every ingredient in it is real and footnoted below.

Save this article before you buy a bot to job-hunt for you, or before your employer buys one to do your job. Then repost it to the colleague who keeps saying AI is “just a tool”. The tools have started filling in application forms.

For newcomers: I’ve spent 17 years in recruitment, the last nine at Bayer and the five before that at the global top-3 CROs ICON and Syneos. I’ve hired in most EU countries and the US, up to SVP level. A declaration of interest, in keeping with this issue’s film: I’m the man in the bunker who trains the resistance, and I charge for it. Everything I offer is on www.morethancareer.de: CV and LinkedIn rewrites in English and German, coaching on job search strategy, interviews and salary negotiation with tailored scripts, and reverse recruitment for roles in Germany. Check my footnotes the way you’d check a stranger who claims he was sent from the future.

Judgment Day arrived quietly and filled in the application form

Meet Horst. He’s invented, which already puts him ahead of a fair share of tomorrow’s candidates (more on that later). Horst is an auto-apply agent, model T-800, applications department. He finds job ads, rewrites your CV for each one, fills in the forms and presses “submit” while you sleep.

According to Fortune’s reporting on Greenhouse, tools that do this for Greenhouse-hosted jobs cost a candidate about $20.¹ Arnold Schwarzenegger needed a time machine to reach his target. Horst needs a credit card.

The result is roughly what you’d expect if everyone at Frankfurt Hauptbahnhof (main station) were handed a megaphone. Greenhouse counts around 175,000 live jobs with an average of 254 applicants each, and applications per recruiter are up 412%.¹ Ashby, another applicant tracking provider, measured about 291 applications per hire, against roughly 100 in early 2021.²

If 291 people queued for one appointment at a German Bürgeramt (citizens’ registration office), the police would be called. On a careers site, it goes into the annual report as employer branding.

Greenhouse’s CEO Daniel Chait calls the result a “doom loop”.¹ Candidates use AI to apply everywhere, recruiters use AI to cope, and rejected candidates answer with more AI. Die Katze beißt sich in den Schwanz (the cat bites its own tail). In 2026 the cat and the tail run on separate subscriptions.

Skynet’s HR department: some rejections now come without a human attached

I’ve spent three issues burying the myth that robots bin 75% of CVs. That myth stays buried. Its smaller, real cousin has now turned up, with a signed confession.

On 14th October 2026, Indeed published a survey of 300 German HR decision-makers. 56% receive more applications per vacancy than a year ago, and they estimate that about a third of those applications are written with AI. 42% officially use software or AI to pre-select. Among those, 11% admit that automatically rejected applications are no longer checked by a human before the rejection goes out.³

Only about half of the software users check every automatically sorted-out application, and 35% of all respondents spend less than a minute per application.⁴ t3n points out that rejecting applications without human review is prohibited under the GDPR.⁵

In plain terms: in roughly one in nine of those companies, a machine decided your rejection and sent it. The only human in the process was you, reading it at breakfast. Holzauge, sei wachsam (wooden eye, stay alert), as the old German warning goes. It now applies to your inbox.

The Americans are no calmer. In data reported by Axios, only 21% of recruiters were “very confident” their systems weren’t weeding out qualified candidates.² The other 79% hope for the best, which, by coincidence, is also the candidates’ strategy.

Keep one number in proportion. In a representative Bitkom survey of 602 German companies, published on 28th September 2026, only 6% use AI for the first review of applications.⁶ Most German employers still read with human eyes and a lukewarm filter coffee. The bunker is smaller in Germany. It exists.

The machines prefer each other

Researchers at the University of Maryland asked an awkward question. If an AI writes your CV and another AI reads it, does the reader play favourites?

With more than 2,200 CVs, large language models preferred CVs they had generated themselves over human-written ones in roughly 67% to 82% of comparisons, depending on the model. In simulated hiring across 24 occupations, candidates who used the same model as the employer’s screener were 23% to 60% more likely to be shortlisted than equally qualified people who wrote their own.^7,8

Gleich und gleich gesellt sich gern (like and like keep company), and that apparently includes things made of matrix multiplication. Germans suspected as much. In a Statista+ survey from February 2026, 44% worried that AI screening might favour AI-optimised documents.⁹ We saw it coming and filed it under “interesting”, the German national response to approaching weather.

There’s a quieter casualty too. On Freelancer.com, employers receive around 60 applications within minutes.¹⁰ After the platform launched a one-click AI cover letter tool in April 2023, cover letters became less correlated with who actually got hired.¹¹ Once every applicant can generate the perfect letter in a second, the letter stops telling the employer anything.

In Germany, Horst’s carefully tuned love letters mostly land on a human desk anyway. 41% of German HR decision-makers say they bin applications that were obviously written entirely by AI.³ Horst expected a hot date with an algorithm and got Frau Krause from Personnel. Pustekuchen (puff cake: no chance).

First they apply for your job. Then they do it

Here’s the plot twist the auto-apply ads leave out. While you use an agent to apply for jobs, employers are testing agents to do those jobs.

In the US, Challenger, Gray & Christmas counted 120,136 announced job cuts attributed to AI through September 2026, about 21% of all cuts, which makes AI the leading reason this year.¹² For the whole of 2025, the figure was 54,836.¹³ Nine months of 2026 have more than doubled all twelve months of 2025.

Each announcement comes with Krokodilstränen (crocodile tears) and the phrase “investing in AI”. Challenger itself notes that what counts as an AI cut is ambiguous.¹⁴ In my experience, “AI” on a restructuring slide is sometimes the reason and sometimes the alibi. Your desk is empty either way.

Young people feel it first. Stanford’s Digital Economy Lab tracks US payroll data in a project called “Canaries in the Coal Mine”.¹⁵ Its August 2026 update puts employment of 22- to 25-year-olds in the most AI-exposed occupations about 19% below where it would be had it kept pace with less exposed peers.

The gap comes mainly from fewer hires, and experienced workers in the same occupations show no comparable decline.¹⁶ Den Letzten beißen die Hunde (the dogs bite the last one through the gate). In this market, the last ones through the gate are 23.

Germany, as usual, is slower and more complicated. On 13th October 2026, the IAB (the Federal Employment Agency’s research institute) reported no general employment slump in AI-exposed occupations so far. It found a slightly negative trend for 20- to 29-year-olds and named the weak economy as the bigger driver.¹⁷ Reporting on the study put the economic effect at more than twice any plausible AI effect.¹⁸

Indeed’s Hiring Lab has the less comforting number. Entry-level job ads in July 2026 sat 46% below their 2019 level, while the overall market was still 8% above it. In occupations where AI could take over or support more than 60% of tasks, beginner ads ran at roughly half their 2019 level.¹⁹

Den Teufel an die Wand malen (to paint the devil on the wall: to tempt fate by predicting the worst) is something I’ll resist. Germany hasn’t had its Judgment Day. It has had a hiring freeze for beginners, which looks rather similar from inside a student flat.

Pharma isn’t immune. It just moves at the speed of a validated system. IQVIA’s global practice lead for pharmacovigilance technologies has described the goal as saving 50% of pharmacovigilance cost while getting quality above 99%.²⁰ If you process adverse event cases for a living, that sentence is a headcount plan in a lab coat.

Software vendors already pitch voice-screening agents to pharma talent teams for CRAs and field reps.²¹ Your first interview for a CRA role may be with a machine, and so, eventually, may your recruiter’s exit interview.

Now the good news, which I’d frame above every pharma graduate’s desk. In Indeed’s category for nursing, medicine and pharmacy, entry-level postings grew 18% since 2022.²² Apprenticeships and dual study programmes ran about 50% above 2019.¹⁹ The machines are coming for the work that fits on a screen. Work that needs hands, a clean room or a signature is holding up better.

Ein Wolf im Schafspelz (a wolf in sheep’s clothing): the T-1000 problem

In Terminator 2, the deadliest machine is the one that can look like anybody. Hiring has met its T-1000.

Gartner predicts that by 2028, one in four job candidates worldwide could be fake, with AI-generated faces, invented histories and chatbot-whispered answers.²³ In the UK, 61% of employers in the Institute of Student Employers’ 2025 survey reported candidates using AI during interviews without disclosure or permission.²⁴ In its latest survey, 92% suspected or confirmed student AI use somewhere in assessment and selection.²⁵

Employers have answered with the most advanced weapon available: a table, two chairs and a person. Google, Cisco and McKinsey brought back in-person interview rounds to check skills and catch fraud.²⁶

Meanwhile, 63% of job seekers in Greenhouse’s 2026 multi-market survey, which included Germany, had been interviewed by an AI, and 70% weren’t told AI was involved.²⁷ So you may sit opposite a machine whose job is to find out whether you’re a machine. Kafka would have asked for royalties.

Horst, incidentally, got his first invitation this week. In person, Frankfurt, 9:00. Horst has no legs and believes Frankfurt is a keyword.

The resistance, country by country

Andere Länder, andere Sitten (other countries, other customs). The machines are the same everywhere. The damage has a local accent.

• United Kingdom: graduate employers received 141 applications per graduate role, and 172 in London. Internship applications rose 27% while internship and placement hiring fell 8%.²⁵ The British have met the robot uprising with their national weapon: a longer queue.
• United States: AI leads the list of reasons for layoffs,¹² and recruiters handle 291 applications per hire.² America runs the experiment first and sells the survival kits afterwards.
• European Union: the AI Act’s high-risk obligations for recruitment AI were pushed from August 2026 to 2nd December 2027.²⁸ Brussels gave HR software sixteen extra months to sort out its paperwork. Candidates got no extension on their rent.
• China: a record 12.7 million university graduates entered the market in 2026, and the government has said it will use AI as part of its employment strategy.²⁹ Youth unemployment stood at 15.6%, according to a summary of The Guardian’s reporting.³⁰ Beijing’s answer to AI eating graduate jobs is more AI, which at least has internal consistency.
• India: hiring is strong, yet 74% of recruiters told LinkedIn it has become harder to find qualified people, and 53% of those blame AI-generated applications.³¹ 94% of Indian job seekers planned to use AI in their search.³² Plenty of jobs, and the recruiters still can’t find the humans inside the pile.

How to fight Skynet: five moves with evidence behind them

“Wer kämpft, kann verlieren. Wer nicht kämpft, hat schon verloren” (whoever fights may lose; whoever doesn’t fight has already lost). Germans love to credit the line to Brecht. The attribution is shaky. The arithmetic holds.

The basics from earlier parts still apply. This list starts where they stop.

1. Apply less, and make each application expensive to fake

The commodity advice says job hunting is a numbers game, so send more. Horst sends more than you ever will, for $20.

The specific action: once AI makes every letter sound competent, the signal moves to what a bot can’t invent for you.¹¹ UK careers services now tell students to send fewer, better-targeted applications.³³ Apply only where you meet the core requirements. For each role, write proof points only you could write: a number, a named context (therapeutic area, study phase, system, market) and a decision you made yourself.

This is how I run reverse recruitment: a few well-matched applications per week instead of dozens of maybes. One Clinical Trial Manager went from zero interviews to four from five applications. That’s an exceptional result, not the norm, but most of my clients see a clear increase once quality replaces volume.

2. Reprogram your Terminator

The commodity advice says either “let AI do it all” or “never touch AI”. In Terminator 2, the same machine model turns from hunter into bodyguard once someone reprograms it.

The specific action: let AI decode the job ad, research the company and tear your draft apart. Write the final version yourself, and keep any agent away from the “submit” button. In my experience, recruiters spot a clumsy AI rewrite within seconds, and the giveaways go far beyond vocabulary: structure, rhythm and suspiciously perfect symmetry. 41% of German HR decision-makers say they bin applications that were obviously written entirely by AI.³ Read every application aloud, and cut any sentence you wouldn’t say to a colleague over lunch.

3. Make sure a human reads you

The commodity advice says stuff your CV with keywords to please the software. Keyword soup is precisely what makes your CV look like Horst wrote it.

The specific action: write for the person, with clear job titles, dates in month and year, and results in numbers. A machine can parse that easily, and a human can believe it. When a rejection from an EU employer arrives minutes after you applied, ask politely whether a person reviewed it. Unchecked automatic rejections are prohibited under the GDPR,⁵ and a polite question costs you one email.

4. Train for the room where the machine can’t follow

The commodity advice says practise your elevator pitch. The pitch gets you to the room. Your answers decide what happens inside it.

The specific action: structured interviews are the strongest single predictor of job performance in the largest recent re-analysis of selection research,³⁴ and in-person rounds are coming back.²⁶ Map each competency in the job ad to one story with a measurable outcome, and rehearse it aloud with a human who interrupts you. AI first rounds are common now,²⁷ so test your camera and microphone and ask the employer beforehand how the first round works.

5. Become the human the machine needs

The commodity advice says “upskill in AI”. That’s what everyone’s LinkedIn headline already claims.

The specific action: Stanford’s data shows employment flat or rising in occupations where AI mostly complements workers, especially for experienced staff.³⁵ List the tasks of your target role and mark which ones an agent could do this year. Then aim for the roles that carry judgment and a signature: patient safety decisions, GMP batch release, regulatory interpretation. Bring one real example to the interview of an AI error you caught. Supervising the machine is a job description in the making.

Who should read which part twice

Juniors and graduates, including those who need visa sponsorship: the entry rung is shorter, and the pharma rung is one of the few that grew.²² My reading of the data is that regulated, hands-on entry roles in QA, QC, production and clinical operations are the safer bet, because Horst can’t hold a pipette. That’s my interpretation, not a measured result. With fake candidates on the rise, have your degree recognition and permit documents ready before the first interview, so nobody has to wonder whether you exist.

Professionals with 2-3 years up to senior: the tasks you did as a junior are the ones being automated first. Your CV should show the layer above them: the call you made, the deviation you closed, the audit you survived.

Executives: application noise rarely reaches your level, but you decide where agents replace teams. Keep Challenger’s caveat in mind before an “AI” label goes on your own restructuring slide.¹⁴ And in a market full of deepfakes, your own next move depends even more on people who have met you in person.

50 and over: Die Flinte ins Korn werfen (to throw the shotgun into the corn: to give up) is the one move I’d ask you to skip. For once, the data is on your side. Experienced workers in AI-exposed occupations show no comparable decline in Stanford’s numbers.¹⁶ In Part IV I suggested selling your experience as risk management. The bots have just supplied the evidence: you’re the person in the room who knows when the machine is confidently wrong.

For recruiters and hiring managers: finding humans in the rubble

Wer die Wahl hat, hat die Qual (whoever has the choice has the agony), and recruiters in 2026 have choice by the lorryload. What works:

• Score structured interviews against the role’s real tasks. They lead the validity tables.³⁴
• Keep at least one live, identity-checked step, in person for final rounds wherever you can.^23,26
• Put a human on every automatic rejection. 11% of German software users admit they don’t, and t3n reports it’s prohibited.^3,5
• Tell candidates where AI sits in your process. In Greenhouse’s survey, 70% weren’t told, and when AI interviews were done well, 38% came away with a more positive view of the employer.²⁷
• If a language model ranks your CVs, test it for self-preference. The Maryland team cut that bias by more than half with simple interventions.⁸
• Swap the cover letter for one role-specific question a bot can’t answer without the candidate’s real experience. 46% of German HR decision-makers want shorter formats without one anyway.⁵
• Put 2nd December 2027 in the calendar.²⁸ Compliance documentation written in a panic in November 2027 will read exactly like it.

If Horst can pass your process, Horst will pass it, several thousand times a night.

Horst’s last application

Horst was eventually reprogrammed. He now researches companies, criticises drafts and stays away from the “submit” button. He applied once more, for a senior talent partner role. He was screened out at an early stage because he couldn’t explain why a CRA would rather drive four hours to a site than spend one more day in the eTMF (electronic trial master file).

He did leave a note for the hiring team, in his best Austrian accent. Three words, which you can probably guess.

Your turn

A like tells me you scrolled. A comment tells me you’re still fighting. Pick one:

• Job seekers: have you received a rejection so fast that no human could have read your application? How many minutes was it?
• Recruiters: what gave away the last application or interview you were sure a bot produced?
• Pharma professionals: which task in your current job could an agent do by next year?
• 50+ readers: has your experience become easier or harder to sell since AI arrived in your function?

Save this before your next application. Repost it to someone who’s about to buy their own Horst, or to the manager about to buy one for their team. Subscribe to this newsletter so the next part finds you.

For more than LinkedIn shows you, join the Inner Circle at https://morethancareer.de/#newsletter.

Subscribers get 15% off my upcoming book, “Surviving the Pharma Bloodbath: The Ultimate Job Search Guide for Pharma in DACH & US”, and 5% off coaching and reverse recruitment. In future, I’ll also share career tips there that won’t appear on LinkedIn.

You can follow MoreThanCareer.de on LinkedIn too: linkedin.com

Want a human in your corner? On www.morethancareer.de you’ll find CV and LinkedIn rewrites in English and German, coaching on job search strategy, interviews and salary negotiation with tailored scripts, and reverse recruitment for roles in Germany: https://morethancareer.de/#programmes. Written by a human. Horst was not consulted.

Prost. Und lasst euch nicht wegrationalisieren. (Cheers. And don’t let anyone rationalise you out of a job.)

#Pharmajobs #CareerAdvice #AIagents #FutureOfWork

Sources

Primary and independent sources (some via secondary summaries, as marked)

[6] Bitkom press release, 28th September 2026: Führt künftig KI das Bewerbungsgespräch?: bitkom.org

[7] University of Maryland, Maryland Today: Study: AI used in hiring is biased toward itself: today.umd.edu

[8] Xu, Li, Jiang: AI self-preferencing in algorithmic hiring (arXiv): arxiv.org

[10] Tuck School of Business (Dartmouth): What happens to hiring when AI writes job applications? (Galdin and Silbert): tuck.dartmouth.edu

[11] Signaling in the age of AI: evidence from cover letters (arXiv): arxiv.org

[12] Fast Company, on the Challenger, Gray & Christmas September 2026 report: fastcompany.com

[13] Challenger, Gray & Christmas: February 2026 report (full-year 2025 AI-attributed cuts): challengergray.com

[14] Challenger, Gray & Christmas: July 2026 report (ambiguity of AI attribution): challengergray.com

[15] Stanford Digital Economy Lab: Canaries, interest rates and timing: digitaleconomy.stanford.edu

[16] Brian Heger: summary of the Stanford “Canaries” August 2026 update (secondary): brianheger.com

[17] IAB press release via idw, 13th October 2026: KI bisher kein Jobkiller, aber Berufseinsteiger zunehmend betroffen: idw-online.de

[18] BornCity: IAB-Studie, kein Beleg, dass KI Berufseinsteigern die Jobsuche erschwert (secondary): borncity.com

[23] NBC New York / CNBC: Gartner prediction on fake job candidates by 2028: nbcnewyork.com

[24] Times Higher Education: ISE Student Recruitment Survey 2025 coverage: timeshighereducation.com

[25] Times Higher Education: ISE survey 2025-26 coverage: timeshighereducation.com

[26] OECD.AI incident record: return of in-person interviews at Google, Cisco, McKinsey: oecd.ai

[28] Gibson Dunn: EU AI Act omnibus agreement, postponed high-risk deadlines: gibsondunn.com

[29] Yahoo Finance / Investing.com: China pivots to AI to manage record 12.7M graduate influx: finance.yahoo.com

[30] UDLAP Observatorio Global, summarising The Guardian (13th July 2026): China’s graduate glut (secondary): observatorioglobal.udlap.mx

[33] University of St Andrews Careers Centre: Graduate recruitment in 2025: careers.wp.st-andrews.ac.uk

[34] SIOP: Is cognitive ability the best predictor of job performance? (Sackett et al., 2022): siop.org

[35] Let’s Data Science: Stanford finds no broad AI job displacement (secondary): letsdatascience.com

Vendor, platform and directional sources (commercial interest or self-reported surveys)

[1] Yahoo Finance, on Fortune’s Greenhouse reporting: Greenhouse CEO calls it a “doom loop”: finance.yahoo.com

[2] Axios, 21st August 2026: Applying for jobs is easier than ever. Getting one isn’t (Ashby data, recruiter confidence): axios.com

[3] infranken.de (dpa), on the Indeed/Censuswide survey of 300 German HR decision-makers, 14th October 2026: infranken.de

[4] onlinemarketing.de, on the same Indeed survey: onlinemarketing.de

[5] t3n, on the same Indeed survey: t3n.de

[9] Statista+ press release, February 2026: Grauzone KI-Bewerbung: de.statista.com

[19] ingenieur.de, on Indeed Hiring Lab’s entry-level analysis of 23rd September 2026: ingenieur.de

[20] Drug Discovery & Development: IQVIA’s AI vision is to cut pharmacovigilance costs by 50% (vendor statement): drugdiscoverytrends.com

[21] Phenom webinar page: pharma AI agents hiring strategy (vendor marketing): phenom.com

[22] ms-aktuell, on the Indeed and IAB data (nursing, medicine and pharmacy +18%): ms-aktuell.de

[27] Greenhouse newsroom, 1st May 2026: 63% of job seekers have faced an AI interview: greenhouse.com

[31] Outlook Money, on LinkedIn’s India recruiter research: outlookmoney.com

[32] Careers360 (PTI), on LinkedIn’s India job seeker survey: news.careers360.com

© 20th October 2026, Andreas Schulz. All rights reserved.