AI Layoffs: Y2K and the Dot-Com Bust Again

Agentic AI layoffs reached a new scale on Monday. The largest software company offered voluntary exit packages to about 60,000 employees, roughly a quarter of its workforce, while pouring money into artificial intelligence. The offer follows a year of cuts across the industry, from the largest IT services companies to the largest consulting firm. Developers are the headline casualties, but the list still runs through testers, project managers and support staff.

New here? I publish one agentic AI governance post every weekday. Subscribe to the blog and it lands in your inbox the moment it goes live.

I have seen this movie twice. In 2000 the Y2K army went home. Then in 2001 the dot-com bubble took 560,000 tech jobs with it. Both times the bust came first, and both times the boom that followed was bigger than the one that broke. This post follows my $234 billion forecast piece, because agentic AI layoffs are the other half of that story.

The Y2K army went home in January 2000

The United States spent about $100 billion fixing two-digit dates between 1995 and 2001, according to a Commerce Department estimate. Then the clock rolled over and nothing broke. So the work ended on a single night, and so did the need for the people who did it.

A Connecticut life insurer took its offshore Y2K programmers and formed a new unit around them, the same article reports. Instead of going home, the unit built call center and web applications for outside clients. The unit recouped $25 million in Y2K costs and expected $25 million more in sales that year. Everyone on the project kept a job, because the skill was never the date fix. It was knowing the systems.

The dot-com bust took 560,000 jobs

The second wave hit a year later, and it was bigger. The American Electronics Association counted about 560,000 high-tech jobs lost from the peak of the dot-com era through the end of 2002. Manufacturing lost 415,000 of them, while telecom lost 135,000. A later university study put the information technology sector’s loss at 403,300 jobs between March 2001 and April 2004, a cut of nearly 19 percent. San Francisco lost half its tech employment, for example, and Dallas lost 30 percent.

The firms that came out of it were the ones that had already built the next practice. The largest computer company of the day had bet on the web before the crash. By 1999 it was providing e-business services to more than 10,000 clients, its own history page says. Meanwhile a half-billion-dollar campaign told enterprises to do real business on the web. Revenue grew nearly 40 percent from 1994 to 2000. The IT services companies made the same bet from the other side. The offshore ERP and CRM practices they built in those years became the industry I spent two decades inside.

This time the cuts say AI out loud

The 2025 and 2026 rounds differ in one way, because the companies now name the cause. The largest IT services company cut about 12,261 middle and senior roles in 2025 while saying it was deploying AI at scale. One analyst told Reuters that 400,000 to 500,000 IT services professionals are at risk over two to three years. The largest consulting firm’s headcount fell by about 7,000 in a single quarter.

A frontier lab’s own data shows why developers sit at the front of the agentic AI layoffs. Coding is still the single largest use of its models. Its March 2026 Economic Index report found that coding work is moving from the chat window to the API, where the workflows are directive and the human is mostly absent. That shift, the report says, “may signal more imminent transformation of work for the associated jobs.” In short, the agent now does the task, not the suggestion.

The practice that follows the bust

Every bust in this industry has ended the same way so far. The firms that survived built a practice around the thing that caused the bust. After Y2K it was e-business. After the dot-com crash it was offshore ERP, CRM and web enablement for the Fortune 1000. That is how the IT services companies grew from date fixers into a $283 billion industry that employs 5.7 million people. The pattern is not subtle. The bust clears the old work, and the practice sells the new work to the same clients.

So the question for the IT services firms is not how many people to cut. It is whether they are building the agentic AI practice now, the way the computer giant built e-business, or waiting for the agentic AI layoffs to finish first. The enterprise does not need another model. It needs the harness around the model and the controls that let an agent act inside a regulated process. It needs a partner that can build both. That was the e-business job, too, because nobody hired that company to explain the web.

What I do with this

In my work in regulated loan origination the same people who ran the legacy integrations staffed the first agent projects. They knew where the loan file lived and who signed it. That knowledge did not become worthless when the agent arrived. Instead it became the specification the agent runs against, written as intent, enforced by a hook, and logged as evidence. The Y2K programmer knew the systems, while the agentic AI architect knows the controls.

My flat opinion: the agentic AI layoffs of 2026 are the Y2K layoffs of 2000 with the cause stated out loud. The firms that treat them as a cost exercise will miss the practice that follows. Build the agentic AI business now with the people who know the clients’ systems. Sell it to the Fortune 1000 the way the e-business divisions did. Instructions in, results out was IT. Intent in, outcomes out is agentic AI, and somebody has to build the loop for every enterprise that cannot.

Which practice did your firm build after the last bust, and who is building the agentic AI one now?

Was this post useful? One click tells me what to write next.

← Back

Thank you for your response. ✨

Subscribe to the Blog: I publish a new post every weekday on one agentic AI governance question. Enter your email and each post arrives in your inbox the moment it goes live.

For more on this topic, please check out my three books on agentic AI for regulated enterprises:

Intent In, Outcomes Out: the thesis, why intent in, outcomes out replaces instructions in, results out.
Earned Autonomy: the method, the EAI-1000 framework that scores how much autonomy an agent has earned.
The Governed Loop: the architecture, the six layers that decide outcomes, govern the harness, not the model.
All three books: reading order, formats and buy links on one page.

Dr. Harish Kotadia, Ph.D., is an Enterprise AI Architect with 20+ years of IT consulting experience serving Fortune 100 clients, specializing in agentic AI systems built on Anthropic Claude, AWS Bedrock, and Google Vertex AI.

Disclaimer: This blog post is based on publicly available academic publications, vendor documentation, open standards, and news items from reputed media sources linked above. This post is intended for educational purposes, to help the enterprise agentic AI community build a shared vocabulary from public, authoritative sources.

Views and opinions expressed here are my own and do not represent those of any employer or client, past or present. The analysis presented is my independent interpretation of the published sources linked above and does not constitute legal, financial, or consulting advice of any kind.