Most technology innovations are incremental improvements—the next faster chip, the next bigger hard drive, a new feature for your CRM, a better email client, etc. Every now and then, something comes along that opens up huge new white spaces of innovation by enabling us to solve previously intractable problems. Problems people had been working on for a long time, but the tech just wasn’t good at solving them.
Desktop computers combined many previously stand-alone devices into one and added new capabilities never before available (previously intractable with dedicated devices) through software applications, resulting in a surge in productivity among desk workers. Mobile came along, making those apps portable in your pocket at all times and adding tons more that were previously unavailable on the desktop (previously intractable, such as location-aware apps).
Now comes AI. AI certainly has enabled incremental improvements in existing apps (CRM with AI! Zoom with AI!). While there is a lot of investment in AI incrementalism, I am more interested in the previously intractable problems that are now tractable with AI. The blue oceans that will be built. Whole new categories that were not possible with prior architectures and tools. Some categories are starting to emerge, and I am investing in them. I am sure there will be more. Here is what I am seeing today in the middle of 2025.
The primary thing AI infrastructure is better at than humans is processing massive amounts of unstructured data (across data types, text, video, audio, image, etc.), extracting insights, and making it structured. Their brains are simply larger than ours and have a much wider context window. Just as the GPS chip and wireless radios in mobile phones enabled whole new app categories, this unstructured -> structured intelligence layer has enabled massive new app categories, including:
Worker augments
Every tech trend threatens to “throw a bunch of people out of work”. What history shows us is that previously inefficient industries become much more efficient, and more new areas of opportunity open up than close. For example, the mechanization of agriculture:
A century of innovation has completely rewritten America’s farm landscape. In 1920, about 26 percent of the U.S. labor force worked in agriculture, tending small, largely manual operations. (gilderlehrman.org) Today, precision-guided tractors, biotech seeds, satellites, sensors and AI let fewer than 1 percent of workers—roughly 0.9 percent—produce the nation’s food. (bls.gov) These technologies have multiplied output: USDA estimates show total farm production is now almost three times its 1948 level, a roughly 300 percent gain, even as farmland and labor inputs have fallen. (ers.usda.gov) In effect, technology turned farming from a mass occupation into a high-tech engine whose productivity per worker has soared more than fifteen-fold, supplying more food than ever with a fraction of the hands once required.
Where did those workers go? To other growing industries that were enabled by their own tech advances, manufacturing, and services.
Where the Farm Hands Went: A Century-Long Labor Migration
1. From fields to factories (1920-1960)
- Great relocation to industry. In 1920 roughly one-quarter of working Americans still earned their living on farms (about 10 million people). Mechanization and rising urban wages pushed them toward the booming industrial cities of the North and West. During World War II alone, the Department of Agriculture recorded more than 2 million men leaving agricultural jobs between 1940-42, and the nation’s farm population fell by 6 million by 1945 — yet food output still rose 32 percent thanks to tractors, hybrid seed and fertilizer. (archives.gov)
- Factory work absorbed them. By the time the post-war boom peaked in 1960, manufacturing plants employed 26 percent of all U.S. workers, up from about 18 percent in 1920. (visualcapitalist.com) Ex-farmers (and their children) found jobs building cars and appliances, pouring concrete for the Interstate Highway System, and staffing the new petro-chemical, meat-packing and textile plants that clustered around rail hubs.
- The Great Migration. Millions of African-American share-croppers in the South traded tenant farming for unionized production lines in Detroit, Chicago and Los Angeles, while Dust-Bowl farmers headed west to California’s defense plants. (nfwm.org)
2. The service-sector era (1960-2025)
- Shift from goods to services. Once machines began replacing assembly-line labor, the same story that had played out in farming repeated in factories. The Bureau of Labor Statistics shows service-providing industries jumping from 31 percent of employment in 1900 to 78 percent by 1999; today they account for roughly four-fifths of all jobs. (bls.gov)
- Where former farm families work now. Successive generations spread into:
- Construction and logistics as suburban housing and interstate trucking exploded.
- Retail, hospitality and food service that grew alongside car-centric shopping malls and tourism.
- Health care, education and government, which expanded with an aging population and the GI Bill–fueled college boom.
- Information and professional services — everything from coding crop-insurance software to designing the GPS-guided combines that replaced their grandparents.
3. Why such a smooth absorption?
- Productivity freed labor. USDA data show total farm output almost tripled while labor hours fell more than 80 percent between 1948-2017. (agamerica.com) Higher yields per worker let millions exit agriculture without creating food shortages.
- Expanding demand elsewhere. Wartime procurement, baby-boom consumption, Cold-War research funding and, later, the digital revolution created new job slots faster than mechanization destroyed old ones.
- Education & mobility. High-school completion rates soared from 17 percent (1930) to 90 percent (today), equipping rural youth for non-farm work, while affordable cars and the FHA mortgage program made relocation feasible.
4. Who still farms?
- Less than 2 percent of today’s U.S. workforce is on the farm, and only about 1 percent are hired wage workers. (ers.usda.gov) That small core is increasingly supplemented by seasonal H-2A visa holders and immigrant families, while the descendants of yesterday’s farmhands mostly sit behind computer screens, cash registers, hospital beds or steering wheels.
Bottom line: Mechanization didn’t strand a generation of farm workers; it set off a two-step migration — first to the factory, then to the service economy. A job market that keeps creating new kinds of work has continuously re-absorbed the labor freed by technology on the land.
Every employee at every company I have ever invested in or worked at has more to do than time to do it. AI, especially when applied to specific industries with specific skill, will enable many workers to become 10x workers. What will the companies do with that productivity? Some will hire fewer people, but most will sell more and become larger companies with flat employee growth, in my opinion.
I have been putting my money where my mouth is. Here are a few companies I have backed in Worker Augmentation:
- Quintess AI. Voice agent for heavy equipment maintenance organizations (rail, truck, etc.). These workers were using desktop apps (took them out of the field) or paper to document their work. Documentation took 10-15% of the day away from doing actual work. The average maintenance tech is 60 years old. Giving them a voice agent to update tickets and document work (unstructured data to structured database) saves 90% of documentation time.
- Grw.ai. 24/7 sales coach for every SDR. It reads every email, message, CRM comment, and can give much better context on deals than even a human VP of Sales. Costs 10% of traditional 1:1 sales coaching from a human.
- Pinnacle Global LLC, 24/7 AI coaching for any manager.
- Procurable.ai augments a procurement analyst who is taking data from multiple systems and using Excel to report. Also, building new features like a “should cost” model from a PDF BOM.
- Enhanced.ai, there is a shortage of AI engineers to build enterprise-scale AI workflow apps. This platform enables nonprogrammers to build and deploy production-ready AI workflows.
Systems integration at scale
While there is considerable discussion about AI agents replacing systems of record, history shows us that systems of record are tough to replace. I bet that the replacement will not come at scale in the 5-10 year timeframe. Instead, the AI layer will provide systems integration at scale and extend these systems of record. Examples of my bets on this thesis include:
- WithCiviq.com, the world’s first structured database of $1T local government spend (cities, counties, school districts). AI collects data from 180,000 websites, public record requests (PDF files), and public video meetings to structure who is spending what and who is supplying what.
- Module, a dashboard for 3PL warehouses that integrates with multiple systems of record to provide a better dashboard for customers on the exact status of their inventory. This was previously done by data analysts in Excel.
- Vega Cloud, AI agents connect to all your various cloud providers, monitor usage across production and dev environments, identifying overprovisioning that humans miss due to the huge amount of data. Refactor the cloud environment in real time to save expenses.
- Tembo Plus, Single API platform to build and launch fintech products across Africa’s multiple countries and currencies in a compliant way. Like Plaid, one API platform saves app developers integration and maintenance time.
- Nuel.ai, supply chain optimization explicitly built for process manufacturing industries, integrating historical and real-time data. Replacing spreadsheets.
Summary
Overall, I am most interested in how AI unlocks previously intractable problems. I am seeing more of these every month and it is an exciting time to build in blue oceans.
