Our predictions for 2026 to 2030

1. 012026 to 2027High confidence

   Big companies keep growing without adding people

   We expect big companies to keep growing revenue while total headcount stays flat or falls, which breaks the old assumption that growth needs proportional hiring. Layoffs will stop being emergencies and turn into routine rebalancing, where a company cuts in one area, hires for AI-adjacent roles in another, and ends up at roughly zero net change.

   The evidence

   • Microsoft, Google and Meta all grew revenue by double digits in 2024 to 2025 while headcount stayed roughly flat.
   • Intuit's 2024 approach (cut 1,800, then hire 1,800 different people) has been copied across SaaS.
   • Executives at Salesforce, CrowdStrike and HP have all said publicly that AI productivity means they need to hire fewer people.

   Watch for

   • Big-tech earnings showing revenue per employee climbing sharply.
   • Companies describing hiring as a 'reallocation' instead of an expansion or a freeze.

   September 2026 check-in · On track

   Industry-wide counts passed 2025's full-year total on September 10 with almost four months left, and the companies doing the cutting (Uber, Apple, Meta, Oracle) all kept growing revenue. Meanwhile 78% of tech leaders plan to add headcount in H2 2026. Cut here, hire there, stay flat overall. That's the model we described.

2. 022026 to 2028High confidence

   Governments step in on entry-level hiring

   Junior hiring is down roughly half from 2019, and we think that stops being a problem for individuals and becomes a structural one. In 5 years there'll be a missing cohort of mid-level professionals, graduate underemployment will be higher, and governments will eventually pay attention, whether through apprenticeship subsidies, hiring incentives or curriculum reform.

   The evidence

   • Several datasets show entry-level postings at large tech firms falling about 50% between 2019 and 2025.
   • AI takes over exactly the routine tasks juniors used to train on, like boilerplate code, research summaries and first drafts of just about everything.
   • In 2024 to 2025, unemployment among recent graduates started running above the general rate, which reverses the historical pattern.

   Watch for

   • Universities reporting that placement rates for CS graduates keep falling.
   • The first big government programs that subsidize junior tech apprenticeships.
   • Companies rediscovering junior pipelines once mid-level talent gets scarce and expensive (around 2028).

   September 2026 check-in · On track

   Employers say they want experienced specialists rather than trainees. 65% report that skilled talent is harder to find, while junior postings stay depressed. 'Residency'-style entry programs are spreading company by company, and governments are funding AI talent pipelines. No major hiring subsidy for juniors has landed yet, though.

3. 032026 to 2027High confidence

   The extra pay for AI skills shrinks as everyone learns them

   We expect today's outsized salary premium for AI-literate professionals to shrink. The skill won't lose its value. It'll become universal, the way 'knows Excel' or 'can use Google' did, and what sets people apart will move from using AI to building with it, and from building to owning judgment over it.

   The evidence

   • The salary premium in AI-skill job postings peaked in 2024 to 2025, while AI tool adoption across all professionals is heading toward saturation.
   • 'AI-first' mandates like Shopify's and Duolingo's turn using AI into an expectation, so it stops setting anyone apart.

   Watch for

   • Job descriptions dropping AI fluency as a 'nice to have' because it's simply assumed.
   • Interview loops adding AI-collaboration tasks, the way they once added coding screens.

   September 2026 check-in · Mixed

   The premium hasn't compressed yet. PwC measured a 56% wage premium for AI-skilled roles, up from 25% a year earlier. Job descriptions are treating fluency as the baseline, as we forecast, but pay for people who build with AI keeps widening. We're early on the mechanism and, so far, wrong on the timing.

4. 042026 to 2029Medium confidence

   IT outsourcing firms stop growing their headcount

   Firms on the scale of TCS, Infosys and Accenture make their money selling human hours, and that headcount-billing model erodes as clients expect AI-driven delivery at AI-driven prices. We expect industry headcount to stay flat or decline through the late 2020s even if revenue holds, with the biggest absolute job losses landing in India.

   The evidence

   • TCS's 12,000-person cut in 2025 was the largest in Indian IT history, and it was explicitly about skills.
   • Accenture made AI reskilling a condition of employment while exiting more than 11,000 staff.
   • Clients are now negotiating outcome-based contracts that break the link between fees and staffing levels.

   Watch for

   • Net hiring at the big Indian IT firms turning negative and staying there.
   • Major contracts moving from time-and-materials pricing to outcome-based pricing.

   July 2026 check-in · On track

   Rackspace cut 15% in a 'shift toward AI focus' in June 2026, and managed-services restructuring kept going through H1. The headcount-billing model is eroding on schedule.

5. 052026 to 2030Medium confidence

   New AI jobs show up in numbers, but not for the same people

   As in every technology transition before this one, new roles will partly make up for the ones destroyed: AI operations, agent orchestration, AI safety and evals, human-AI workflow design, AI auditing and compliance. The catch is that the new roles need more skill than the ones they replace, and the people displaced won't automatically be the people hired.

   The evidence

   • The EU AI Act and sector-specific regulation are creating compliance roles that didn't exist in 2023.
   • 'Agent ops' and 'AI evals' titles started showing up in job postings in 2025.
   • History points the same way. ATMs led to more bank branches, and spreadsheets led to more analysts. The lag between displacement and new jobs is typically 5 to 10 years.

   Watch for

   • Titles standardizing. Once 'AI Operations Engineer' shows up at 100+ companies, the category is real.
   • Whether displaced mid-career professionals actually move into these roles, or whether the jobs go to a younger, AI-native cohort.

   September 2026 check-in · On track

   AI engineer is now the fastest-growing job title in the US, Indeed finds AI in 6.3% of all postings, and Robert Half says 'AI adoption and automation workflows' tops employers' hardest-to-fill list. You can see the new-role economy now. The open question hasn't changed: are the people being cut the same people being hired?

6. 062026 to 2028High confidence

   Companies keep cutting layers of middle management

   The layers between individual contributors and executives will keep getting squeezed. AI takes over the coordination work (status updates, reporting, scheduling, summaries) that used to justify a lot of management roles, and flat-org ideology now has cost-cutting economics behind it.

   The evidence

   • Meta, Amazon, Google and Microsoft all went explicitly after management layers in their 2023 to 2025 cuts.
   • Amazon required higher IC-to-manager ratios across the company.
   • AI tools for meeting summaries, project tracking and reporting overlap directly with what a coordination-focused manager spends the day on.

   Watch for

   • Span-of-control norms moving from about 6 to 8 reports toward 10 to 15.
   • Management job descriptions starting to ask for 'player-coaches'.

   September 2026 check-in · Ahead of schedule

   Uber's September memo turned the flattening into explicit policy: 20% fewer managers, half as many one- and two-person teams, and no role more than seven layers below the CEO, with some managers going back to individual-contributor work. Span-of-control norms are moving faster than our 2028 horizon assumed.

7. 072027 to 2030Speculative

   AI moves work between countries, in both directions

   Two forces pull against each other here. AI eats into the labor-cost advantage that built offshore hubs (why offshore work a model can do?). It also lets smaller, cheaper markets compete for higher-value work, since an AI-augmented engineer in Lahore or Lagos can get closer to Silicon Valley output than ever before. Our guess is that where you live ends up mattering less than what you can do, but this is a speculative call.

   The evidence

   • Stress in the outsourcing sector (TCS, service-desk automation) is hitting routine offshore work first.
   • Remote work plus AI tooling measurably narrows the productivity gap between markets.

   Watch for

   • Whether AI-native startups in the global south win international contracts at scale.
   • Whether the US and EU 'reshore' functions they once offshored, or the opposite happens and senior-level work globalizes further.
8. 082027 to 2030Medium confidence

   Politicians start writing rules for AI layoffs

   If white-collar displacement keeps going at the 2025 rate, expect a political response, such as rules requiring disclosure of AI-driven layoffs, retraining funds paid for by taxes on AI productivity, and employment-impact assessments for large deployments. Companies will stop bragging about 'AI efficiency' and start choosing their euphemisms carefully.

   The evidence

   • In 2025, executives were already walking back or softening AI-layoff statements after backlash (Klarna's reversal, IBM's clarifications).
   • The EU AI Act sets a template for obligations on AI deployment, and labor provisions would be the natural next step.
   • Every general-purpose technology transition so far (mechanization, offshoring) eventually produced labor policy.

   Watch for

   • The first jurisdiction to require AI-impact disclosure in layoff filings, along the lines of the WARN Act.
   • Union contracts adding AI clauses, which has already started in media and entertainment (the WGA/SAG template).

Published January 2026 · last reviewed September 2026. We score these against the data as it arrives, honest forecasting means being seen to be wrong sometimes. Predictions without a check-in are still too early to call.