Evos Raises $9M From Uncork and Field Ventures to Build the Physical Economy's AI Workforce

As many as 1.9 million US manufacturing jobs could go unfilled by 2033, according to Deloitte and The Manufacturing Institute, while the people retiring out of those industries are taking decades of unwritten judgment about how their companies actually run with them. Evos, a New York startup building autonomous AI operators for the businesses that make, move and power the economy, came out of stealth today with a $9 million pre-seed round to capture that knowledge before it disappears. The round was co-led by Uncork Capital and Field Ventures, with Crosslink Capital, Utopia Capital Management, Sandwith Ventures and angel investors also taking part. It lands after Evos passed $1 million in annual recurring revenue just four weeks after going live.

Evos operators already answer broker calls, build freight loads, chase missing paperwork and reconcile invoices, the kind of busy work that small industrial companies rarely have enough people or hours to finish. Each operator learns a job from a company's most experienced staff, works inside the software the business already uses and only takes on a task by itself once its decisions consistently match the team's. Evos will use the new capital to grow its team and bring operators to the small and mid-sized companies in physical industries that software has been slowest to reach.

The Market for AI Operators

Businesses are moving quickly from AI that answers questions to AI that does the work. The global market for AI agents is forecast to grow from $7.84 billion in 2025 to $52.62 billion by 2030, a compound growth rate of 46.3% a year according to MarketsandMarkets, while Gartner expects a third of enterprise software applications to include agentic AI by 2028, up from less than 1% in 2024. Gartner also predicts that at least 15% of day-to-day work decisions will be made autonomously by AI agents within the same timeframe.

Most of that spending has so far gone to office software and large enterprises, yet the physical economy offers a far bigger pool of repetitive operational work. US trucking alone generated $987 billion in revenue in 2023, according to the American Trucking Associations, while behind every load sits a chain of calls, quotes, paperwork and invoices that still depends on people working through inboxes and spreadsheets.

The Physical Economy's Long Tail

The companies doing that work are overwhelmingly small, as 99.6% of US trucking companies run fewer than 100 trucks and 95.5% run 10 or fewer, according to the ATA's latest trucking trends report, while 93% of US manufacturers have fewer than 100 employees and 74% have fewer than 20, according to the National Association of Manufacturers.

Each of those businesses runs its own way, on decades of decisions about pricing, carriers, customers and exceptions that were never written down. That is a large part of why enterprise software has struggled to reach them, since a tool built for a 5,000-person company rarely fits a 40-person fabricator or a 12-truck carrier without months of customization that a small business cannot afford.

The Knowledge Gap

The people who hold that knowledge are leaving faster than they can be replaced. US manufacturers could need as many as 3.8 million new employees between 2024 and 2033, with half of those roles could go unfilled if skills and applicant gaps persist, while 65% of manufacturers named attracting and keeping talent as their top business challenge in the National Association of Manufacturers' 2024 outlook. A record 4.1 million Americans turned 65 in 2024, the peak of a retirement wave that is pulling experienced workers out of factories, dispatch desks and back offices.

When a dispatcher with 20 years on the job retires, the company loses more than a headcount, because it loses the person who knew which carrier to call for a late Friday load, which customer always disputes accessorial charges and which invoices can be approved without a second look, while that judgment usually leaves without ever being recorded anywhere.

How Evos Works

Evos does not ask a business to trust an AI agent on day one. Every Evos operator starts behind an approval gate, where people approve, reject or modify each decision it proposes, with every one of those choices becomes training data that teaches the operator how the team actually works. As the operator proves its accuracy, it earns autonomy one task and one scenario at a time, while anything genuinely new is sent back to a person. Across its customers, Evos says its operators reach roughly 82% decision accuracy and handle about 70% of their work without human intervention.

The operators learn from behavior rather than written instructions, studying the decisions a team makes and the patterns behind them, while working inside more than 300 connected systems including SAP, NetSuite, Salesforce and the transport management systems freight companies already run. They are also given a function rather than a prompt, which changes how a team uses them. Most AI tools wait for someone to ask a question and then review the answer, whereas an Evos operator owns a job, so the team manages outcomes instead of inputs.

Speed is the other half of the pitch, since the Evos platform is fully self-serve. Work that would take a consultant or implementation partner 9 to 18 months to deliver, an Evos operator takes on in 24 hours from discovery to deployment. Evos calls this the Trifecta, made up of one hour with a company's experts, a working operator the next day and measurable results within a month, with the business deciding how far to go, from a person signing off on every step to an operator running entirely on its own.

The Founder Thesis

Urav Shah, founder and chief executive of Evos, argues that the most capable system inside many industrial businesses is not a piece of software at all but the person who has been there for 20 years, whose knowledge has never been written down. In his view an Evos operator should learn the job from those people, improve on it and prove it can match them before it takes on any work by itself, so that their expertise can be applied to every job, every day, so smaller companies can run the way the best operators in the world do.

Shah is 23 and has already spent five years building companies in deep tech and AI, founding four and exiting two. He built his first company at 18, developing generative world models for prompt-based game generation, later commercialized quantum algorithms for Fortune 500 customers at Terra Quantum and co-founded Navon, which builds modular data center infrastructure for the Global South. Evos, founded in 2026 as Evolutionary Operating System, Inc., describes itself as the agentic operations platform for the businesses the technology industry has paid least attention to.

Early Proof in Freight

The clearest results so far come from freight, where MyFreightWorld, a freight brokerage, runs an Evos operator that audits every less-than-truckload carrier invoice against its transport management system. In its first two months the operator made nearly 3,500 audit decisions at 97% accuracy, while in the last five weeks alone it closed out more than $600,000 of invoice value.

That is a job where small errors add up quickly, since an audit of nearly 250,000 LTL invoices by AFS Logistics found errors in about 4.5% of them, with inaccurate accessorial charges making up 42.8% of all mistakes and fuel surcharges another 14.4%. Mike Chalfant, president of MyFreightWorld, says Evos changed the company's billing process with no upfront investment, in a matter of days and weeks, raising accuracy and efficiency enough that the team now plans to give the operator more work.

The Investor Thesis

Richard Kerby, partner at Field Ventures, argues that for decades the software sold to industrial businesses has helped people organize and track their work but stopped short of actually doing it. He sees Evos changing that by learning the judgment behind how a business runs, working inside existing systems and taking ownership of entire operational functions, which is why he believes Evos can become the operating layer for the physical economy. Kerby also points to the speed at which Shah and his team turned that vision into measurable value for customers.

Tripp Jones, general partner at Uncork Capital, makes a similar case from the customer side, noting how quickly teams are putting Evos operators to work on core parts of their business without a long technology rollout, which brings powerful technology within reach of companies that software has long overlooked. The size of the round reflects that conviction, as US startups on Carta raised an average of $276,000 per pre-seed SAFE or convertible note in the second quarter of 2026, which makes Evos's $9 million round roughly 33 times the typical pre-seed instrument, at a time when AI companies took 49% of all pre-seed dollars in the first half of the year.

What Comes Next

Evos will formally launch its platform, Workhorse, which it describes as the workforce for physical industries, in the coming weeks. Its operators already cover freight brokerage work from carrier sourcing and dispatch to track and trace, shipper relations and billing, along with manufacturing tasks such as running the quote desk, order entry, purchasing, scheduling and quality paperwork for metal fabricators, plastics molders and electronics assemblers. If the model holds, the knowledge that small industrial companies have relied on for decades will stop walking out of the door at retirement and start compounding inside software that keeps learning from the people who built those businesses.

Don’t forget to like and share the story!