45K tech workers were laid off in March 2026. Amazon's AI agent then took down their own site. Here's the real lesson for small businesses.
Article text
Amazon cut 16,000 corporate employees in January. CEO Andy Jassy said the company would "need fewer people doing some of the jobs that are being done today." The money would go to AI agents instead.
Two months later, one of those AI agents took down Amazon's retail website for six hours. Customers couldn't check out, couldn't see their account info, couldn't even view product prices. The cause, according to internal documents: an AI agent followed advice from an outdated internal wiki and made changes to critical production systems that a human would have questioned.
That's not an AI success story. That's a $200 billion lesson in what happens when you skip steps.
The Headlines Don't Tell the Whole Story
If you run a business and you've been watching the news this month, the numbers are alarming. According to tracking data from RationalFX and Layoffs.fyi, 45,363 tech workers lost their jobs since January 2026 . Of those, 9,238 were directly attributed to AI and automation. That's about 20%.
The companies making the cuts aren't small. Block slashed its workforce from 10,000 to 6,000. Pinterest cut 15% of its team to pursue an "AI-forward strategy." eBay automated listings, pricing, and customer service, then let 800 people go. HP is targeting $1 billion in savings by 2028 through a combination of AI and 4,000-6,000 fewer employees.
If you're an SMB owner reading those headlines, you might be thinking one of two things: "Should I be doing this too?" or "Am I going to lose my business to someone who does?"
Both reactions are understandable. Both miss the point.
If your business has 20, 50, or 100 employees and you're wondering how to think about AI and your workforce, we do free 30-minute calls where we walk through exactly that. No sales pitch. Just clarity on what makes sense for your size.
Why the Enterprise Playbook Doesn't Work at Your Scale
Amazon can cut 16,000 people because they have 1.5 million. They can absorb the chaos. They can afford to put AI in charge, watch it crash their site, and then add humans back into the loop. That's what they did, by the way. After the outages, Amazon introduced what they called "controlled friction" into AI-assisted deployments. Translation: they put humans back in the decision chain because the AI alone wasn't ready.
When you have 35 employees, you don't have that margin. Every person on your team wears multiple hats. Your office manager also handles vendor relationships. Your lead tech also trains new hires. You can't cut roles. The institutional knowledge lives in people's heads, not in documented processes.
This is the gap most AI coverage ignores. The headlines say "AI is replacing workers." The reality at companies your size is different. AI works best when it makes your existing team faster, not when it makes your team smaller.
What Amazon Actually Got Wrong
The Amazon crash is worth studying not because Amazon is bad at technology, but because they're very good at it and still made this mistake.
Here's what went wrong, reconstructed from Fortune's reporting and internal documents:
They had outdated documentation.
An internal wiki hadn't been updated. The AI agent treated that wiki as a source of truth.
The AI agent acted on bad information.
It followed instructions from the outdated wiki and made changes to production systems.
No human checked the work.
The AI had enough autonomy to make critical changes without a human reviewing them first.
The result was catastrophic.
Four high-severity incidents in one week, including a six-hour outage of core retail functions.
After this happened, Amazon tried to remove the reference to "GenAI-assisted changes" from their postmortem before a senior leadership meeting. They eventually acknowledged the issue and added human oversight back into the process.
The lesson isn't "AI is dangerous." The lesson is: AI is only as good as the process it's built on. Outdated docs, unreviewed autonomy, no guardrails. Those are process failures, not technology failures.
The SMB Advantage Nobody Talks About
Here's something counterintuitive: being small actually gives you an advantage here.
At Amazon, that outdated wiki probably existed for years without anyone noticing. In a company with 1.5 million employees and thousands of internal documentation pages, stale information is inevitable. Nobody owns everything.
In a 40-person company, you probably know where the problems are. Your ops manager can tell you which spreadsheet has wrong formulas. Your bookkeeper knows which reports take too long. Your dispatcher knows which part of the scheduling process causes the most callbacks.
That knowledge is gold. It means you can skip the part where big companies spend months doing "AI readiness assessments" and instead go straight to the specific friction point that costs you the most time or money.
We see this every week. A property management company knows their tenant communication process is broken. A plumbing contractor knows dispatching takes too long. An accounting firm knows month-end close involves 14 manual steps that could be 4. They don't need a consultant to tell them the problem exists. They need help building the right fix.
Four Questions Before You Let AI Touch Any Job
If you're thinking about where AI fits in your business, here's the framework we use with every client. It takes about 15 minutes and will save you from Amazon's mistake.
1. Is this task documented?
If the process only exists in someone's head, AI can't do it. Full stop. Before any automation, the steps need to be written down. This is the boring part. It's also the part Amazon skipped. Their documentation was stale, and their AI built on a rotten foundation.
Start here: pick one process and have the person who does it write down every step, every decision point, every "it depends" moment. That document is worth more than any AI tool.
2. What happens when AI gets it wrong?
For every task you're considering automating, ask: if the AI makes a mistake, what's the blast radius?
AI writes a draft email that sounds off? Low risk. A human reviews it before it sends.
AI changes a field in your CRM? Medium risk. Wrong data propagates to invoicing and reporting.
AI modifies your production database? High risk. That's what crashed Amazon.
Start with low-blast-radius tasks. Get comfortable. Then move to medium. Leave high-risk tasks for when you have strong guardrails and a human checkpoint.
3. Does this task require judgment or just execution?
Some tasks are pure execution: data entry, report generation, appointment reminders, invoice formatting. AI handles these well today.
Other tasks require judgment: deciding whether to extend credit to a customer, choosing which project to prioritize, handling an upset client. AI can provide information to support these decisions, but the decision itself should stay with a human.
The companies that get in trouble are the ones that hand AI a judgment task and treat it like execution. Amazon's AI was executing changes, but the situation required judgment about whether the source information was reliable. Nobody built that check into the process.
4. Will this free time or cut headcount?
This is the question that matters most for businesses your size. The right answer, almost always, is "free time."
When your bookkeeper spends 6 hours a week on manual reconciliation and AI cuts that to 45 minutes, you didn't eliminate a bookkeeper. You gave your bookkeeper 5 hours back. Those 5 hours go to analysis, client communication, process improvement, or the three other tasks they never had time for.
Block cut 4,000 people. Jack Dorsey said "a significantly smaller team, using the tools we're building, can do more and do it better." He's running a 6,000-person company. That math works differently when you have 30 people and everyone is essential.
What Smart SMBs Are Actually Doing
The businesses we work with aren't cutting teams. They're doing something more interesting: they're making their teams capable of things that used to require hiring.
A 28-person electrical contractor was spending 12 hours a week on estimate follow-ups. Calls that went to voicemail, emails that needed to be sent three times, notes that had to be manually entered into their CRM. We automated the follow-up sequence and the CRM entry. The estimator kept their job. They just stopped losing deals to slow follow-up.
A 45-employee staffing agency was manually matching candidates to job requirements by reading through resumes one at a time. We built a matching system that surfaces the top 5 candidates for each role in seconds. The recruiters still make the final call. They just make it faster, with better information.
Neither of these companies eliminated a role. Both of them grew revenue without growing headcount.
The Real Risk Isn't AI. It's Doing Nothing.
If the current trend continues, RationalFX projects approximately 264,730 tech layoffs by the end of 2026. The big companies are restructuring around AI whether the technology is ready or not.
That creates an opportunity for smaller businesses. While enterprise companies stumble through messy, high-profile AI implementations, an SMB that gets it right, that automates the right things with the right guardrails, quietly becomes more competitive. Lower costs per customer. Faster response times. Fewer errors. Same team.
The risk isn't that AI will take your employees' jobs. The risk is that your competitor figures this out first.
Where to Start
You don't need a massive AI strategy. You need one process, documented clearly, with a low blast radius and high time cost. Start there.
If you're not sure which process to pick, or you want someone to look at your operations with fresh eyes, book a free workflow call with us . We'll walk through your day-to-day and tell you honestly where AI makes sense and where it doesn't. Sometimes the answer is "not yet." That's fine too.
The businesses that win with AI aren't the ones that move fastest. They're the ones that move deliberately.