Samit Kalra is a partner at 1984 Ventures, a $70 million pre-seed and seed fund. The firm backs founders rebuilding legacy industries with vertical AI, plus digital health and infrastructure software. Before 1984, Kalra invested at Bain Capital Ventures and AngelList. More than 75% of the 1984 portfolio has gone on to raise a Series A.

His core idea is about pricing. Automating a job lets you charge far more than software ever could. Charge 5x more and a sector that had no budget for tooling turns into a real market. Industries nobody wanted to touch three years ago suddenly look venture-scale.

That is why his view matters for the Physical AI and hard-industries thesis. The dirty, dusty, and dangerous work of the physical economy still runs on people. Once AI can read, write, speak, see, and listen, most of that work is up for grabs. Kalra spends his days working out which parts, and how founders should build so a competitor cannot copy them in a quarter.

The conversation

Jay: Make the case. Why are vertical labor markets the best place to build and invest over the next decade?

Samit: A few reasons. One, these markets often lack a clear software player, so you are not coming up against someone with strong existing distribution. Two, you can charge a lot more than you could before. You can take a sector that was unprofitable to sell software into and turn it into something where you charge 5x more. And labor automation is a deep pain point. What we look for as early-stage VCs is a deep problem to solve. Automating somebody's job, to put it crudely, is a deep pain point.

Jay: Are there certain niches that better fit that paradigm?

Samit: Right now, if I generalize, AI can read, write, speak, see, and listen. A lot of tasks are well suited to it today. Many jobs involve repetitious workflows, and those sit pretty well inside the AI bullseye. What is not in the bullseye yet is a job that can be automated but carries some stamp of trust. Tax services, advisory services, where you are not just buying the work, you are buying the peace of mind that it passes a quality threshold. Nearly everything else offline, labor-wise, is automatable.

Jay: What would have to change for people to trust AI on those high-trust jobs?

Samit: I do not think that happens quickly. KPMG can make a large mistake and KPMG still sticks around. But if a fully AI KPMG makes a large mistake, the expectation is there are many more mistakes underneath it. It is going to take years. I would guess it is more than five years out.

Jay: You have written about focusing on revenue lifts, not cost cuts. Why?

Samit: AI can make businesses more money and it can save them a lot of cost. We care about those two more than time-saving. Five years back, SaaS ideas were about saving time or a small workflow improvement. That is not a high enough bar anymore. In many cases AI means making more money, which opens up whole sectors. In other cases it means cutting jobs, which brings down cost. Topline up or bottom line up, both are enough for us. If someone says, I can make this workflow a little faster, we are less excited.

Jay: How do you think about model improvements through that lens?

Samit: One key learning this past year is that it is easy come, easy go. You grow really fast doing one thing. But if it works that well for you, it works well for everyone else too. So you get a lot of competition for every idea, and you had better move beyond that first thing fast. The way to think about long-term stickiness is to do hard things outside of the AI. AI is flashy and it works well. The question is what you are doing that is not AI that makes you a sticky business.

Jay: The old Paul Graham line, do things that don't scale.

Samit: It is do hard things over and over that others would also have to do to catch you.

Jay: You have written that to survive in vertical AI, you have to get multi-product within the first year. Why the first year?

Samit: Everybody is obsessed with revenue growth. But we are one year into a five-year transformation. Just because you had no rivals this year does not mean that holds next year. Multi-product, all-in-one is a strong pitch, especially when that first business line gets commoditized, and not just by other startups. Incumbents come after you too. The vibe-coding tools make that possible now. We have seen founders build a five-business-line company in five months.

Jay: Give me an example where that worked, and one where the second product was harder.

Samit: We backed a company in the rippling-for-offline-workers space. Two-person team. They built scheduling, time tracking, payroll, and others. We did not think anyone could do that until we saw it. Now they can go to offline customers and say, why would you not use us? One all-in-one tool, underpriced relative to the five things you were paying for. On the other side, we have had companies fall for the idea that revenue growth was going so well. They had one thing they found before others did, and incumbents offered something similar, and the margin compression is real.

Jay: When are dirty, dusty, and dangerous businesses good venture-backable companies versus better suited for a PE roll-up?

Samit: We just had a couple of companies get bought out post-seed by incumbents for pretty sizable dollars. Founders who sold only 20% of the business walked away with a lot of money. There is a lot of appetite from incumbents to buy AI-first upstarts that might threaten them long term. What makes an offline business venture scale is whether you can really tap the labor automation idea and increase ACVs. If something used to cost 5K and you can now charge 50K, that is a 10x bigger market. Everyone has figured this out, which is why you see investors backing companies in sectors they would never have touched three years ago.

Jay: What's an example?

Samit: Debt collections. Selling software to collections agencies, except now you can do the human's job. Many call-center businesses. We backed a company in recruiting that automates interviews. Surprisingly, candidates love being interviewed by it. They feel it is a safer interviewer and that they are better heard. We did one in pest control too. It is a system-of-record opportunity. ServiceTitan IPO'd, a roughly $10 billion company, but it focuses on large enterprise home-services businesses. The smaller shops are five-person teams. They are not going to buy software. The pitch was, we will automate every call for you. 24/7 call answering. Then, what if we also do dispatching? What about sales coaching for your reps?

Jay: How much does vertical-specific knowledge and data matter, or does this roll up into one horizontal AI play?

Samit: ServiceTitan actually bought a pest control system of record, and it is bad. All the pest control companies hate using it. It is easy to assume the big player just adds AI. But many industries are more nuanced. Pest control is recurring usage, whereas landscaping is more one-off, and that changes the dispatch routes. That is a very different algorithm. My view after five-plus years investing is that everything is more complicated once you get into it. So why does one company win when the offering looks similar? Founder execution. Solving people's needs better, going deeper.

Jay: How do you rank a founder? What is highest priority?

Samit: Is the founder brilliant in some way, does that link to the business they are building, have they found a deep enough pain point, and can they ship software fast? Domain expertise is not essential. Many people get up to speed on a space in months. The software skill set really matters, because engineers can often see solutions others cannot the moment they hear the problem. So we back engineering-first teams. And we index on a hair-on-fire problem. If the pain is not deep enough, we are not interested.

Jay: That is a hot take. Most people at seed talk about founder-market fit. How does that square with your point that industries are more complex than people think?

Samit: The main thing people lack right now is insight. The world exists the way it does for a reason, even when it looks silly. We would wait until someone said something that made us sit up. Sometimes that insight comes from someone who knows the space deeply. Sometimes from a smart person who started poking around and kept asking why. There are many more young founders now, and many have fewer insights. Maybe you need less of an insight these days because there are so many hair-on-fire problems. But if you do not have an insight, your chance of finding a true problem to solve is lower.

Jay: Do VCs overindex on market size?

Samit: You need your companies to be funded by follow-on investors, so you cannot be alone in thinking something is big enough. But many markets are expanding now by tapping labor budgets. My real observation is that it is so easy to overfocus on the first thing a company does. Usually, if you do the first thing well, that same customer offers up other problems to solve. So we do not focus on market size too much, because the best founders keep finding more around it.

Jay: What is a seed anymore? When do you tell a founder to raise a Series A versus a jumbo seed?

Samit: It is very hard to raise a Series A right now, and you get no credit for things you say you will do but have not done. The million-dollar topline benchmark from the past is a seed benchmark now, not a Series A benchmark. You are probably aiming for $2 million-plus to raise an A. For revenue, it has to be high-margin, low-churn. It can be usage-based. It is not just recurring revenue, but it has to be good revenue. We have seen margins drop to around 60%. What is super risky is the negative or zero gross margin thing, where people count pass-through inference as their own revenue. If your gross margin is zero or below, you are in trouble unless something changes.

Jay: What makes an exceptionally well-run Series A process?

Samit: Do not pitch a handful of firms. You have to pitch 30 to 40 firms to make sure you get that one term sheet. This is a full-time job for the CEO for nearly two months. Find people who actually like what you are doing. It is nearly impossible to convince someone skeptical about your space. And you get no credit right now for things you have not done yet. So only go out when you have the best story and the best recent traction.

Jay: If a raise lasts longer than two months, did the founder do something wrong?

Samit: If someone is dragging their feet, they are probably not going to invest. The best way to get multiple people excited at the same time is to pitch them at the same time. A week or two of back-to-back meetings, every first meeting at the same point. That builds FOMO. Investors hate losing to someone else.

Jay: I tell my founders never to send a data room. A fundraising process is trading information for engagement. If they are not engaging, what do I have left?

Samit: The fundraising narrative matters so much. When you send a pile of information, you cede the story you are trying to tell. People analyze the data and say, that churn number is higher than I thought, and you are in those questions on their terms, not yours. It is storytelling. It is getting people greedy about your business rather than fearful.

Jay: What do you wish more founders would stop doing when they pitch you?

Samit: If someone bombards you with facts, it is difficult to take them all in, especially in a space you barely know. The purpose of a first meeting is to get the second meeting. Think about any interesting meeting you have had. You walk out with one or two things that were unexpected, and you go check if they are true. I like when founders lead with a narrative disruption. One example: you think selling into law enforcement is a long sales cycle? We go from sale to onboarding in 30 days. If that is the only thing I remember about you after the call, I will be thinking about it for days.

Jay: How will category-defining vertical AI companies get built?

Samit: Find a wedge that is super valuable. Quickly become multi-product. Aim to become the system of record. Play strategically with the existing system of record, because you are often leeching off them at first. It is a higher bar to win now. You have to be a faster builder, a better fundraiser, a good forward thinker, with an ambitious product roadmap.

Jay: How much of that can be learned versus batteries included?

Samit: It is rare for an engineering founder to go learn sales. I do not think you go from bad storyteller to great one in the year between seed and A. So you do not want to back founders you have to shine up too much. You would rather back someone you know is great that others have not met yet. That is why we back people with some kind of brilliance. A brilliant salesperson, a brilliant marketer, the most ambitious person in the space. If things work, that person has something to hang their hat on.

Jay: A year out, what would have to change for the Series A to flip, or does it just get harder?

Samit: I think it will flip. It is too restrictive right now. There are many good companies not meeting the Series A bar that have product-market fit. That is why the jumbo seed class exists. In many cases 20% dilution makes sense. A 5-on-25 round can become 6-on-30 or 7-on-35 in the blink of an eye if you get multiple funds bidding. The Series A bar has gotten too high. There is an over-rotation on topline growth right now. There are a lot of worthy companies not yet getting the eyes of Series A investors. Those companies just need to keep doing what they are doing. If you get to $5 million or $10 million of revenue with high quality, those are impressive numbers. I have quiet hope for those quiet compounders.

Jay: Looking a year ahead, what are you most confident will still be true, and what are you least confident about?

Samit: Most confident: the way to build a long-term sticky business is to do a bunch of unsexy, non-AI things that are hard for others to replicate. Painful integrations. Proprietary partnerships. Curating an offline network someone else would also have to go build. All the pre-AI ways you built defensibility are coming back into vogue. Least confident: maybe I am too pessimistic about whether these companies can keep up their revenue paths. I am skeptical you can call a company at sub-$20 million of revenue the guaranteed winner. In a lot of spaces the A, B, and C all happen within six months, and VCs love the idea of kingmaking. I do not believe you can kingmake that early. But I might be wrong.

Jay: Samit Kalra, thank you for joining me on CLIMB.

Samit: Thank you for having me.

Pull quotes

1. "You can take a sector that was unprofitable to sell software into and turn it into something where you charge 5x more."

2. "The way to think about long-term stickiness is to do hard things outside of the AI."

3. "We assume smart founders will not rest until they find a deep enough problem to solve."

4. "The million-dollar topline benchmark from the past is a seed benchmark now, not a Series A benchmark."

5. "A fundraising process is trading information for engagement. If they are not engaging, what do I have left?"

Source

From CLIMB Episode 080 with Samit Kalra (1984 Ventures). Transcript cleaned from the published episode. Watch the full episode: https://youtu.be/Xw23boFxG58