Fears of agentic AI layoffs are running ahead of the evidence, and history has a plain answer for them. The last time a machine took over the writing trade, the trade grew. Johannes Gutenberg’s press put the scribes of Europe out of work after 1450. Within fifty years the continent had more books, more readers and more people working around books than ever before.
The same pattern is forming around software. Agentic AI now writes a large share of the code in many shops, and programmers feel it first. But the numbers from the Federal Reserve, the Bureau of Labor Statistics, Gartner and Anthropic say something else. The agentic AI layoffs story is bigger than the layoffs. The press did not shrink the book trade. It built an industry on top of it.
The scribes lost their trade and the book trade exploded
Before the press, a trained scribe copied each book by hand, one copy at a time. Only a few people could do the work, so only a few books existed. The scribe was the bottleneck. The price of a book was the price of his time.
Then the press arrived, and the bottleneck was gone. In Florence in 1483, the Ripoli Press charged three florins for 1,025 copies of a text, in a price comparison compiled by AI Impacts. A scribe wanted about one florin for a single copy. The same analysis of Buringh and van Zanden’s book-production series finds that European book output doubled every 43 years once printing spread. Before the press, the doubling time was about 104 years.
Book prices in the Netherlands fell roughly five-fold between 1460 and 1550. Output per worker rose about twenty-fold in the first two centuries of print. The scribes did lose their work. Yet printers, typefounders, editors, booksellers and newspaper publishers took their place, and then a reading public arrived. Knowledge spread wider and faster than any one could have imagined.
So the entry-level coder is the scribe in this story. The question is whether it spreads to the whole trade, Or whether it stays at the edge, where the copying work was.
The broader evidence says exaggerated
The broader evidence on agentic AI layoffs points the other way. A Federal Reserve Bank of Chicago working paper found that jobs in the occupations most exposed to AI rose 4.05 percent from 2019 to 2024. Less exposed jobs trimmed payrolls in the same period, and wages rose across the board. The Yale Budget Lab’s tracker, updated on October 14, still finds no clear evidence of AI-related disruption in the labor market as a whole.
Anthropic’s March 2026 Economic Index draws on a million Claude conversations. It found computer and mathematical tasks up 14 percent on the API but down 18 percent on Claude.ai between August 2025 and February 2026. Coding is moving from chat into the pipeline, which is where agentic AI lives. The company’s study of agent autonomy found that about half of all agentic activity is software engineering. Still, 73 percent of tool calls had a human in the loop.
The press did not just copy faster, it made new work
The deeper lesson from print is that cheap copying created demand nobody had priced. Lower book prices drove literacy, and literacy drove demand for newspapers. Newspapers then needed editors, reporters, printers and advertisers. The copying job vanished. Everything around the copying job multiplied.
Software is following that curve. Uber’s engineering team describes a software factory where agents generate code and people govern the line. The lifecycle around the agents is changing faster than the headcount. That is why I keep writing about what agentic AI does to the SDLC and the ten assumptions it breaks.
What I see from inside the loop
I have watched this from inside regulated loan origination. Agents now draft most of the code there, and humans review intent rather than diffs. The copying work went to the agent. The judging work did not, and there is more of it than before. Someone has to write the intent, run the evals and read the audit trail. Someone has to set the permission boundaries and own the outcome when a regulator asks.
That is where the jobs are moving. I wrote about this shift in the IT services industry and in code review built on intent, not diffs. The teams that cut coders and kept the org chart will be the ones Gartner is talking about in 2029. Those that retrain coders into harness engineers, eval engineers and governance owners will ship more software than ever. The cost of a line of code just did what the cost of a page did in 1483.
My opinion: agentic AI layoffs are a transition cost, not the destination. Instructions in, results out was IT. Intent in, outcomes out is agentic AI, and intent still needs a human who means it. The scribes lost. The book won. So did everyone who learned to work the press.
Did your firm cut coders this year, or did it start hiring for the loop around them?
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.
My Books: Three books on agentic AI for regulated enterprises: Intent In, Outcomes Out, Earned Autonomy and The Governed Loop. Reading order, formats and buy links are on one page.
© Dr. Harish Kotadia, Ph.D., All Rights Reserved, 2026
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.