On a Monday morning in late October 2026, Microsoft sent a memo to approximately 60,000 of its employees. The memo was written by Chief People Officer Amy Coleman and it described, in carefully considered corporate language, something that had never happened before in the company's fifty-year history: a voluntary buyout offered to more than a quarter of the entire global workforce.
Sixty thousand people. That is not a restructuring. That is a signal.
The offer — severance pay, continued stock vesting, six to twelve months of healthcare, and career services — is available to employees in the United States, United Kingdom, Canada, and India. Workers have until early December to decide. Those who accept leave before the year is out. The program, Microsoft said, gives employees who are "considering leaving" a new choice to make that transition with support.
The polite framing does not obscure what is happening. Microsoft has opened 31 new data centres in a single quarter. It has unveiled its Maia 300 inference chip, which it claims delivers forty percent more AI compute per unit of energy than leading alternatives. It has spent billions building the infrastructure to run AI at scale. And it has concluded that it needs to reduce the number of humans on the payroll to fund that buildout.
The math is that direct. Revenue grew eighteen percent year over year to $90 billion. Profit is rising. The company is not cutting because it is struggling. It is cutting because it has decided that certain categories of human labour are being replaced by something faster and cheaper, and it would rather pay for the infrastructure than the headcount.
This is the clearest statement yet from the largest software company in the world about where it believes the value in technology now lives. Not in the people who operate systems. In the systems themselves.
It Is Not Just Microsoft
Microsoft is the loudest signal, but it is not an isolated one. In June 2026, Google offered buyouts to select employees. Amazon laid off approximately 30,000 workers across late 2025 and early 2026. Microsoft itself had already cut 4,800 roles in July and more than 15,000 in 2025. Phil Spencer, who ran Xbox for a decade, retired. Ryan Roslansky, who built LinkedIn into one of the world's most important professional networks, announced his departure. These are not middle-tier managers being shown the door. These are the people who built the modern tech industry.
What is happening is a generational handover — except the handover is not to the next generation of humans. It is to infrastructure. The engineers, product managers, marketers, and strategists who built these platforms are being quietly told that the platforms can now run with fewer of them. For the professional working in technology — or adjacent to it — in 2026, there is a version of this story that is easy to tell and easy to believe: AI is taking jobs, the industry is contracting, and there is nothing to be done except update your CV and hope. That version of the story is incomplete. And the part it leaves out is the part that changes everything.
The Part Nobody Is Telling You
Here is what the Microsoft memo did not include, because no corporate restructuring memo ever does: a description of the problem that Microsoft, Google, OpenAI, Anthropic, Meta AI, and every other company building AI systems at scale is currently unable to solve by building more AI.
They cannot get their models to stop making mistakes.
Not random noise mistakes. Plausible, confident, well-formatted, professionally worded mistakes. The kind of error that sounds correct unless you already know enough about the subject to recognise that it is wrong. An AI system that generates a software architecture recommendation will produce suggestions that look exactly like good engineering advice and occasionally contain a critical flaw that an experienced engineer would catch in thirty seconds. An AI system that drafts marketing copy will produce content that reads professionally and sometimes makes claims that are factually incorrect or legally problematic.
The only reliable way to catch these errors is to have someone who knows the subject deeply enough to recognise them. Not someone who can read the output and say it looks reasonable. Someone with the professional expertise to evaluate whether it is actually correct.
That person is you. If you spent the last decade building expertise in a field — any field — while also developing the digital fluency that the modern workplace demanded, you are the precise professional profile that the AI evaluation market is currently under-supplied with and actively paying to find.
Microsoft is cutting the roles AI can now perform. It is simultaneously unable to fill the roles that require a human expert to evaluate what its AI produces. Both of these things are true at the same time. The question is which side of that equation you choose to stand on.
Microsoft employees offered buyouts (Oct 2026)
Microsoft AI capital expenditure, fiscal 2026
Share of Microsoft's workforce affected
What the Market Is Actually Paying
AI evaluation is not a side hustle. It is a structured, growing professional market with tiered rates based on domain expertise and the complexity of the evaluation work. As of October 2026, the platforms actively recruiting domain experts are paying:
1) Software engineers reviewing AI-generated code for correctness, security vulnerabilities, and architectural soundness: $50–$100/hr on Outlier AI, DataAnnotation, and AfterQuery
2) Product managers and strategists evaluating AI-generated business analysis, market research, and strategic recommendations: $45–$85/hr on Mercor and SME Careers
3) Marketing professionals reviewing AI-generated campaigns, copy, and brand strategy for accuracy, legal risk, and strategic coherence: $40–$75/hr on DataAnnotation and OpenTrain AI
4) Financial analysts verifying AI-generated financial models, earnings summaries, and investment analysis against primary source filings: $55–$100/hr on Outlier AI and AfterQuery
5) Lawyers and legal professionals validating AI citation accuracy, jurisdictional applicability, and regulatory compliance: $55–$110/hr on Mercor and SME Careers
6) Healthcare and clinical professionals evaluating AI-generated clinical notes, medication recommendations, and diagnostic outputs: $65–$120/hr on OpenTrain AI and SME Careers
These rates are not exceptional. They are standard market rates for qualified evaluators. Professionals working twenty hours per week at the mid-range of their domain can replace a significant portion of a full-time technology salary — without a long-term employment commitment, without office politics, and without waiting for a headcount approval that is no longer coming.
Meet the People Who Saw It First
Chisom is a product manager who spent nine years at a Series B fintech startup in Lagos. She owned the roadmap. She knew how user research got done, how competitive analysis was conducted, how pricing decisions were made. When her company's Series C round fell through and the team was halved, she found herself with a LinkedIn profile, a decade of expertise, and a market that seemed to have no interest in either.
A colleague mentioned AI evaluation. Within three weeks, Chisom had passed qualification assessments on two platforms and was reviewing AI-generated product strategy documents for a US-based software company. The work was familiar — she was doing what she had always done, which was read analysis and determine whether it was actually correct. The difference was that the analysis had been produced by a model rather than a junior analyst, and the platform was paying her $70 per hour to catch what the model had missed.
Or consider David, who spent twelve years in B2B sales at a global software company, most recently as a regional director. He understood CRM architecture, sales process design, and how AI-generated outreach looked to an experienced buyer — which is to say, often generic, sometimes off-message, and occasionally containing errors that would have killed a real deal. When his role was eliminated in a round of cuts framed as an "efficiency initiative," he moved into evaluating AI-generated sales content and coaching materials for one of the major evaluation platforms.
He works part-time. He earns more per hour than he did as a director. He is building a portfolio that documents his evaluation expertise — a new professional credential for a new professional market.
The professionals finding income in this market are not doing something foreign to their experience. They are doing the same thing they were always doing — applying professional judgment to professional outputs. The outputs just come from a model now, not a colleague.
The Gap Between Knowledge and Earning
There is one thing worth being honest about, because the professionals who are actually earning well in this market have figured it out and the ones who are struggling have not: having domain expertise is necessary but not sufficient.
An AI evaluation platform is not going to pay $80 per hour for someone who reads an output and says "this seems fine" or "this looks wrong to me." It is going to pay $80 per hour for someone who can articulate precisely why something is wrong, in what specific way it is wrong, and what a correct response would contain. That requires methodology, not just knowledge.
The difference between a professional who earns $40/hr on these platforms and one who earns $90/hr is usually not years of experience. It is the ability to translate professional judgment into structured, documented evaluation that the platform can use to train its model. That is a learnable skill. It is not intuitive for most professionals coming from traditional employment, where the expectation was that judgment was enough.
This is the gap that training closes. Not the knowledge gap — you already have the knowledge. The methodology gap. The ability to convert what you know into evaluation outputs that meet the quality standards that move you up the rate tiers.
Five Steps to Start This Week
First, identify your evaluation domain. The platforms segment by field, and the rates vary significantly. If you come from technology, software engineering and product evaluation pay well. If you have a professional background in finance, law, medicine, or marketing, those are the tracks with the most open demand. Do not try to evaluate across domains — depth pays. Breadth does not.
Second, build your evaluation methodology before you apply. A qualification assessment tests whether you can produce structured, consistent, documented evaluations — not whether you have expertise. Practice evaluating AI outputs in your domain before you sit an assessment. Write down your reasoning process explicitly. The platforms score on consistency and documentation, not just correctness.
Third, treat the first three months as your portfolio-building phase. The income in those months is meaningful, but the credential is more valuable. Documented evaluation work, with pass rates and quality scores from reputable platforms, is the professional track record that opens higher-rate projects. Start now even if the first rates are lower than you want.
Fourth, apply to multiple platforms in parallel. Mercor, Outlier AI, DataAnnotation, AfterQuery, SME Careers, and OpenTrain AI each have independent qualification processes and independent project pipelines. Running parallel applications reduces your time to first income. The onboarding timelines are two to six weeks per platform.
Fifth, do not treat this as a fallback. The professionals earning the most in AI evaluation are the ones who are treating it as a primary professional track, not a temporary bridge while they wait for a traditional role to appear. The traditional roles being eliminated at Microsoft this quarter are not coming back. This market is growing. Position accordingly.
The Longer View
Microsoft's voluntary buyout is being covered as a story about corporate restructuring, about AI investment, about the future of work. All of those framings are accurate. But for the professional reading this, the most important fact is a simpler one.
Microsoft is spending $175 billion this fiscal year to build AI systems that will produce outputs across every professional domain it operates in. Those outputs will need to be evaluated, verified, and quality-assured by professionals who understand the domains well enough to know when the system is wrong.
The company is simultaneously reducing the headcount it maintains to do the repetitive work that AI has learned to do, and it is unable to find enough domain experts to evaluate the work that AI cannot yet do reliably. Both of these things are happening inside the same company, in the same quarter, in the same memo.
You do not have to wait for the technology sector to decide what your expertise is worth. There is a market that has already decided. It is paying for what you know, right now, at rates that reflect how hard it is to find someone who genuinely knows it.
The skills are there. The window is open. The question is whether you walk through it.
Your expertise is already what this market needs.
Crested Academy's AI Evaluator certification trains professionals in the evaluation methodology that converts domain knowledge into platform income. CAE-certified evaluators consistently place in the top rate tiers.
→ Start with the free AI Evaluator Starter Guide at crestedacademy.com
Pingback:The Skills Are Still There. The Jobs Are Not. What Millennials Should Do Next. – crestedacademy