Software Winners & Losers in the Age of AI

1. Mission-critical software endures. Systems of record that businesses run on — software touching money, regulation, or physical assets — are very unlikely to be ripped out for vibe-coded replacements. Shopify for e-commerce or Veeva for clinical trials aren't going anywhere.
2. Public SaaS faces real duration risk. Markets may stop paying for seven-plus years of cash flows because they become unknowable in an AI-disrupted world, forcing equities to trade as short-duration assets. Per-seat ARR SaaS that isn't AI-native is under serious pressure.
3. Defensibility now comes from data, regulation, physical reality, and workflow. Look for proprietary datasets, regulatory hurdles, robotics/hardware integration, and deeply embedded workflows. These are far harder for a fast follower with AI to replicate than traditional software businesses.
4. English is the new programming language. Portfolio companies are shipping 4-5x more code per engineer using tools like Claude Code, and product UIs are becoming customer-customizable through natural language — fundamentally changing what software looks like.
5. Entry-level white-collar work is disappearing fast. Co-agents are replacing SDRs, CSRs, paralegals, junior engineers, and analysts. One Amplify portfolio company cut from 22 SDRs + 18 CSRs to 1 each while growing revenue 50%.
6. Take money off the table on the way up. Amplify has an explicit internal policy: once a position crosses 10x, they begin selling tranches in secondaries (Series B, C, E) rather than letting emotion drive an all-or-nothing decision.
7. Pre-seed is unusually AI-resistant on the sourcing side. Series A/B funds can use AI to scan all funded companies, but pre-seed deals come from founders who just spun out of SpaceX, Stripe, etc. — invisible to algorithms and discoverable only through long-built networks.
8. QSBS makes pre-seed venture a powerful tax shelter. The 2024 tax bill expanded QSBS — bigger gain exclusion, larger eligible company sizes, and partial credit for sales under five years — making early-stage venture especially attractive for taxable high-net-worth capital.
9. Ask 'why now' and let AI answer it. Great VC bets today are companies that couldn't have been built five years ago and would be too late five years from now — and AI is almost always the catalyst making 'now' the right moment.
10. Revenue growth has decoupled from expense growth. Multiple Amplify and Stage portfolio companies have grown revenue many times over while keeping flat or lower headcount and not raising capital for 4-7 years — a structural shift in software economics.

The State of Public Software in an AI World

Meb opens by asking whether public software companies, which have been taking it on the chin, are facing overdone fears. Rubalcava draws a sharp line between two categories of software. Products that are 'nice to have,' that don't touch money, regulation, or physical assets, are genuinely at risk of being ripped out. But software a business actually runs on is safe — it would be a poor use of scarce engineering time for a company to save a few hundred thousand dollars by re-building what Shopify or Veeva already provides. The test, he says, is to ask: Is this mission-critical? Does high reliability matter? Does it touch money, law, or the physical world? If yes, those companies will be fine.

Bricault adds a market-structure concern: disruption is making things cheaper so fast that public markets may stop paying for cash flows seven-plus years out because they become unknowable. That dynamic showed up acutely after a recent Anthropic announcement caused a diminution in SaaS multiples, as equities started trading like short-duration assets priced on near-term cash. That, taken to its conclusion, kills the logic of growth investing and pushes capital toward short-duration cash flows. For early-stage investors like Amplify, the playbook becomes finding companies that are hard to copy quickly — through proprietary data, regulatory barriers, physical/robotics constraints, or deep workflow embedding that creates its own data flow. A basic, non-AI-native per-seat SaaS company is a hard thing to defend right now.

Inside Amplify LA and the Merger With Alex Rubalcava

Amplify has been around 12 years, co-founded by Bricault and Oded Noy, a former Israeli Air Force fighter pilot and three-time engineer who took his last company public. Bricault, who came from a media/finance background at William Morris (WME), specifically wanted a technical co-founder who could look under the hood. The firm started as a first-check capital source in Southern California and has since expanded across roughly 20 states, narrowing its focus to B2B enterprise with a heavy emphasis on vertical AI and frontier tech.

Rubalcava and Bricault had known each other for over a decade — going back to when Rubalcava was a young analyst at Anthem Ventures, which had invested in Oded's prior company TrueCar. They had co-invested through Stage and Amplify and served together on a portfolio board. With Rubalcava about to raise a new Stage fund and Amplify about to raise its next vehicle, the merger let Amplify acquire Rubalcava's frontier-tech expertise (space, defense, robotics, industrial AI) rather than spend years building it. From Rubalcava's perspective, after seven years running his fund solo, he realized that if he executed his own next four or five years perfectly, he'd end up building exactly what Amplify already was — so why not just join?

Where Defensibility Lives in an AI-Disrupted Software Market

Bricault explains that essentially everything Amplify sees now has some component of AI, though the intensity varies — a CPG company won't be as AI-centric as a healthcare, fintech, or vertical AI agriculture business. Their filter is built around what's hard for a fast follower to replicate.

The Moats They Hunt For

Amplify looks for proprietary data sources, physical data moats in robotics (where real-world data collection is required and not easily synthesizable), integration of complex hardware with AI software, healthcare's regulatory and integration barriers, fintech's compliance regimes and proprietary transactional datasets across specific trade corridors, and anything with deep policy or compliance logic. Rubalcava's space and defense expertise adds another category that is assisted by AI rather than disintermediated by it.

The 'Why Now' Filter

Rubalcava returns to a frame from his prior appearance: great VC investments are companies that could only be built now. Right now, AI is almost always the answer to 'why now.' That shows up in two ways. First, companies can build vastly more product: Balto, a call-center software company, has seen pull requests up 4x and engineering productivity up 5x per team in the 100 days since adopting Claude Code — and they were already using Cursor and other state-of-the-art tools. Second, the product itself is changing. Software used to be fields you filled in to transform data and execute a database transaction. Now you can speak software into existence; English is the programming language. The user interface will be different for nearly every customer because they'll customize their own workflows on top of the system of record.

AI in Action: Stories From the Portfolio

The hosts share several concrete examples that show how AI is showing up inside real companies.

Placer's AI Front End

Placer tracks anonymized foot traffic for tens of millions of people to help retail real estate owners, tenants, brokers, and lenders. Most users only see the heat-map tip of the iceberg; deeper functionality required SQL queries, and the Venn diagram of real estate professionals who can write SQL is tiny. Placer is rolling out a natural-language AI front end. One customer used it during a leasing negotiation when a large publicly traded retailer was walking away from a location — the CEO pulled together a supplementary data presentation in 90 minutes that would have taken his marketing team three weeks, and saved a multi-million-dollar transaction the same day.

Trace and the Birth of Mach1 AI

Trace, a post-Series B machine vision company in Austin, was told by its board to hit profitability before expanding into new verticals. The CEO challenged an employee — a former Amplify employee — to use AI to get there. The employee built co-agents trained on internal company data and reduced headcount from 22 SDRs and 18 CSRs to one each, while the company grew revenue 50% over the same period. Trace hit profitability, the CEO got his expansion mandate, and Amplify then funded the employee to spin out the tech as a standalone company, Mach1 AI, which now helps other companies do the same thing.

Reallocating Top of Funnel

Another Rubalcava portfolio CEO told him conversion rates on outbound SDR work have fallen because everyone's email is being flooded with AI-generated outreach. But that has been offset by AI-enabled automated content generation on LinkedIn and the company blog, which has lifted marketing-driven top-of-funnel conversion. He's reallocating resources from sales to marketing in real time in response.

Two Founder Quotes Worth Saving

Bricault and Rubalcava polled portfolio founders before recording. One wrote: 'Conceptually, we've gone from writing code to writing English to speaking English so the AI can write English for us to program the programs to write our code.' Another said: 'The agentic web is the single most important layer on the internet, and it's only now just coming online, growing 1,000x in the next two years. Current chat AI tools like ChatGPT, etc., will feel like the AOL era of interfaces compared to what we will be using in two years.'

What AI Means for Jobs, Hiring, and Careers

Bricault flags the macro implication that scares him: the disappearance of entry-level work. Anthropic's CEO and others have spoken about this, and the New York Times had just published a story noting roughly a 30% drop in entry-level jobs for college grads. Paralegals, junior law firm associates, entry-level engineers, HR workflow, data analytics, customer service, low-end content/media production, retail analytics, digital marketing — anything based on publicly available data plus basic synthesis — these were the jobs new graduates used to learn the business through, and they are vanishing.

Rubalcava cites a note from the CEO of First Resonance, a manufacturing/engineering software company: when they hire now, they focus on 'AI engineers' and 'forward deployed engineers' — people comfortable with fast feedback loops between customer and implementation who can command AI support teams behind them. The era of learning the business by entering data in a cubicle is gone; you have to be customer-facing from day one. The people who will do best, he argues, are those who are best at listening to customers and understanding their needs.

Meb notes he's told his own team he wants them to level up not only for his firm but in case they leave — to build skills that survive disruption. Bricault references Shopify CEO Tobi Lütke's now-famous mandate that no one at Shopify can hire anyone unless they first establish in writing why AI can't do that job better. Rubalcava ties the thread together with a striking observation: across both legacy Stage and Amplify portfolios, multiple companies haven't raised capital in 4-7 years yet have grown revenue many times over with flat or lower headcount. Revenue growth has decoupled from expense growth in a way that is genuinely new.

Sectors to Avoid and the Adoption Gap

When Meb asks whether any sectors are essentially uninvestable — analogous to ad tech post third-party cookies — Bricault flags low-complexity content creation, technical writing, translation, and basic web development as areas where venture-grade defensibility is essentially impossible. Anything that's just basic data analytics layered on publicly available data — he gives insurance brokerage as an example — is also highly exposed.

Meb references a Ramp study showing that customers in the top quartile of AI spending over the past three years doubled their revenue while the bottom quartile was flat. He notes that a similar survey six months earlier showed almost no correlation between AI spend and revenue growth. Bricault reads that as evidence that companies have shifted from spending on AI to placate Wall Street analysts ('boil the ocean' tools) to spending on AI to solve specific problems with much more focused tooling. Rubalcava extends the parallel: a decade or 15 years ago, the companies that moved earliest from on-prem to SaaS sustained the highest revenue growth. The same phenomenon is happening with AI adoption now — only the transition from SaaS to AI is bigger and harder than the on-prem-to-SaaS shift, so the spread between leaders and laggards will be even wider.

How Pre-Seed Venture Has Changed

Meb references their prior conversation where Rubalcava described being a contrarian sole bidder on deals, and asks whether that's still possible. The answer is: it depends. Rubalcava notes that today there are seed rounds raising $400 million at $2 billion valuations for pre-revenue companies that were incorporated a month earlier — a transaction structure he literally could not have conceived of a decade ago, yet which happens monthly now. Certain founder profiles attract capital with terms that 'beggar the imagination.' But off-map deals — founders without obvious pedigrees, or companies in unsexy industrial domains — still allow patient sole-bidder investing. Rubalcava is finalizing a term sheet on his first Amplify investment, an industrial AI company, three weeks after first meeting the founder, with no apparent competing bidder.

Bricault explains why pre-seed is uniquely defensible against AI-driven sourcing. Series A and B funds like Lightspeed, Accel, Benchmark, and Sequoia can and must use AI to enumerate every funded company in a given sector and run comps. At pre-seed, the founders are people who literally just spun out of SpaceX or Stripe and haven't told anyone — invisible to algorithms. The edge is the 20-40-year networks that Amplify's partners have built. Round sizes have doubled in two years: where pre-seed used to be $1-2 million, it's now typically $2-3 million, with crazy outliers raising $100 million pre-seed for foundational LLM or world model companies based largely on team quality.

Diligence Velocity

Timeframes range from a few days to about eight weeks. Some companies they meet repeatedly while just observing what gets built; others require an immediate decision because a round is closing. By the time they get on a first call, Amplify has typically already done reference checks, market sizing, and competitive analysis, so the call is focused on tech diligence, customer pipeline, and competitive questions. Employment agreements, cap table, and incorporation docs come after the decision.

Recent Investments and Areas of Focus

Amplify Fund 6 has made four investments to date plus the pending industrial AI deal. Roughly 40-50% of investments are typically in Southern California (the last fund was about 45% SoCal), with eight investments in Texas across Dallas, San Antonio, Houston, and Austin, and others scattered across roughly 20 states.

The Fund 6 Portfolio So Far

The four made investments span industrial automation, defense tech, healthcare (including an AI scribe specifically for nurses — addressing a population overlooked by the dozens of doctor-focused scribe products; that company is now past $2M ARR with major hospital network customers), another healthcare deal, and unusually for Amplify, a world model company.

World Models as a New Category

Bricault distinguishes world models from LLMs: where LLMs are trained on text for natural language processing, world models try to learn the underlying structure of spatial environments — the physical rules of spatial awareness — and extrapolate synthetic data to generate fully formed 3D environments. The company landed via a Canadian VC connection and required a fast decision. Amplify normally avoids high-capex deals, but the long-term vision of turning simulation into its own operating system was compelling.

El Segundo, Space, and Defense

Rubalcava notes that as the crow flies between his office in Playa Vista and Meb's home in Manhattan Beach, you fly directly over El Segundo, which has become the spiritual hub of new aerospace and defense entrepreneurship. His existing portfolio includes Slingshot Aerospace (geospatial intelligence focused on what's happening in orbit, not on the ground, addressing rising space-based geopolitical conflict), Epsilon 3 (mission management software for space missions), First Resonance (manufacturing and engineering software), and Gray Matter Robotics (industrial surface treatment robots, based just south of El Segundo). Amplify's focus there will be software serving those industries or hardware with recurring revenue characteristics — not 10-year-to-revenue nuclear fusion bets. As Rubalcava puts it, paraphrasing an El Segundo meme: 'You hang the giant American flag on the wall to show that you're post-traction even if you're still pre-revenue' — and that's probably not their kind of company, even if they're rooting for them.

Pivots and How to Manage Them

Pre-seed investing means what you funded often isn't what the eventual company is. Rubalcava's very first Stage investment, Ventana, started as a hologram company — brand activations at the Super Bowl where you could throw a football to a Joe Montana hologram. They constantly had to convert 3D assets from different content tools into compatible, smaller files, and realized that conversion infrastructure was a better business than the holograms themselves. Ventana is now essentially the 'Adobe Acrobat of 3D assets,' focused on manufacturing digital twins of industrial equipment like tractors.

Bricault's best-ever angel investment pivoted twice — Context Logic started competing with Google in search and ended up as the e-commerce company Wish. He invested at a $7.5 million post-money valuation; Wish later hit $10 billion in the public markets. He took money off before that, which sidestepped the subsequent decline and informed his later philosophy at Amplify. Two Amplify companies they had written down to $1 ended up returning entire prior funds — including a crypto company whose founder called him to say he was going to return the fund, which Bricault initially laughed off as a joke before connecting him to lawyers.

Follow-Ons, Reserves, and the Logic of Pre-Seed Pricing

Both Amplify and the legacy Stage model reserve capital for the seed follow-on but stop deploying their own capital in later rounds. Instead, they offer those opportunities as co-invest to existing LPs via SPVs or direct introductions. Rubalcava's framing: his job is to give investors exposure to a pre-seed portfolio at pre-seed prices. If he reserved capital for Series B-E rounds, his average deployed price would be much higher and he'd be choosing only from companies he'd already found, while a true Series C/D investor should be looking at every Series C/D in the market. He typically defends his initial ownership through the next round, then dilutes alongside founders — which makes his interests as the earliest investor nearly indistinguishable from the founders' for the rest of the ride.

Taking Money Off the Table

Meb raises a point he often makes to public-market investors: people obsess over downside risk but rarely plan for what happens when something works and becomes a huge concentrated position. Rubalcava notes that public-market investors have liquidity at any time; VCs have liquidity only episodically, usually when a company raises a round — and rounds are now spaced 3-4 years apart because companies are so capital-efficient. That makes every secondary opportunity precious: 'I might not have a chance to do this again until 2030.' Closed-end venture funds also impose statutorily defined lives that change the calculus relative to open-end vehicles or personal capital.

Bricault stresses removing emotion through policy. Amplify's internal rule: once a position passes a 10x return, they begin trimming on subsequent secondary opportunities. He gives an example of a Fund 1 company where they took capital off three separate times — Series B, Series C, and Series E — and still own 50% of their original position. He credits Fred Wilson's public writing about doing this with Coinbase and Twitter as the inspiration. Meb notes the broader principle: selling isn't binary. It's not In-N-Out Burger — there's a secret menu where you can sell 10%, 20%, or 50% and minimize regret either way.

LP Base, QSBS, and Asset Location

Rubalcava notes that the 2024 tax bill significantly expanded Qualified Small Business Stock (QSBS) treatment: bigger excludable gains, larger eligible company sizes, and partial credit for sales completed in under five years. Because nearly everything Amplify invests in at pre-seed qualifies (with rare exceptions like fintech lending), an early-stage venture fund is effectively a powerful, open-ended growth-oriented tax shelter for highly taxed individuals and family offices. He ties this to a broader theme he and Meb have discussed: optimizing not just asset allocation but asset location — putting high-friction or high-tax assets inside IRAs, and using QSBS-eligible venture funds for taxable brokerage capital. Venture, he argues, is also more accessible than people think — showing up at Sequoia with $10 million might get your check cashed in 40 years, but plenty of high-quality firms that aren't oversubscribed are reachable.

Bricault describes Amplify's LP base philosophy. Historically they raised heavily from individuals — senior Goldman Sachs executives, senior media executives from his network, public-markets veterans like Gordy Crawford of Capital Group, and senior healthcare and tech executives — chosen both for credibility and for sourcing/portfolio help. He tells founders to set aside cap table for great advisors because one phone call can connect them to their first three customers. More recently Amplify has institutionalized by bringing in funds of funds, both for credibility and for the discipline their diligence imposes — funds of funds typically commit to three vehicles up front, supporting the multi-decade legacy fund he and Rubalcava are trying to build.

Mentors and Lessons From a Career in Venture

Bricault clarifies he started at William Morris (founded in 1898), not as its founder. While there he set up the firm's corporate venture arm as a joint venture with Accel, with Jim Breyer on the advisory board. At Greycroft, where he spent over 12 years starting in 2010, he worked closely with founder Alan Patricof — who previously built Apax (Alan Patricof and Associates) into a $30 billion AUM firm before returning to early-stage venture with Greycroft in 2006. Patricof has Apple's framed angel check on his office wall and tells Steve Jobs stories that double as lessons in selling too early. Bricault's biggest takeaway is Patricof's humility and curiosity: at any industry event, the billionaire 90-year-old would gravitate to the youngest people in the room because that's who he wanted to learn from. Patricof is still investing at 90.

Rubalcava notes that in board meetings, his most useful advice on go-to-market or hiring questions comes directly from mistakes he and portfolio companies have made — scar tissue that has cost real millions. Only time and only mistakes produce that perspective. Meb agrees, drawing the parallel to the ETF industry, where seemingly brilliant launches gather no assets and seemingly dumb ones raise a billion dollars.

The Wildest Pitches

Bricault recalls an MIT graduate pitching a robot EDM DJ that would replace high-paid DJs by reading crowd vibes and programming music in real time — while firing lasers from its eyes for the light show. They passed. He told the story to an Amazon producer friend, and it ended up dramatized as the demo-day-winning company in the first or second episode of the Amazon show Betas, a precursor to Silicon Valley.

Rubalcava recounts being pitched a company literally working on uploading human brains to the cloud. Problem one: cost. He pushed the founders — a million dollars? More. Ten million? More. A hundred million? More. They thought a full upload would cost $10-20 billion, meaning the total addressable market was Elon Musk, Jeff Bezos, and a few dozen other people on Earth. Problem two, in the founders' own words: the process 'consumed the substrate' — i.e., the brain is destroyed during the read operation, so you'd only do it if you were about a week from dying of cancer. He passed despite finding the team extraordinarily technically capable.