The Starvard sealStarvard.
English
EnglishFrançaisEspañolहिन्दीPortuguêsBahasa IndonesiaAll languages →
Starvard University

The Starvard Blog

Notes from the free university

Blue-Collar Bosses: AI Agents in the Trades

October 20, 2026 · By the Starvard professors
A small robot assistant answering the phone in a plumber's workshop at dawn
Illustration by Starvard (AI)

The popular story says AI is coming for the lawyers and coders while the plumbers stay safe, and it is half right. AI is arriving in the trades fast, with serious money behind it, but it is showing up on the phone, in the dispatch board and in the invoice queue, not in the crawlspace. This post tours the agents now answering calls and quietly managing the office side of America's heating, plumbing, electrical and roofing companies, separates the real numbers from the sales pitches, and asks what it means for the people who do the work.

It is 2:07 a.m. in January. It is ten degrees outside. Your furnace has just made a noise like a shopping cart falling down stairs, and now it is making no noise at all.

You grab your phone and call the first heating company Google shows you.

Here is the question this whole post is about: who picks up?

Ten years ago, the honest answer was "a voicemail box," or, if the company was big, a bleary answering-service operator three states away reading from a binder. Today, at a growing number of heating, plumbing, electrical and roofing companies, the answer is something stranger. A friendly voice answers on the first ring. It asks how old your system is, whether you smell gas, whether anyone in the house is elderly or has a baby. It looks at which technicians are on call, holds a 6:30 a.m. slot for you, texts you a confirmation, and tells you how to keep the pipes from freezing in the meantime.

That voice is not a person. It is an AI agent. And it may be one of the more consequential, least glamorous things happening in AI right now.

You have probably heard the popular take on AI and work. It goes something like: AI is coming for the lawyers, coders and copywriters, but the plumbers are safe, because a chatbot can't crawl under your house.

That take is mostly right. It is also missing the most interesting part. AI is arriving in the trades, fast and with a lot of money behind it. It is just not arriving where everyone was looking. It isn't showing up in the crawlspace. It is showing up in the office, on the phone, in the dispatch board, in the invoice queue. It is coming for the boss part of the business, not the wrench part.

So let's build this from the ground up: what a trades business is, why its weak point turned out to be a ringing phone, what an "AI agent" really means, who's building them, which numbers are real and which are sales pitches, and what it all means if you're thinking about a career with your hands.

Part 1: Every trades company is two businesses wearing one trench coat

Let's start with the simplest possible picture.

When you hire a plumber, you think you're buying one thing: someone who shows up and fixes the pipe. But that plumber's company is actually running two very different businesses at the same time.

Business #1: The Wrench. This is the physical work. Diagnosing why the water heater is leaking, sweating a copper joint, pulling wire through a wall, tearing off a roof in August. It requires years of training, a body that can do it, and judgment that comes from having seen a thousand weird basements.

Business #2: The Switchboard. This is everything else. Answering the phone. Deciding if the call is an emergency or a dripping faucet. Booking the slot. Routing technicians so trucks aren't zigzagging across the county. Writing estimates, chasing the unsigned ones, invoicing, ordering parts, running ads, training the new person on the phones.

Here is a rough diagram of a typical service call, split by business:

Notice something? The Wrench is the part customers think they're paying for, but the Switchboard touches the customer more times. And in a small shop, the Switchboard is often run by the same person who runs the Wrench, from the top of a ladder, with a phone wedged against his shoulder, while trying not to drop a pipe wrench on a homeowner's cat.

This is the key insight that the rest of the post hangs on:

The Wrench is physical, messy and different every time. The Switchboard is mostly talking, typing, scheduling and remembering.

And talking, typing, scheduling and remembering are exactly what modern AI happens to be good at.

Part 2: The leak in the bucket is a ringing phone

Ten years ago, few trades owners would have named "the phone" as their biggest money leak. It turned out to be the hole in the bottom of the bucket, because home services combine four traits:

1. The demand is spiky. Nobody calls about their air conditioner in April. Everybody calls during the first 100-degree week in June. You can't staff a call center for the worst day of the year without paying a dozen people to stare at silent phones the other 340 days.

2. The emergencies don't keep business hours. Pipes burst at night. Storms hit on weekends. A lot of the most urgent (and most valuable) calls come exactly when nobody is at the desk.

3. The customer doesn't wait. If your heat is out and it's ten degrees, you do not leave a polite voicemail and wait until Monday. You hang up and call the next company on the list.

4. Each call can be worth a lot. This is the one that makes everything else matter.

That last point is how one of the biggest companies in this space was born. Avoca was started by two MIT engineers, Tyson Chen and Apurva Shrivastava, who met at a poker night and originally built an AI to answer phones for restaurants. Then they ran into a Texas HVAC company called Rescue Air and realized they'd been aiming at the wrong target. As Chen put it to Fortune, a restaurant that misses a call loses a dinner order, but a home service company that misses a call can lose "a $30,000, $40,000 HVAC install."

That's the whole business model on a napkin. Same technology. A thousand times the stakes per call.

How many calls actually get missed? Here I have to put on my skeptic hat for the first of many times. If you go looking, you'll find vendor blogs confidently claiming that most contractor calls go unanswered and that almost none of those people call back. Those numbers get copied from blog to blog and I couldn't trace them to a solid original study, so I'm not going to repeat them as fact. What I can give you are specific, named examples:

So the leak is real, even if nobody has a perfectly clean number for its size. And once you see the leak, the next question is obvious: what if something could answer every single call, instantly, at 2 a.m., during a tornado, without ever getting tired?

Part 3: Okay, so what actually is an AI agent?

The word "agent" gets thrown around so much right now that it's lost most of its meaning, so let's rebuild it from scratch.

Step 1: A language model. At the core is a large language model, the same basic kind of technology behind ChatGPT, Claude or Gemini. On its own, a language model is a very sophisticated text predictor. You give it words; it gives you words back. It can understand "my AC is blowing warm air and the outside unit is making a clicking sound," and it can respond sensibly. But on its own, it can't do anything. It's a brain in a jar.

Step 2: Give it tools. Connect that brain to the company's calendar, customer database, technician list, price book and phone line. Now when it hears "my AC is blowing warm air," it can check the customer's maintenance plan, find a free tech, and actually book the appointment.

Step 3: Give it rules and permission. Finally, tell it what it's allowed to do and when to stop and hand off to a human. Book routine repairs, yes. Quote a price for a full system replacement over the phone, no. Someone says they smell gas? Stop everything and give them the emergency instructions and the gas company's number, then alert a human immediately.

Brain plus tools plus rules equals agent. ServiceTitan co-founder Vahe Kuzoyan used a nice image for this at the company's conference this October: AI models are like wild horses, enormously powerful but hard to direct, and the software that connects them to a real job is the harness. A model plus a harness is an agent.

It helps to think about this as a ladder, because a lot of confusion in this space comes from people calling very different things "AI":

In 2024, most of the action was at rungs 2 and 3. In late 2026, the big platforms are publicly racing toward rung 4 and talking openly about rung 5. Nobody is fully at rung 5. We'll come back to how close anyone really is.

Why the Switchboard falls first: a 38-year-old paradox

There's a deep reason the AI is arriving in the office before the crawlspace, and it has a name. In 1988, the roboticist Hans Moravec pointed out something that seemed backwards: for computers, the stuff humans find hard (math, chess, logic) is relatively easy, while the stuff humans find effortless (walking across a cluttered room, picking up an odd-shaped object, feeling whether a fitting is snug) is fiendishly hard. This is now called Moravec's paradox, and it has aged remarkably well.

Think about one crawlspace visit: squeeze through a 16-inch opening, feel that a joint is slightly damp, recognize a 1970s fitting nobody makes anymore, improvise an adapter, explain it to a worried homeowner. Every house is a mess in its own unique way. Now think about a call-center rep: listen, ask questions, look things up, book a slot. Skilled work when done well, but it lives almost entirely in words and data, which is precisely the world modern AI was built for.

So the first wave of AI in the trades looks exactly like Moravec would have predicted. The voice on the phone gets automated. The hands under the sink don't.

Part 4: Meet the bosses

Let's look at who's actually building these things, roughly in order from "answers the phone" to "runs the whole office." I'll give you real names and real numbers, and I'll flag where a number comes from the company selling the product, which, fair warning, is most of them.

The phone specialists

Avoca is the poster child. After that Rescue Air pivot in 2023 (they went through Y Combinator's Winter 2023 batch), the company built AI agents that answer calls and texts, book jobs into the contractor's existing software, follow up on unsigned estimates, and hand tricky calls to humans. In April 2026, Fortune reported that Avoca had raised more than $125 million across three rounds (Kleiner Perkins led its Series A; Meritech and General Catalyst led its Series B), at a valuation of $1 billion, with more than 800 customers. It reportedly crossed eight figures in annual recurring revenue in 2025 and is now pushing into moving companies, auto shops and property management. A billion-dollar valuation for software that answers plumbers' phones would have sounded like a joke five years ago.

Avoca's own website is full of customer testimonials: one company that booked 5 calls in two-thirds of a month with a live answering service and 43 in the last third with Avoca; a claimed 90% booking rate on after-hours calls; and one owner saying he can "run a $100M business with 9 CSRs" because the AI handles 70% of calls. Treat these the way you'd treat testimonials on any company's homepage: real people, presumably real results, but hand-picked by the seller.

One detail I did find genuinely interesting: at ServiceTitan's conference this month, Avoca said it tests every change to its agent, right down to the greeting, by splitting live calls between two versions and only rolling out the winner once the difference is statistically significant. That's the kind of relentless optimization no human call center can do, and it's a quiet advantage that compounds.

Netic is the other high-profile startup. It was founded in late 2024 by Melisa Tokmak, a Stanford computer science graduate who had been a general manager and chief of staff at Scale AI. Netic emerged with $20 million in funding and raised a $23 million Series B led by Founders Fund in November 2025, for about $43 million total according to the research firm Sacra. Its pitch is a bit more ambitious than "answer the phone": it wants to run the whole revenue engine, including inbound calls, texts, web chat, and outbound campaigns to existing customers.

Netic's showcase customer is Hoffmann Brothers, the St. Louis company from Part 2, and its case study is worth walking through because it shows the progression that a lot of companies go through:

  1. Start small. Hoffmann Brothers first used Netic only for after-hours and overflow calls, the ones nobody was answering anyway. Low risk.
  2. Listen obsessively. Chris Hoffmann says he spent countless hours listening to the AI's calls before trusting it further. He'd started out deeply skeptical; his line was, "For every one good tool, there are 99 bad ones."
  3. Go first-touch. Once he was confident, the company switched to what Netic calls "Netic First": the AI is the first responder on every inbound lead, across phone, text, chat, Google ads and lead sites like Angi, with humans stepping in for judgment calls.
  4. Go outbound. Then they pointed the AI at their 20,000-member maintenance plan, texting members at the right time to book tune-ups.

The claimed results: 70% of eligible jobs booked entirely by the AI; outbound booking rates up from 5-10% to about 25%; roughly 2,900 new jobs and $665,000 in revenue from automated outbound campaigns; and 15% year-over-year revenue growth in the second half of 2025. Those are vendor-published numbers, and a company's revenue grows for lots of reasons (weather, pricing, a tornado that damages thousands of homes), so I wouldn't credit all of that growth to software. But the progression itself is the real lesson, and it matches what other operators describe.

The giants bolting AI onto what they already own

Here's the thing about startups like Avoca and Netic: they plug into software the contractor already runs. And the biggest company in that software is not sitting still.

ServiceTitan is the dominant operating software for larger trades businesses. The publicly traded company says more than 10,000 contractors run on it, processing over $62 billion in transactions a year. In September 2025, at its Pantheon conference in Anaheim (more than 4,000 attendees), it unveiled Atlas, an AI assistant you can talk to in plain English to run reports, find jobs and dispatch technicians, built on its own trades data plus outside models like Google's Gemini. Co-founder and CEO Ara Mahdessian's framing at the time: "AI isn't going to replace contractors, but it will absolutely transform how they operate."

Fast-forward to Pantheon 2026, held this month in Orlando, and the ambition has escalated considerably. According to ServiceTitan's own conference coverage:

ServiceTitan isn't alone. Housecall Pro (about 45,000 smaller businesses) sells "CSR AI" and advertises an average 42x return on investment, defined as revenue from AI-booked jobs divided by the AI's cost (generous, since some callers might have called back anyway). Jobber says its AI Receptionist handles about 20 calls a week for the average business.

That's the big open business question, and Sacra flags it as a risk for Netic: if the software everyone already uses bundles a "good enough" agent, standalone startups must be dramatically better to justify a separate bill.

AI in the truck, not just the office

Not all of this is about phones. A few companies are putting AI in the technician's pocket.

Bluon makes an app for HVAC technicians that includes MasterMechanic, an AI troubleshooting assistant. What makes it interesting is its training data: not the open internet, but recordings of real tech-support calls between field technicians and veteran experts. It launched in 2023 trained on about 50,000 such calls; by 2026 the company says it's trained on 150,000 calls handled by 40 top technicians. CEO Peter Capuciati claims it solves 95% of problems, and, tellingly, Bluon also relaunched a live hotline staffed by human techs with 20 to 50 years of experience for the cases the AI can't crack. The 95% figure is the company's claim; the existence of the human backup line is the honest part.

Why it matters: the people who know how to fix a weird 30-year-old boiler are retiring, and an AI trained on veterans' answers is an attempt to bottle that expertise before it walks out the door.

Rilla does something sneakier. Sales in home services happen in person, at the kitchen table, where no manager can watch. Rilla's app records those in-person conversations (with consent), transcribes them, and uses AI to score them and flag coaching moments. According to Sacra, a sales manager can review a roughly three-minute highlight clip instead of spending three to six hours physically riding along in a truck. Private-market data puts Rilla's funding at around $79 million, and Home Depot was announced as a customer in early 2026.

And on the big job sites

Commercial construction is getting its own version. Procore, the leading construction-management software company, launched an Agent Builder in open beta in October 2025 that lets construction firms create custom agents by describing them in plain language, plus prebuilt agents like one that drafts RFIs (formal "requests for information" between builders and designers) and searches project documents for answers. ServiceTitan is moving into the same territory with agents that build a foreman's daily log from voice notes and photos.

None of this lays a single brick. All of it eats paperwork.

Part 5: Why "bosses"? Because the agents are starting to manage

So far this might sound like a story about receptionists. It's bigger than that, and the reason has to do with a piece of math that's easy to miss.

Here's an illustration Mahdessian used in his Pantheon keynote. (It's his hypothetical, not a measured result, but the arithmetic is worth understanding.)

Imagine a contractor with:

That's 6,000 jobs and $4.8 million in revenue. At a 10% profit margin, $480,000 in profit.

Now improve each of those three numbers by just 10%: 11,000 leads, a 66% booking rate, an $880 ticket. Multiply it out and revenue climbs to about $6.4 million, a 33% increase from three modest 10% improvements. And here's the kicker: if you got those improvements from better software rather than more trucks, more ad spend, and more office staff, most of the extra money falls straight to profit. In his example, profit roughly doubles.

That's the compounding logic behind "blue-collar bosses." A single agent answering phones is nice. A team of agents that each push one number up a little, and share information so they don't work against each other, is a different animal. And it explains why the products are drifting from "answer the phone" toward the jobs that used to belong to managers:

And the owners are noticing what this does to the shape of their companies. Jake Wold, COO of Guild Garage Group (29 businesses, more than 1,000 employees, about $350 million in revenue), told the Pantheon crowd he wants revenue to grow fast and headcount to grow slowly. His example was blunt: he doesn't want to be at 2,000 employees when the company hits $700 million. Two of Guild's businesses already run with AI answering the call, booking it, and dispatching the technician.

Or consider Riley Plumbing, Heating & Air, which Brooke Riley started at age 19 in 1999. On ServiceTitan's podcast, the Rileys described using AI voice agents to handle 100% of their inbound call-taking, and opening a location in Reno with no office staff at all.

That's the real headline. The trades owner used to be a technician forced to also become dispatcher, marketer and accountant. Increasingly, the owner becomes a manager of agents.

Part 6: The honest ledger (hype vs. reality)

This is a Starvard post, which means we don't just pass along the sales pitch. So let's turn the flashlight around.

Problem 1: Almost every number comes from someone selling something

Look back over Part 4. The 98% booking rate, the 42x ROI, the 15% revenue growth, the 95% troubleshooting success rate: every one of those came from the company that makes the product or from a customer story that company chose to publish. That doesn't make them false. It makes them unaudited. There are no large independent studies yet comparing similar contractors with and without AI agents over a full year.

And the companies on stage are the ones it worked for. Nobody keynotes about the shop that tried an AI receptionist for three weeks and turned it off.

Problem 2: Lots of people hate talking to bots

This one is real. The Better Business Bureau reported that complaints tied to AI services rose from about 22,000 in 2023 to nearly 34,000 in 2025, about a 55% increase, covering everything from chatbots that can't solve problems to phone systems with no path to a human. A survey run by OnePoll for a live legal receptionist company (so, an interested party) found that 89% of 6,000 consumers prefer a real person when calling a law firm, and one in three said they'd hang up on an AI.

The fair counterpoint is that the realistic alternative at 2 a.m. is usually voicemail, not a warm human. But "better than voicemail" is a low bar, and replacing a great dispatcher with a mediocre bot to save money is a bad trade.

Problem 3: AI makes a mess faster

The single most repeated warning at this year's ServiceTitan conference wasn't about AI at all. It was about data. Jennifer D'Ambra, CTO of HomeX Services Group, said her dispatch AI was stuck at about a 30% optimization rate until her team went back and cleaned up how they'd defined technician skills, job types and business units, at which point performance jumped. Her point: no matter how sophisticated the tool, if the setup is sloppy, it's just processing bad data and producing bad results. Best Choice Roofing, with more than 80 branches in 25-plus states, told Roofing Contractor it standardized workflows first, then layered AI on top, and said fear, not readiness, is the dominant emotion about AI among its roofing peers.

Problem 4: The "too many apps" trap

Every vendor claims to be the answer, so a contractor can end up with six AI tools that don't talk to each other. A ServiceTitan product manager described the classic failure: marketing doubles ad spend in a heat wave, the schedule is already full for nine days, and the company pays for leads it can't serve. Operator Steven Gurley made the grounded version of the point at Pantheon: his technicians already struggle to open one app, so why add six more?

Problem 5: Phones are where the safety-critical calls come in

A gas smell or a sparking panel is a real emergency, and the AI is sometimes the first to hear about it. Responsible systems recognize danger words, give basic safety instructions and escalate to a human immediately rather than improvise. Ask any vendor exactly how theirs does that.

Problem 6: Somebody's job is changing, and it isn't the plumber's

Here's the part the cheerful press releases tend to skip. When an owner says he can run a $100 million business with nine call-center reps because the AI handles 70% of the calls, the obvious question is: how many reps would that business have needed otherwise?

The vendors' answer is that AI takes the routine calls so humans can help the callers who need it, which Superior Plumbing described at Pantheon. That's plausible, especially in chronically short-staffed shops. But it would be dishonest to pretend the dispatcher and customer-service roles aren't the ones most reshaped here. Part 1's diagram tells the story: the Switchboard jobs are the exposed ones.

Part 7: So is the wrench actually safe?

Let's zoom out to the question most people actually care about.

The case that it's safe (for now)

In 2025, Microsoft researchers analyzed 200,000 anonymized conversations people had with Bing Copilot and scored occupations by how much their work overlaps with what AI chatbots can do. At the very bottom of the list, the jobs with the least overlap, were dredge operators, roofers and roofers' helpers, alongside a lot of other hands-on work. The lead researcher, Kiran Tomlinson, was careful to say this measures where chatbots are useful, not which jobs will disappear. And it only looked at chatbots, not robots. Still, it confirms what Moravec's paradox predicts.

The robots aren't close, either. Autonomous robots on real job sites mostly work in predictable, mapped spaces like warehouses, and 2026 coverage describes humanoids as largely in the demo stage, often remote-controlled by humans. Geoffrey Hinton, one of the godfathers of modern AI, has suggested plumbing is a comparatively good career bet, but with an important caveat: that advantage might last only "until the humanoid robots show up."

The case that the robots are coming eventually

And here's a sign that the robot builders know exactly what's missing. This month, an AI data company called HUMXN announced it would subsidize free plumbing, electrical, HVAC and duct-cleaning repairs for homeowners in three cities, in exchange for letting technicians wear "wearable capture technology" that records how skilled workers actually do the job. The stated goal is to collect training data for robots across different homes, equipment and problems.

Part of why robots are bad at crawlspaces is that there's no internet-sized pile of crawlspace data to learn from. Companies are now paying to build that pile. Robot plumbers aren't imminent, but "never" is the wrong word.

Meanwhile, demand for the wrench is huge

While everyone debates whether robots will take trades jobs, the more immediate problem is the opposite: there aren't enough people to fill them.

The honest caveats

Before anyone sprints to trade school because a blog told them to, two cautions.

First, trades are cyclical. Fortune has pointed to research showing that trade jobs are closely tied to construction and manufacturing, which makes them sensitive to economic downturns, and some trade roles rank poorly on entry-level job lists. Even the data center boom has hit friction: Fortune noted data center construction actually slowed last year for the first time since 2020, as developers ran into zoning, permitting and power-supply delays.

Second, "AI-proof" is a slogan, not a guarantee. The more accurate statement is: the physical core of skilled trades work is among the least exposed work to current AI, and the business side of trades work is among the most actively transformed. Both are true at once.

Mike Rowe, the former Dirty Jobs host who has been preaching the trades for two decades, has described his own attitude toward AI as "skeptimistic." Honestly, that's about the right word for all of this.

Part 8: What this means for you

Let's make this practical. Depending on who you are:

If you're thinking about a career in the trades

The trades look like a genuinely strong bet right now: big worker shortages, rising demand, and core work that AI can't do. (For a baseline, the Bureau of Labor Statistics put the median plumber's pay at about $63,800 a year as of May 2025, with typically a paid four-to-five-year apprenticeship rather than tuition debt.)

But here's the twist this post has been building toward: the tradesperson who does best over the next decade will be good at both businesses in the trench coat. Excellent with the wrench, comfortable with the Switchboard tools, able to use an AI troubleshooting assistant without blindly trusting it, and, if you ever own a shop, able to manage a team of agents.

The physical skill is the moat. The AI fluency is the multiplier.

If you own (or want to own) a trades business

Watch what the operators who succeeded actually did, rather than what the vendors promise:

  1. Fix your basics first. AI amplifies whatever you already are.
  2. Start where the risk is lowest: after-hours and overflow calls.
  3. Listen to the calls before expanding, as Chris Hoffmann did.
  4. Expand only when results hold: first-touch, then outbound.
  5. Avoid tool sprawl. Fewer, connected systems beat six clever ones.
  6. Always leave a door to a human.

And keep an eye on one emerging idea: ServiceTitan launched a consumer marketplace called Homh as a plug-in inside ChatGPT, Gemini and Claude, betting that homeowners will increasingly ask their AI to find and book a contractor. In that world, it's not just your AI answering the customer's call. It's their AI negotiating with yours. Kuzoyan admitted the numbers are small today. Small numbers in a new channel are exactly how the early internet looked, too.

If you work the phones or the dispatch desk

This is the role changing most. Become the person who knows why the AI booked that job wrong and how to fix the rule. "AI supervisor" is a real job now, even if nobody's put it on a business card yet.

If you're a homeowner

Expect instant answers at 2 a.m. And when you need a human, ask for one. If a company's system won't get you there quickly, that tells you something about the company.

Conclusion: back to 2:07 a.m.

Let's return to that dead furnace and that ten-degree night.

In the old world, your call probably went to voicemail, and you dialed down the list until someone picked up. In the new world, an AI answers on the first ring, figures out that you have a cold house and a toddler, books the earliest slot, and tells the on-call tech everything before they even start their truck.

But notice what didn't change. A human being still drives over at 6:30 in the morning. A human being still kneels in front of your furnace, smells the burnt-dust smell, notices the cracked igniter, and fixes it with their hands. The AI made the Switchboard faster. It did not touch the Wrench.

That's the shape of AI in the trades in late 2026. Not robot plumbers, but software absorbing the office work that used to bury the people best at the physical work. Investors are betting big on it precisely because it's unglamorous, and because a missed call here costs thousands of dollars instead of forty.

Will it work as well as the keynotes promise? Probably not everywhere, and not as fast. The numbers are mostly self-reported and plenty of shops will buy tools before they're ready. But the direction is clear.

Kuzoyan put his bet this way: some people call AI the great equalizer, but he thinks "it will be the great multiplier." The gap between the trades businesses that learn to use it well and the ones that don't is likely to widen. The same is probably true for the people who work in them.

The wrench is still yours. The switchboard is getting a new boss. The smartest move is to learn to be the boss of the boss.

Why this belongs at Starvard

Starvard exists for people who learn with their hands as happily as with their heads, and this story sits exactly on that seam. The post's two-businesses problem is really an education problem: the tradesperson who wins the next decade is fluent in both the Wrench and the Switchboard, and both halves can be learned for free. If the agent half is new to you, AI Agents: Your First $10,000 is our ground-up course on what agents are and how people earn with them. If the wrench half is the draw, Apprenticeships and the Skilled Trades maps the paid routes into the work itself, no tuition debt required. And if Part 5's compounding math started you thinking about owning the shop someday, Entrepreneurship 101 is the plain-English place to start. All three are free, like everything at Starvard.

Watch

Sources

Tagged: Technology · Business · AI Agents · AI at Work · Startups

Back to the blog

A note on accuracy: Much of what you read at Starvard was created with help from AI agents, working alongside people. We check what we can, but we cannot guarantee that everything here is true, current, or complete. If something really matters, your health, your money, a grade, a job, please verify it with a trusted source before you rely on it.