# I Lost My Job To AI. 90 Days Later I Am Earning 40% More Because of It. Here is How.
**作者**: NeilXbt
**日期**: 2026-04-29T13:00:10.000Z
**来源**: [https://x.com/neil_xbt/status/2049473820033990842](https://x.com/neil_xbt/status/2049473820033990842)
---
Fourteen months ago, I lost my job.
Not because I was underperforming. Not because the company was struggling. Because someone in a boardroom ran the numbers and realized that AI could do what I did for a fraction of the cost. I was out.
I spent three weeks in the spiral. The anger, the confusion, the low-grade panic about what comes next. I sat with it longer than I should have. Then I stopped doing that and started doing something else.
I spent 90 days earning three Microsoft AI certifications.
The job I got on the other side paid 40% more than the one I lost.
The role did not exist two years earlier. And it exists specifically because organizations are deploying AI at scale and cannot find enough people who know how to do it properly.
This is not a story about resilience.
It is a story about a specific decision made in a specific window of time that produced a specific result. And the reason I am writing it is that the same window is open right now, for you, and it will not stay open indefinitely.
## The Thing Most People Get Wrong About AI Careers
The assumption almost everyone makes about AI career opportunities is that the demand is for people who can use AI tools. Prompting. API integration. Workflow automation.
That is a real market. It is also a crowded one.
The market that is genuinely undersupplied, the one where offers are still moving at a pace that most people in tech stopped seeing two years ago, is not for AI users. It is for AI architects. The people who can design the systems, make the infrastructure decisions, and be accountable for how AI gets deployed across an entire organization.
These are not the same skill set and they are not compensated at the same level.
A person who can use Claude or ChatGPT to produce content is useful. A person who can design and deploy a scalable, secure, compliant AI infrastructure for 50,000 employees, manage the cost model, and explain every architectural decision to a board of directors is irreplaceable.
Microsoft AI certifications are built for the second person. And the second person, right now, is rare enough that the companies who need them will pay almost anything to find them.
## Why Microsoft Specifically
There are hundreds of AI courses available. Courses on prompting, on building workflows, on using specific tools. Most of them teach you how to operate inside a system that someone else built.
Microsoft certifications teach you how to build the system.
More importantly, they teach you how to think about AI at an architectural level. How to make decisions about identity and access management. How to design data storage that is both performant and compliant. How to build AI solutions that scale without collapsing under their own weight. How to explain technical decisions in business terms that a boardroom can evaluate.
That combination is what organizations are actually paying for. And a Microsoft badge is not a course completion certificate or a LinkedIn skill endorsement. It is a verified credential, with a proctored exam, that tells a hiring manager exactly what you know and exactly what you can do.
In a market where everyone claims AI experience, that verifiability is worth an enormous amount. I learned this the hard way by trying to compete without it first.
## The Three Levels and What Each One Opens
The certification path has a clear architecture. Three levels, each one building on the last, together forming a credential stack that most professionals in tech do not have and are not currently building.
Level 1: Azure AI Fundamentals (AI-900)
This is where the path starts and it is accessible to anyone regardless of technical background. It is where I started, and I will be honest: I almost skipped it because it felt too basic. That would have been a mistake.
AI-900 teaches the vocabulary, the mental models, and the foundational understanding of how AI and machine learning work within the Azure ecosystem. What machine learning actually is at a conceptual level. The difference between supervised and unsupervised learning. How computer vision, natural language processing, and conversational AI are structured. What Azure's AI services are designed to do and how they relate to each other.
This is not a deep technical certification. It is a credibility certification. And in the current hiring environment, that credibility is the first filter. A verifiable Microsoft badge tells a hiring manager that you have invested real time in understanding the space. That signal, in a market flooded with people who claim AI experience without evidence, is worth more than most people realize.
Who it is for: anyone who wants to work in and around AI systems without being the person who builds them from scratch. Marketers, business analysts, project managers, salespeople, operations professionals, anyone who wants AI fluency without needing to write production code.
Study time: 10 to 20 hours. The Microsoft Learn free path covers everything on the exam. The official practice assessment is free and the real exam draws heavily from the same question patterns.
Level 2: Azure AI Engineer Associate (AI-102)
This is where the compensation starts to shift meaningfully. This is also the certification that made me feel, for the first time since losing my job, like I was building something that could not be taken from me easily.
AI-102 is a professional level certification. It requires real technical ability. The exam covers building custom vision models, implementing natural language understanding, designing conversational AI systems with Azure Bot Service, integrating knowledge mining solutions, and managing responsible AI principles in deployed systems.
This is not theoretical. The exam includes scenario-based questions that require actual architectural decisions, and you cannot pass by reading alone. Hands-on time in the Azure portal, with the SDK and the services themselves, is not optional. Reading about Azure Cognitive Services and configuring one are two different types of knowledge, and the exam tests the second type.
The people who hold AI-102 are the ones organizations call when they need to build something. Not describe it, not theorize about it, but actually build a working AI system that integrates with existing infrastructure and can be deployed and maintained at scale.
Who it is for: developers, cloud engineers, data scientists, or anyone with a technical background who wants to specialize in AI implementation. Also worth pursuing for anyone coming from AI-900 who wants to cross into technical roles.
Study time: 40 to 80 hours with hands-on Azure practice as the non-negotiable component.
Salary range in the US for roles this certification opens: $110,000 to $160,000. Freelance consulting at $100 to $200 per hour.
Level 3: Azure Solutions Architect Expert (AZ-305)
This is the certification that changed everything for me. Not because it was the hardest, though it was, but because of what it signals.
AZ-305 is not AI-specific. It is enterprise architecture. It is the certification that says you are the person in the room who designs the entire system, not just the AI component. The infrastructure it runs on. The security model. The networking architecture. The governance framework. The business continuity planning. All of it, designed from the ground up, with accountability for every decision.
When a Fortune 500 company decides to deploy an AI solution across its entire workforce, they need someone who can design that deployment comprehensively and defend every choice it involves. AZ-305 is the credential that identifies that person.
The exam covers identity and access management, data storage architectures, infrastructure design, business continuity, and migration strategy. The scenario-based questions require real enterprise experience to answer correctly. Study time is 80 to 120 hours, and the most effective approach combines John Savill's AZ-305 YouTube series which is free and genuinely outstanding, the official Microsoft Learn path, practice exams on MeasureUp or Whizlabs, and study groups for the architectural reasoning the scenario questions demand.
Who it is for: senior engineers, technical leads, cloud architects, and anyone who wants to move into principal or staff-level roles where they are accountable for system design rather than just implementation.
What it opens: Solutions Architect, Principal Engineer, Cloud Strategy Lead, Enterprise AI Architect. This is the certification that gets you into rooms and conversations that most engineers never reach. It got me into mine.
## The 90-Day Schedule That Made This Real for Me
Three certifications in 90 days is a demanding target. I am not going to tell you it was easy because it was not. I am going to tell you it was worth it and that the structure made it possible.
Days 1 through 20 were for AI-900. Ten to fifteen hours of focused study through the Microsoft Learn free path, followed by the official practice assessment. I scheduled the exam for day 19. It is available online from home, no test center required.
Days 21 through 60 were for AI-102. This was the heavy lift, 40 to 60 hours of actual hands-on work in the Azure portal. I made one rule for myself and kept it: every service I configured once, I understood better than ten hours of reading. I built things. I broke them. I configured them again. The exam tests whether you can make real decisions, and you cannot fake that with memorized theory.
Days 61 through 90 were for AZ-305. Eighty hours compressed into 30 days is aggressive. I prioritized the scenario-based content from day one and studied in a small group when I could, because talking through architectural reasoning with other people made the hardest questions click in a way solo study did not.
Three certifications. Ninety days. Free study materials for every level. A credential stack that most professionals in technology right now do not have and are not building.
## Why the Timing Matters More Than Anything Else
In 2019, a cloud certification proved you understood cloud infrastructure. It was valuable, but the landscape was established and the credential was becoming normalized.
In 2026, a Microsoft AI certification proves something different. It proves you can be trusted with decisions that have real business consequences in a technology that most organizations are still genuinely figuring out. The boardrooms are full of executives who know they need AI but do not yet have the people who can design and deploy it responsibly at scale.
That gap between organizational need and available talent is what produces the compensation numbers. It is not permanent. Markets always close gaps eventually. The people who got cloud certified in 2019 built careers that the people who got certified in 2023 had to compete much harder to achieve.
I went from being let go because AI replaced my role to being hired specifically because I understood how to deploy AI at scale.
That does not happen if I wait another year. The role I landed exists because the gap is open right now. A year from now, more people will have made this decision and the gap will be narrower.
I am not telling you this to be inspirational. I am telling you this because I sat in a job that felt secure and it was not, and I now sit in a role that would not have existed two years ago, earning 40% more than I was before I lost everything that pushed me to start.
## One Last Thing
I did not have a technical advantage when I started this.
I was not a developer or a cloud engineer. I was someone who got displaced, recognized a specific opportunity, and executed on it with the kind of focus that most people save for things they feel they have no choice but to do.
Having no choice was the clearest I have ever felt.
You do not have to wait for that clarity. You do not have to lose something before you decide to build something.
The study materials are free.
The path is documented.
The market is real.
The window is open right now in a way it will not be forever.
Bookmark this. Start today.
Free study resources: Microsoft Learn AI-900 path, AI-102 path, and AZ-305 path at learn.microsoft.com.
John Savill's AZ-305 YouTube series for architecture preparation. Official practice assessments at learn.microsoft.com for all three exams.
## 相关链接
- [NeilXbt](https://x.com/neil_xbt)
- [@neil_xbt](https://x.com/neil_xbt)
- [17K](https://x.com/neil_xbt/status/2049473820033990842/analytics)
- [Upgrade to Premium](https://x.com/i/premium_sign_up)
- [9:00 PM · Apr 29, 2026](https://x.com/neil_xbt/status/2049473820033990842)
- [17.8K Views](https://x.com/neil_xbt/status/2049473820033990842/analytics)
- [View quotes](https://x.com/neil_xbt/status/2049473820033990842/quotes)
---
*导出时间: 2026/4/30 09:24:34*
---
## 中文翻译
# 我因 AI 失业。90 天后，我因为它薪水涨了 40%。这是我的做法。
**作者**: NeilXbt
**日期**: 2026-04-29T13:00:10.000Z
**来源**: [https://x.com/neil_xbt/status/2049473820033990842](https://x.com/neil_xbt/status/2049473820033990842)
---
十四个月前，我失业了。
不是因为表现不佳。不是因为公司陷入困境。而是因为董事会的某个人算了笔账，意识到 AI 可以以极低的成本做我做的事。我被淘汰了。
我在崩溃中挣扎了三周。愤怒、困惑，以及对未来的低度恐慌。我在这种状态中沉溺了太久，久过了头。后来，我停止了沉沦，开始做别的事。
我花了 90 天时间考取了三个微软 AI 认证。
我随后得到的工作薪水比我失去的那份高出了 40%。
两年前这个职位还不存在。它之所以存在，是因为组织正在大规模部署 AI，却找不到足够多懂如何正确操作的人才。
这不是一个关于坚韧的故事。
这是一个关于在特定的时间窗口做出特定决定，从而产生特定结果的故事。我写这个故事的原因是，同样的窗口现在为你敞开，但它不会永远敞开。
## 大多数人对 AI 职业的误解
几乎每个人对 AI 职业机会的假设是，市场需求在于会使用 AI 工具的人。比如提示词工程、API 集成、工作流自动化。
这是一个真实的市场，但也非常拥挤。
真正供不应求的市场，那个录用通知发放速度还保持着科技圈两年前节奏的市场，需求的是 AI 架构师，而不是 AI 用户。是那些能够设计系统、做出基础设施决策，并并对 AI 在整个组织中的部署负责的人。
这不是同一套技能，薪酬水平也不同。
一个会使用 Claude 或 ChatGPT 生成内容的人是有用的。但一个能为 5 万名员工设计并部署可扩展、安全、合规的 AI 基础设施，管理成本模型，并能向董事会解释每一个架构决策的人，是不可替代的。
微软 AI 认证正是为第二类人打造的。而现在，第二类人非常稀缺，需要他们的公司愿意支付任何代价来寻找他们。
## 为什么偏偏是微软
市面上有数百门 AI 课程。关于提示词、构建工作流、使用特定工具的课程。它们中的大多数教的是如何在别人构建的系统内进行操作。
微软认证教你的是如何构建系统。
更重要的是，它们教你如何从架构层面思考 AI。如何做出关于身份和访问管理的决策。如何设计既高效又合规的数据存储。如何构建不会因自身重量而崩溃的可扩展 AI 解决方案。如何用董事会能评估的商业术语来解释技术决策。
这种组合才是组织真正愿意付费的地方。而且，微软徽章不是课程完成证书或领英上的技能认可。它是一项经过监考验证的凭据，清楚地告诉招聘经理你懂什么、你能做什么。
在一个每个人都声称有 AI 经验的市场上，这种可验证性价值巨大。我是通过在没有它的情况下尝试竞争才艰难地明白这一点的。
## 三个级别及其开启的机会
认证路径有着清晰的架构。三个级别，每一级都建立在前一级之上，共同构成了大多数技术人员目前不具备且正在构建的凭据栈。
**第一级：Azure AI 基础知识 (AI-900)**
这是路径的起点，无论技术背景如何，任何人都可以接触。我是从这里开始的，坦白说：我差点跳过它，因为它感觉太基础了。如果那样做，我就错了。
AI-900 教授的是词汇、思维模型，以及对 AI 和机器学习如何在 Azure 生态系统中运作的基础理解。机器学习在概念上究竟是什么。监督学习和无监督学习的区别。计算机视觉、自然语言处理和对话式 AI 是如何构建的。Azure 的 AI 服务设计用于做什么，以及它们如何相互关联。
这不是一个深度的技术认证。它是一个可信度认证。在当前的招聘环境中，这种可信度是第一道筛选。一个可验证的微软徽章告诉招聘经理，你投入了真实的时间来理解这个领域。在一个充斥着毫无证据却声称有 AI 经验的人的市场中，这个信号的价值比大多数人意识到的要高。
**适合人群**：任何想在 AI 系统周围工作但不想从零构建它们的人。营销人员、商业分析师、项目经理、销售人员、运营专业人士，任何想要 AI 流利度而不需要编写生产代码的人。
**学习时间**：10 到 20 小时。Microsoft Learn 的免费路径涵盖了考试中的所有内容。官方的练习评估是免费的，真正的考试很大程度上从相同的题模式中抽取题目。
**第二级：Azure AI 工程师助理 (AI-102)**
这是薪酬开始发生有意义转变的阶段。这也是让我自失业以来第一次感觉自己在构建某种无法轻易被剥夺的东西的认证。
AI-102 是一个专业级别的认证。它需要真正的技术能力。考试涵盖构建自定义视觉模型、实现自然语言理解、使用 Azure Bot Service 设计对话式 AI 系统、集成知识挖掘解决方案，以及在部署的系统中管理负责任 AI 的原则。
这不是纸上谈兵。考试包含需要实际架构决策的场景题，仅靠阅读是无法通过的。在 Azure 门户、SDK 以及服务本身上花费的时间是强制性的。阅读 Azure Cognitive Services 和配置一个服务是两种不同类型的知识，而考试测试的是第二种类型。
持有 AI-102 的人是组织在需要构建东西时呼叫的人。不是描述它，不是空谈它，而是构建一个与现有基础设施集成并能大规模部署和维护的真正工作的 AI 系统。
**适合人群**：开发人员、云工程师、数据科学家，或任何有技术背景并希望专攻 AI 实施的人。对于来自 AI-900 并希望跨入技术角色的人来说，也值得追求。
**学习时间**：40 到 80 小时，其中 Azure 实践是不可协商的组成部分。
**该认证开启的职位在美国的薪水范围**：110,000 美元到 160,000 美元。自由职业咨询费为每小时 100 到 200 美元。
**第三级：Azure 解决方案架构师专家 (AZ-305)**
这个认证改变了我的一切。不仅因为它是最难的——虽然确实如此——更因为它的信号作用。
AZ-305 并非专门针对 AI。它是企业架构。这个认证表明你是房间里设计整个系统的人，而不仅仅是 AI 组件。它运行的底层基础设施。安全模型。网络架构。治理框架。业务连续性规划。所有这一切都从头设计，并对每一个决策负责。
当一家财富 500 强公司决定在其整个员工队伍中部署 AI 解决方案时，他们需要能够全面设计该部署并为其涉及的每一个选择进行辩护的人。AZ-305 就是识别此类人员的凭据。
考试涵盖身份和访问管理、数据存储架构、基础设施设计、业务连续性和迁移策略。基于场景的问题需要真正的企业经验才能正确回答。学习时间为 80 到 120 小时，最有效的方法结合了 John Savill 的 AZ-305 YouTube 系列（免费且真正出色）、官方 Microsoft Learn 路径、MeasureUp 或 Whizlabs 上的练习考试，以及针对场景题所需的架构推理进行的小组学习。
**适合人群**：高级工程师、技术负责人、云架构师，以及任何希望进入主要或人员级别的角色，对系统设计而不仅仅是实施负责的人。
**它开启了什么**：解决方案架构师、首席工程师、云战略负责人、企业 AI 架构师。这是让你进入大多数工程师从未触及的房间和对话的认证。它让我进入了属于我的那个房间。
## 让我实现这一目标的 90 天计划
90 天内获得三个认证是一个苛刻的目标。我不会告诉你这很容易，因为这并不容易。我要告诉你的是，这是值得的，而且这种结构使其成为可能。
第 1 天到第 20 天用于 AI-900。通过 Microsoft Learn 免费路径进行了 10 到 15 小时的专注学习，随后进行了官方练习评估。我将考试安排在第 19 天。考试可以在线在家参加，无需去考试中心。
第 21 天到第 60 天用于 AI-102。这是繁重的工作，在 Azure 门户中进行了 40 到 60 小时的实际动手操作。我为自己制定了一条规则并坚持了下来：我配置过的每一个服务，我的理解都比我阅读十小时要深刻。我构建了东西。我弄坏了它们。我又重新配置。考试测试的是你能否做出真正的决策，你无法靠死记硬背理论来造假。
第 61 天到第 90 天用于 AZ-305。将 80 小时压缩到 30 天内是很激进的。我从第一天起就优先考虑基于场景的内容，并尽可能进行小组学习，因为与其他人一起讨论架构推理，能让最难的问题在独自学习无法企及的层面上变得豁然开朗。
三个认证。九十天。每个级别都有免费的学习材料。大多数技术人员目前不具备且正在构建的凭据栈。
## 为什么时机比什么都重要
2019 年，云认证证明你理解云基础设施。它很有价值，但格局已定，该凭据正在变得常态化。
2026 年，微软 AI 认证证明的东西不同。它证明你可以在大多数组织仍在真正摸索的技术中，被托付具有真实商业后果的决策。董事会里挤满了知道自己需要 AI 但尚未拥有能够负责任地大规模设计和部署 AI 的人员的高管。
组织需求与可用人才之间的差距正是产生这些薪酬数字的原因。这不是永久的。市场最终会弥合差距。2019 年获得云认证的人建立职业生涯，而 2023 年获得认证的人则必须努力竞争才能实现。
我从因为 AI 取代了我的角色而被解雇，变成了因为理解如何大规模部署 AI 而被专门聘用。
如果我再等一年，这一切就不会发生。我得到的职位之所以存在，是因为差距现在正好敞开。一年后，更多人会做出这个决定，差距将缩小。
我告诉你这些不是为了以此励志。我告诉你这些是因为我曾坐在一份感觉安全但实际不然的工作中，而现在我坐在一个两年前还不存在的职位上，薪水比失去一切推动我重新开始之前高出 40%。
## 最后一件事
开始时我并没有技术优势。
我不是开发人员或云工程师。我是一个被取代的人，认识到了一个特定的机会，并以大多数人留给他们觉得别无选择只能做的事情时的专注力去执行它。
别无选择是我曾有过的最清晰的感觉。
你不必等待那种清晰。你不必在决定构建东西之前先失去什么。
学习资料是免费的。
路径是明确的。
市场是真实的。
窗口现在敞开，但不会永远敞开。
收藏这篇文章。今天就开始。
免费学习资源：Microsoft Learn 的 AI-900 路径、AI-102 路径和 AZ-305 路径，网址为 learn.microsoft.com。
John Savill 的 AZ-305 YouTube 系列用于架构准备。所有三门考试的官方练习评估位于 learn.microsoft.com。
## 相关链接
- [NeilXbt](https://x.com/neil_xbt)
- [@neil_xbt](https://x.com/neil_xbt)
- [17K](https://x.com/neil_xbt/status/2049473820033990842/analytics)
- [升级到 Premium](https://x.com/i/premium_sign_up)
- [9:00 PM · Apr 29, 2026](https://x.com/neil_xbt/status/2049473820033990842)
- [17.8K 次观看](https://x.com/neil_xbt/status/2049473820033990842/analytics)
- [查看引用](https://x.com/neil_xbt/status/2049473820033990842/quotes)
---
*导出时间: 2026/4/30 09:24:34*