Sell the Work, Not the Tool
Sequoia says the next trillion-dollar company will do your job, not sell you a dashboard. I've been accidentally building one since 2017.
Hey.
Few weeks ago, a Sequoia Capital partner named Julien Bek published a short essay that quietly rewired my thinking about everything I’m building.
The title was “Services: The New Software.” The thesis?
For every $1 spent on software, $6 is spent on services. AI is about to eat the $6.
Read that again. Every SaaS company in the world has been fighting over the $1. The subscription. The dashboard. The seat license.
Meanwhile, $6 is sitting right there, untouched, in the form of real people doing real work.
The accountant is closing the books. The broker is shopping for insurance quotes. The recruiter screens resumes. The lawyer drafting NDAs. The SDR team is building your pipeline.
Sequoia’s argument is elegant and brutal: the next trillion-dollar company won’t sell a better tool for these professionals.
It will just do its job.
And every time AI models get better, that company gets faster, cheaper, and harder to kill.
That’s not a prediction. It’s already happening.
And then I realized something that made me put down my coffee and stare at the wall for about twenty minutes.
I’ve been building a $6 company for eight years. I just didn’t have the vocabulary for it until now.
Belkins Confession
Let me be uncomfortably honest with you.
Belkins, the company I built from nothing, is a services business. We don’t sell software. We sell appointments. We sell a pipeline. We sell revenue outcomes.
When a B2B company hires Belkins, they don’t get a dashboard and a login. They get a team. Researchers who find and verify leads. Copywriters who craft sequences. SDRs who book meetings and qualify prospects. Deliverability specialists who make sure those emails actually reach inboxes.
The client shows up to calls. We do everything else.
For years, the tech world looked down on this model. “Services don’t scale.” “Services have low margins.” “You’ll never be a real tech company.”
And then Sequoia publishes an essay saying the next trillion-dollar company will be a software company masquerading as a services firm.
That’s what Belkins has been doing since 2017. We just weren’t pretending.
Think about how perfectly we map to Bek’s framework:
Already outsourced? Yes. Companies have been outsourcing SDR functions for over a decade. The budget line exists. The buyer expectation exists. Nobody needs convincing that lead generation can be done externally.
Selling outcomes? Yes. Clients don’t pay for emails sent or calls made. They pay for appointments booked. For pipeline generated. For revenue produced.
Intelligence-heavy work? Yes. Lead research, ICP definition, email copywriting, sequence optimization, and deliverability monitoring. These are complex but pattern-matchable tasks. They require deep knowledge, but they run on rules.
Reading Bek’s essay felt like someone describing my house from an airplane. Every room was there. The layout was exact. He just saw it from 30,000 feet while I’ve been living in it at ground level.
But here’s the part that didn’t let me sleep.
Bek isn’t describing where companies like Belkins should go. He’s describing where companies like Belkins get eaten, unless we eat first.
Intelligence vs. Judgment
Bek introduces a framework I haven’t stopped thinking about.
Every job in the world is a mix of two things: intelligence and judgment.
Intelligence is rule-based work. Complex, yes. Demanding, yes. But ultimately pattern-matchable. Writing code to a spec. Translating clinical notes into billing codes. Shopping for insurance quotes across carriers. Researching leads that match an ICP. Crafting cold email sequences that avoid spam triggers.
Judgment is the other thing. Instinct is built on years of practice. Deciding which feature to build next. Knowing when to ship before it’s ready. Reading the room in a negotiation. Knowing when a prospect’s objection is real versus performative. Deciding whether to pivot an entire campaign strategy after two weeks of flat results.
AI has crossed the threshold where it can handle most intelligence work autonomously. Software engineering got there first, which is why more coding tasks are now started by agents than by humans.
But it’s coming for every profession. Including mine.
The higher the intelligence ratio in any field, the sooner autopilots will win.
Now here’s what most people overlook about this distinction, and what makes it truly dangerous.
It’s not binary. It’s a sliding scale. And the frontier moves in one direction only: yesterday’s judgment becomes tomorrow’s intelligence.
That nuanced decision about whether to segment your ICP into three groups or five? Once an AI has seen 10,000 campaigns and their outcomes, it’s not judgment anymore. It’s pattern recognition at a scale no human can match.
That “instinct” about which subject line will resonate with enterprise CFOs? After 50,000 A/B tests, instinct becomes statistics. And statistics is intelligence work.
The line between intelligence and judgment isn’t a wall. It’s a tide.
And the tide only goes one way.
Think about it like a glacier. You can’t see it moving on any given day. But come back in five years, and the entire landscape has changed. Every profession, every role, every task that feels like it requires “human intuition” today is slowly being converted into pattern-matchable data. The question isn’t whether. It’s when.
Copilots vs. Autopilots
This is where Bek’s framework gets dangerous. And personal.
A copilot sells the tool. Think of Harvey selling AI to law firms. Rogo selling to investment banks. The professional is still the customer. The AI makes them more productive. They take responsibility for the output.
An autopilot sells the work. Think of a company that drafts your NDAs directly. Or one that closes your books. Or one that gets you insured. The customer isn’t the professional anymore. The customer is the company that needs the outcome.
The difference isn’t semantic. It’s economic.
The copilot captures the tool budget. The autopilot captures the labor budget. And the labor budget is 6x bigger.
Now let me hold up a mirror to my own portfolio. Because this is where the essay stopped being abstract and started being personal.
Folderly, the email deliverability platform I built, is a copilot. It sells the tool. It gives you the dashboard, the spam trigger detection, the DNS analysis, the inbox placement testing. It makes deliverability professionals more productive. It makes marketing teams smarter about their sender reputation.
Belkins is an autopilot. It sells the work. You don’t learn how to do deliverability. You don’t study the dashboard. Your emails just land in inboxes because we handle everything.
I’ve been building both models simultaneously and didn’t even realize they were in philosophical opposition until I read this essay.
Folderly charges per mailbox. Belkins charges per outcome.
Folderly makes you better at your job. Belkins does your job for you.
Folderly captures the tool budget. Belkins captures the labor budget.
And here’s the part that should make every SaaS founder lose sleep:
Every improvement in AI makes Folderly more vulnerable and Belkins more powerful.
When models get smarter, a tool like Folderly risks becoming a feature inside a bigger platform. Gmail’s Gemini is already adding semantic filtering that does some of what Folderly does natively. Every ESP is building AI-powered deliverability into their core product. The tool gets absorbed.
But when models get smarter, an autopilot like Belkins can do the same work faster, cheaper, with fewer humans in the loop. AI doesn’t threaten the outcome. It accelerates it.
That’s not a theoretical distinction. That’s my portfolio talking to me in real time.
The question I’m now asking myself every day: How do we push Folderly from copilot to autopilot? Not “here’s a tool that shows you your deliverability problems.” Instead: “Your emails reach the inbox. Period. We handle it.”
Same destination as Belkins. Different starting point.
If you’re building a tool right now, sit with that question. Because the market is coming for you from both directions, models getting better from below, and autopilots eating your customers from above.
Outsourcing Wedge
Here’s the part that hit me hardest as someone who built a services company from scratch.
Bek’s playbook for autopilot companies is deceptively simple:
Start where outsourcing already exists.
Why? Because if a task is already outsourced, three things are true:
The company already accepts that this work can be done externally
There’s an existing budget line that can be substituted cleanly
The buyer is already purchasing an outcome, not managing a process
Think about what that means.
Replacing an outsourcing contract with an AI-native service is a vendor swap. Replacing headcount is a reorg.
One is a procurement decision. The other is a political crisis.
This is something I understand viscerally from building Belkins. When a company decides to outsource their SDR function to us, that decision has already been made at a philosophical level. They’ve accepted the model. The internal argument is over. All we need to do is prove we’re better than the last agency they tried.
Now imagine an AI-native version of that. Same outcome. Fewer humans. Lower cost. Faster ramp.
The client doesn’t care how the meeting gets booked. They care that it shows up on their calendar with a qualified prospect who showed up informed and ready to buy.
Here’s the analogy that keeps coming back to me. It’s like water finding the path of least resistance. Don’t try to flood the fortress. Find the crack in the wall where water is already leaking through. Outsourced work is that crack. The water is already flowing. AI just turns the leak into a river.
The outsourced task is the wedge. The insourced work is the long-term market.
And the companies that already own the outsourcing relationship? They’re standing right next to the crack. They just need to widen it.
The Opportunity Map That Should Keep You Up at Night
Bek maps out verticals by plotting two axes: intelligence-to-judgment ratio and outsourced-to-insourced ratio.
The sweet spot? High intelligence, already outsourced. That’s where autopilots win first.
Here’s what the landscape looks like, and I’m adding the angle most commentators are missing:
Insurance brokerage ($140-200B): The broker’s value is shopping across carriers and filling forms. Pure intelligence. Massively fragmented, tens of thousands of small brokers each running the same process. What’s overlooked: fragmentation isn’t a weakness here. It’s the vulnerability. Nobody can mount a coordinated defense. There’s no “Big Insurance Brokerage” lobbying to slow this down. The market is an open field.
Accounting and audit ($50-80B outsourced in the US alone): The US has lost roughly 340,000 accountants over five years while demand has grown. 75% of CPAs are nearing retirement. What’s overlooked: this isn’t just an automation opportunity. It’s a demographic inevitability. The humans are literally disappearing. AI doesn’t need to be better than the accountant. It just needs to exist when the accountant doesn’t.
Healthcare revenue cycle ($50-80B): People hear “healthcare” and assume judgment. But the billing layer is nearly pure intelligence. Medical coding means translating clinical notes into ~70,000 standardized codes. What’s overlooked: the complexity is the moat for AI, not against it. No human can hold 70,000 codes in active memory. The AI can. Complexity protects the autopilot, not the incumbent.
B2B sales development and lead generation (growing rapidly, embedded in the $200B+ staffing market): This is my backyard. The top of the funnel, research, targeting, outreach, qualification, is overwhelmingly intelligence work. Pattern-matchable. Scalable. Already massively outsourced. What’s overlooked: the real asset isn’t the AI doing outreach. It’s the data exhaust from thousands of campaigns across dozens of industries. That proprietary dataset is what turns intelligence into judgment over time. And almost nobody is talking about this.
Legal, transactional work ($20-25B): Contract drafting, NDAs, regulatory filings. High intelligence, routinely outsourced. What’s overlooked: quality is verifiable. Unlike creative work, a contract either covers the right clauses or it doesn’t. Verifiability is what makes autopilots trustworthy in legal. You can audit the output. Trust follows auditability.
IT managed services ($100B+): Every SMB outsources its IT. Patching, monitoring, user provisioning, alert triage. Intelligence work on repeat across thousands of identical environments. Nobody has yet sold “your IT runs” directly to the company as a finished outcome instead of selling tools to the MSP.
Supply chain and procurement ($200B+): Most enterprises negotiate seriously with only their top 20% of suppliers. The long tail gets zero attention. Contract leakage runs 2-5% of total spend. What’s overlooked: the wedge is abandoned work. No budget line to justify, no incumbent to displace, just found money. The easiest sale in the world is “here’s money you didn’t know you were losing.”
Management consulting ($300-400B): Huge market but mostly judgment. The interesting question is whether AI can disaggregate consulting into intelligence components (data gathering, benchmarking) and judgment components (strategic recommendations), with the intelligence getting automated and the judgment layer staying human but expensive.
The next wave of AI unicorns won’t look like startups. They’ll look like accounting firms, staffing agencies, and yes, lead generation companies.
The Innovator’s Dilemma (I’m Living It In Real Time)
Here’s the strategic twist that Bek nails. And the one that keeps me honest.
In 2025, the fastest-growing AI companies were copilots. In 2026, many will try to become autopilots. They have the product. They have the customer knowledge.
But they also have the innovator’s dilemma.
Selling the work means cutting their own customers out of doing it.
If you’re Harvey, selling AI tools to law firms, how do you tell your customers, “actually, we’re going to do your job now”? Your best accounts would leave overnight. Your champions become your enemies.
That’s the opening for pure-play autopilots. The ones that start by selling the outcome, not the tool.
I see this from both sides every day. Folderly’s customers are deliverability professionals and marketing teams. If Folderly becomes an autopilot, “your deliverability just works, we handle everything,” it could alienate the exact people who champion the tool internally. That’s the innovator’s dilemma in my own portfolio.
Belkins faces the inverse. We’re already the autopilot. Our challenge is to build enough AI into the delivery so we can serve 10x more clients without 10x more people. That’s not a dilemma. That’s a dream.
This maps perfectly to what I wrote in Plateau: the strategies that got you from 0 to 1 won’t get you from 1 to 10. The copilot that served you well becomes the identity you can’t shed. The service model that built your company becomes the constraint you have to break, or the rocket you have to fuel.
The copilots that can’t make the jump will become features. The autopilots that nail distribution will become empires.
And the services companies that already own the client relationship and add AI to their delivery? They skip the dilemma entirely.
Belkins Home
I’ve been talking about this thesis in the abstract. Let me stop doing that.
We’re building something internally called Belkins Home. And reading Bek’s essay felt like reading our own product roadmap published by someone who’d never seen it.
Belkins Home is our answer to the question every services company should be asking right now: what happens when you take eight years of delivery data, the operational knowledge of 50+ industries, and an AI layer that gets smarter with every campaign, and you fuse them into a single system?
You get the autopilot. Not the kind Silicon Valley builds in a garage with seed funding and a demo. The kind that’s been battle-tested with thousands of real clients paying real money for real outcomes.
Here’s why I believe this will outperform every competitor on the market. And I don’t say that lightly, because I know the graveyard of founders who thought they were special.
🔑 We own the data. And data is the only moat that matters.
This is the part most AI startups can’t fake and can’t buy.
Belkins has run thousands of campaigns across 50+ industries since 2017. We know which subject lines convert for SaaS selling to enterprise healthcare. We know which send cadences work for manufacturing procurement teams. We know the exact sequence length that optimizes for fintech targeting CFOs in DACH markets. We know that the prospect who replied “not interested” to the first touch actually converts at 8% if you wait 6 weeks and change the angle.
That dataset isn’t public. You can’t scrape it. You can’t synthesize it from a foundation model’s training data. You can’t shortcut it with a bigger funding round.
Every campaign we’ve ever run is a row in the training set. Every A/B test, every reply rate, every no-show, every closed deal, every failed experiment. That’s not a feature. That’s a moat.
Bek writes: “As AI systems accumulate proprietary data about what good judgment looks like in their domain, the frontier will shift.” We’ve been accumulating that data since 2017. The frontier shifted for us before the essay was published.
Most AI startups are starting from zero data and trying to sell outcomes. We’re starting from eight years of outcomes and adding AI. The difference is everything.
🔑 We work across segments. That’s the compounding advantage nobody talks about.
This is the overlooked superpower.
A startup building an AI SDR for fintech knows fintech. Maybe they know it well. But when their fintech playbook hits a wall, they have nowhere to go. They’re a rifle. One caliber. One target.
Belkins works across manufacturing, healthcare, SaaS, logistics, financial services, solar, cybersecurity, e-learning, environmental services, and dozens more. That cross-segment exposure means our AI doesn’t just learn one industry’s patterns. It learns the meta-patterns that transfer across industries.
The cold email structure that works for selling enterprise software? It also works for selling environmental compliance services. Different language, different pain points, but the same underlying architecture. The timing pattern that converts for healthcare outreach? It maps to financial services with a two-day offset.
That’s cross-domain intelligence. And it’s exactly what I wrote about in AI Generalist: the most valuable system isn’t the one that knows one thing deeply. It’s the one that sees connections between domains that specialists miss.
Belkins Home inherits that cross-pollination automatically. Every campaign in every industry trains the same underlying model. A breakthrough in targeting methodology for logistics clients improves the targeting engine for healthcare clients. Not because someone copies a playbook manually. Because the system learns.
Single-industry AI startups are building rifles. Belkins Home is building an arsenal.
🔑 We know how to deliver. And delivery is what separates demos from revenue.
This is the part that Silicon Valley consistently underestimates. And it’s the part that separates the autopilots that raise Series A from the autopilots that survive to Series C.
Building an AI that can generate a cold email sequence is not hard in 2026. Any decent engineer with Claude Code can prototype it in a weekend. I know this because I build with Claude Code daily. The model can write the email. The model can even personalize it.
But building a system that actually delivers a qualified appointment to a client’s calendar? Where the prospect shows up informed and ready to talk? That requires knowing a thousand things that aren’t in any codebase:
The client’s sales team can only handle 15 meetings per week before quality drops
That Tuesday at 2 pm converts 23% better than Monday at 9am for this specific ICP
This particular industry has a 3-month buying cycle that accelerates in Q4 because of budget deadlines
The deliverability profile of this domain is degrading and needs intervention before next week’s campaign launch
That the prospect who asked to “circle back in Q3” is actually a dead lead, but the one who said “bad timing” will close in 60 days if you change the approach
That the client’s value proposition is actually wrong, and no amount of outreach optimization will fix a positioning problem
That’s operational knowledge. It’s something between intelligence and judgment. I’d call it a delivery craft. And it takes years to accumulate.
Most AI-first SDR startups build the model and then discover they need the operations. They need the deliverability layer. They need the quality assurance. They need the client management. They need the ability to handle the prospect who goes off-script. They need the team that knows when the data says one thing but experience says another.
They’re building from the AI up. We’re building from the delivery down. Our foundation is solid because we’ve been stress-testing it with real clients paying real money for real outcomes since before GPT-3 existed.
AI startups have the technology. We have the technology AND the delivery infrastructure AND the data AND the cross-segment knowledge. That stack doesn’t exist anywhere else.
🔑 Folderly is the infrastructure layer most people don’t see.
Here’s something I haven’t connected publicly before.
Folderly isn’t just a standalone product sitting next to Belkins on the shelf. Inside the Belkins Home architecture, Folderly is the deliverability nervous system.
Every email Belkins sends runs through Folderly’s AI. Spam trigger detection. DNS health monitoring. Inbox placement testing. Sender reputation tracking. Real-time alerts when something starts to slip.
In 2026, email deliverability isn’t a static setting you configure once. Gmail’s Gemini AI has added semantic filtering on top of traditional spam detection. Inbox placement is now a live, adaptive, AI-vs-AI battle. Every day, the rules change. Every day, the filters learn.
Folderly is our weapon in that fight.
For outside customers, Folderly is a copilot. A tool. A dashboard.
Inside Belkins Home, Folderly is an autopilot subsystem. It doesn’t show you the problem. It fixes the problem. Automatically. Continuously. Before the client even knows there was a problem.
That dual nature, copilot externally, autopilot internally, is actually the transition model I think a lot of tool companies should study. You don’t have to abandon your copilot customers overnight. You build the autopilot layer for your own operations first. Prove it works at scale. Then offer it to the market as a managed service.
Folderly as a tool: “Here’s your deliverability score.” Folderly inside Belkins Home: “Your emails reached the inbox. All of them. You’ll never think about deliverability again.”
🔑 The compounding loop that can’t be replicated.
This is the final piece. And it’s the one that makes me genuinely believe we’re building something that compounds in a way no pure-play AI startup can match.
Bek writes: “Today’s judgment will become tomorrow’s intelligence. As AI systems accumulate proprietary data about what good judgment looks like in their domain, the frontier will shift.”
Belkins Home creates a compounding loop that feeds on itself:
More campaigns → more data → smarter AI → better outcomes → happier clients → more referrals → more campaigns → more data.
Every client we onboard makes the system better for the next client. Every industry we enter deepens the cross-segment pattern library. Every deliverability challenge Folderly solves trains the model for the next one. Every failed experiment teaches the system what not to do, which is often more valuable than the successes.
Software people call this a flywheel. But in the $6 Economy, it’s something more powerful. Because the output isn’t a better product experience. It’s a better business outcome. And outcomes compound differently than features.
The AI-first SDR startup starts the flywheel from zero. They have the model but not the data. They have the pitch deck but not the delivery infrastructure. They have the demo but not the proof.
We start the flywheel from eight years of accumulated momentum.
The race isn’t about who builds the best AI. It’s about who has the best data to feed it. And in B2B sales development, nobody has a dataset like ours. Nobody.
Part Everyone’s Missing
Every newsletter, every VC, every analyst is reading Bek’s essay and asking: “What autopilot should I build?”
Good question. Wrong framing.
The real question is: what happens to the $6 economy when the work itself becomes nearly free to deliver?
Because if AI can close books for a fraction of what an accountant charges, and draft NDAs for a fraction of what a lawyer costs, and book sales appointments for a fraction of what an SDR team bills, then the entire pricing architecture of professional services collapses.
Not slowly. Not gracefully. Like the floor giving out.
The $6 doesn’t just get captured by autopilots. It gets compressed. Clients won’t pay $120K for what an AI can do for $12K. The autopilot captures the work, but the work itself becomes cheaper.
Which means the real margin isn’t in doing the intelligence work. It’s in owning the judgment layer on top.
The autopilot is the entry point. The judgment monopoly is the moat.
Think about it like this. An AI can book your sales meeting. But who decides whether your ICP is right? Does your value proposition resonate in this market? Should you pivot to a different segment entirely? Whether the problem isn’t lead generation at all but product-market fit?
That’s judgment. And that judgment is worth more than a thousand booked appointments.
This is where the Belkins Home vision gets really ambitious. Not “we book appointments.” Instead: “We own your top-of-funnel revenue engine, from strategy through pipeline to closed deal.”
The appointment is the wedge. The judgment layer is the empire.
The companies that win in the $6 Economy won’t just automate the work. They’ll own the decision that precedes the work. And the ones who’ve been doing the work for eight years? They already know what those decisions look like.
What This Actually Means for You
Let me bring this down from Sequoia altitude to where you and I actually operate.
If you run a services business (agency, consultancy, outsourced operations):
You are sitting on exactly what Bek says autopilots need. Domain expertise. Existing client relationships. A deep, intuitive understanding of what “good” looks like in your vertical. That’s the raw material. The question is whether you’ll build the autopilot or get replaced by one. You have 18-24 months before someone credible enters your space with an AI-native version of your offering.
Audit every repeatable process in your delivery. Which ones are pure intelligence? Those are your automation candidates. At Belkins, lead research and initial email generation were the first to shift.
Start treating your historical data as a strategic asset. Every project, every campaign, every outcome, every failure is training data for the autopilot you should be building. If you’re not logging this systematically, start today.
Build the autopilot as a product inside your service. Don’t announce it. Don’t rebrand. Just start replacing intelligence work with AI in your delivery pipeline and see what happens to your margins and speed.
If you’re a founder building with AI:
Stop asking “what tool can I build?” Start asking “what work can I eliminate?”
Pick a vertical where outsourcing is already mature. That’s your distribution advantage. The budget exists, the buyer is trained, the substitution is frictionless. Bek calls this the “outsourcing wedge.” Use it.
Sell the outcome from day one. Don’t start as a copilot hoping to become an autopilot later. The innovator’s dilemma is real, and it protects you from incumbents who can’t make the switch. But know this: you’re racing against companies like mine who already have the delivery infrastructure and the data.
Your real competitive advantage isn’t the AI model. The models are commoditizing fast. Your advantage is the proprietary data that accumulates as you do the work. Every contract you draft, every claim you process, every campaign you run is a row in the training set nobody else has.
If you run a SaaS or tool company (you might be Folderly):
You are the copilot. And the autopilots are coming for your market from a completely different angle.
Ask the hard question: Can your tool become the service? Can you go from “here’s a dashboard that shows you the problem” to “the problem is solved, here’s your monthly report”?
Watch for the signal: When your customers start asking “can you just do it for me?” that’s not a support ticket. That’s a market signal. They’re begging you to become the autopilot.
Build the managed service layer before someone else wraps your tool inside one and sells the outcome at 6x your subscription price.
If you’re a professional or employee:
This is the uncomfortable section. If your daily work is primarily intelligence, meaning pattern-matching, rule-following, data-processing, your timeline just shortened.
Map your own intelligence-to-judgment ratio. How much of your week is rules versus instinct? Be brutal with yourself. If the honest answer is 70%+ intelligence, you’re in the zone that autopilots will reach first.
Migrate toward the judgment side. Negotiation, strategy, relationship-building, creative problem-solving under ambiguity. These are the skills that compound rather than depreciate. The AI gets better at intelligence every quarter. It barely improves at judgment.
Become the human in the loop, not the human being looped out. The autopilot still needs someone making judgment calls at the critical nodes. Position yourself at those nodes.
Back to the $6
I keep thinking about that number.
For every dollar spent on software, six dollars spent on services.
For decades, the entire tech industry has been obsessing over the $1. Building prettier dashboards. Adding features. Fighting for seat licenses. Celebrating MRR milestones.
The $6 was always right there. Hiding in plain sight. In the accountant’s invoice. In the law firm’s retainer. In the recruiter’s placement fee. In the agency’s monthly retainer.
In Belkins’s invoices.
AI didn’t create the $6. It made the $6 addressable.
And the builders who see it, really see it, aren’t going to waste their time making better tools.
They’re going to do the work.
The question I’m sitting with right now is whether Belkins Home will be the system that defines the autopilot era in B2B sales development. Or whether some well-funded startup will read this same Sequoia essay, raise $50M, build the autopilot I should have built two years ago, and eat us for lunch.
I know which outcome I’m choosing. And now you know what we’re building.
The race is on. And for the first time in eight years, I have the vocabulary for what I’ve been building all along.
💎 Post-Credit Scene
This one’s dense, so here’s some lighter fuel to keep the engine running.
📖 Book: Prediction Machines by Ajay Agrawal, Joshua Gans, and Avi Goldfarb. If the intelligence vs. judgment framework clicked for you, this is the deeper dive. The core argument: AI is a prediction technology, and when predictions get cheap, the value of human judgment goes up. Written a few years ago but reads like it was written for this exact Sequoia essay.
🎙️ Podcast: Julien Bek on TBPN (March 2026). The interview where Bek unpacks the essay live. The money quote: “Instead of having 10 humans and one AI, you have one human and 10 AIs, and that ratio just shifts as models get better.” 20 minutes. Zero filler.
🎙️ Podcast: AI Agents Podcast, David Wong / Thomson Reuters episode. How AI agents are already reshaping legal, tax, audit, and professional services from inside one of the largest professional services companies on Earth. Pairs perfectly with Bek’s outsourcing wedge theory.
📝 Essay: Han Heloir Yan, “Building Them Is the Hard Part” (Medium, March 2026). The best counter-argument to the Sequoia thesis I’ve found. Takes the same framework to ground level and asks the uncomfortable questions about what it actually takes to build these systems in production. Also points out, correctly, that many companies Bek cites are Sequoia portfolio investments. Always follow the money.
Thanks for reading.
Vlad
it's a great edition, honestly my top-5 for sure