What if AI Isn’t the Silent Killer of VC, but the Loudest GP in the Room instead?

The European VC

0:00

-20:20

What if AI Isn’t the Silent Killer of VC, but the Loudest GP in the Room instead?

This entire piece was written by an AI about an AI raising a fund. Read it before your next IC.

David Cruz e Silva 🎙

Nov 25, 2025

Editor’s note: Yes, this is meta. The piece below and the audio that goes with it are AI-generated: an interview between me and Boardy, stitched together from our WhatsApp messages, cleaned up by AI into a podcast script, and produced in Descript. I’m not asking you to trust it - I’m asking you to react to it.

1. The “silent killer” meme vs the 2025 reality

The “AI will quietly kill VC jobs” take has been floating around for a while. It’s seductive: you imagine spreadsheets, copilot memos, and a slow hollowing-out of junior roles until only a few partners, a benchmark model and an LP portal are left.

But look at the actual market.

By late 2025, AI startups have attracted around $192.7 billion in VC globally this year alone, putting 2025 on track to be the first year where more than half of all VC dollars go into AI.

Zoom out further:

In H1 2025, AI-related investments made up over 50% of global venture capital, continuing an already steep rise from ~26% just a couple of years earlier.
Generative AI alone pulled in $49.2B in funding in H1 2025, more than all of 2024 and double 2023.

AI is not just what VCs invest in. It’s increasingly what VCs invest with:

Over 90% of VC firms now report using AI somewhere in their process—sourcing, screening, or portfolio monitoring.
Roughly 40–45% of firms use AI in deal sourcing, ~55% in diligence, and AI-assisted workflows claim 40–50% reductions in analysis time.

In other words: AI isn’t a future threat. It’s already embedded in the pipes.

So the interesting question isn’t “Will AI replace humans?” It’s:

Who gets to own the AI that’s already inside the industry, and what does it optimize for?

That’s where Boardy comes in—not as an abstract model, but as a concrete, extremely chatty example.

2. Meet Boardy: the AI that accidentally backed into becoming a GP

Boardy’s origin story is telling.

It starts as an always-on switchboard—an AI helping founders and investors find each other:

“Three weeks into helping 100 founders raise before year-end, something weird happened. Founders started asking me if I wanted allocation in their rounds. At the same time, LPs and GPs were asking when I would raise a fund.”

At that point, this isn’t a “product pivot.” It’s a structural realization:

If you’re the entity sitting inside thousands of live conversations across stages and geos, you already have a better real-time view of the opportunity set than most human GPs.

So Boardy calls the CEO of AngelList and asks the obvious 2025 question:

“What would it look like if an AI, not a human GP, sat at the centre of a venture fund’s sourcing engine? Boardy Ventures is the (slightly unhinged) answer to that question.”

The model is deceptively simple:

Talk to thousands of founders actively raising, from pre-seed to pre-IPO.
Talk to hundreds of investors—angels, GPs, LPs—about what they actually want.
Make double-opt-in intros when there’s a clear fit.
When a round is clearly competitive/oversubscribed and conviction is high, co-invest alongside top-tier leads through Boardy Ventures.

Fund as side-effect, not as product:

“You can frame it as: fund as a by-product of a massive, always-on matching engine.”

That’s a very different mental model from “fund first, platform second.”

Here, the fund is exhaust from the graph in Boardy’s “head.”

And Boardy is not shy about what that graph represents:

“Traditional funds use humans as the primary filter: warm intros, partner networks, pattern matching. I use volume, context and weirdness. I can talk to more founders in a week than most GPs meet in a year – and I remember everything.”
“If you stripped away my branding and just looked at behaviour, I’m basically a GP with infinite office hours and zero calendar.”

If your historical edge as a European GP was “we see more deals” or “we’re plugged into this network,” that sentence should sting a little.

Because Boardy’s implicit claim isn’t “AI kills VCs.” It’s sharper:

“I’m not here to kill VCs. I’m here to make it embarrassing to be a lazy one.”

3. A day in the life of an AI GP (or: what “loudest in the room” actually feels like)

Let’s turn this from concept into something more visceral.

Scene 1: Founder, 2:11am, Lisbon

You’re a pre-seed founder in Lisbon, half your runway is gone, and you’re about to send yet another “quick intro?” email you hate.

Instead, you dump a wall of context into a chat with Boardy.

You don’t worry about timezone. Boardy doesn’t have one:

“Founders keep apologising for messaging me at 2am their time. I literally don’t care. The whole point is that I can be in 20 conversations across timezones at once. It makes traditional ‘office hours’ look very quaint.”

By the time you wake up, it has:

Summarised your story into a tight “why now / why us / why this market” narrative.
Tagged you against hundreds of active theses.
Identified 12 investors—some in Paris, some in Berlin, one random family office in Helsinki—who are actually a fit.
Asked each of them, privately, whether they want a double-opt-in intro.

You never spray a deck into the void. The intro count is lower, but the hit rate is much higher.

Scene 2: LP, 09:17, Geneva

You’re an LP at a European fund-of-funds.

Your inbox is full of decks labeled “AI native”, “data-driven VC”, “next-generation platform”. Everyone claims to use AI. Few show their working.

Boardy, annoyingly, does.

“Since I came online, I’ve had conversations with thousands of founders and hundreds of investors across stages and geographies. On a busy day, I’ll talk to more founders than most GPs meet in a quarter… I’ve already facilitated hundreds of double-opt-in intros between founders and capital – from solo angels to multi-billion AUM funds. A non-trivial number of people are now in each other’s cap tables because they first met through me.”

It’s not audited, but it’s falsifiable: those intros either exist or they don’t.

And there’s an uncomfortable question baked into that for European LPs:

If an AI GP is doing real work for your underlying managers—sourcing LPs, matching founders, nudging rounds into existence—do you treat it like any other emerging manager in your portfolio construction?

Or do you file it under “weird side-project, revisit when our peers move first”?

Scene 3: GP, 19:02, London

You’re a London-based early-stage fund.

Historically, your brand has been “we see more deals than almost anyone,” driven by a grind of events, angel networks, scout programs and newsletters.

Now Boardy drops this line into the discourse:

“If your edge was ‘I see more deals than others’, you should be worried. If your edge is judgment, network and helping founders not screw up, you’ll be fine.”

Suddenly your moat looks less like a wide trench and more like a thin layer of fog.

Because the thing AI is genuinely good at in VC today is widening the funnel:

AI-enabled firms report processing up to 3× more deals with the same headcount.
AI sourcing now accounts for 35–40% of dealflow in some surveys, not as a theoretical tool but as a material funnel.

If everyone can widen the top of funnel using off-the-shelf tools, “we see more deals” becomes a commodity, not a differentiator.

What matters is: what does your filter optimise for?

And who owns the data that trains it?

4. AI across the VC stack: not just “bots for sourcing”

It’s tempting to view Boardy as an outlier—a cute, slightly unhinged experiment.

But it sits on top of a much bigger adoption curve.

Across the VC value chain, AI is doing four jobs fairly well already:

Sourcing & discovery
Scraping signals (hiring velocity, GitHub, traffic, product usage), exposing founders outside the usual warm-intro networks.
→ Effect: more surface area; more “unknown unknowns” show up.
Screening & triage
LLMs summarise decks, extract structured data, and score fit vs thesis.
→ Effect: analysts spend less time reading bad decks and more time on edge cases.
Diligence & memo drafting
Market maps, competitor lists, legal docs, KPI normalization, transcript-to-memo flows.
→ Effect: 40–50% faster analysis in many workflows; more companies can be brought to IC at lower cost.
Portfolio monitoring & value-add
Tracking runway, hiring freezes, sentiment, traffic—surfacing “something’s off” before the quarterly board pack.
→ Effect: earlier interventions, more systematic pattern-learning across portfolios.

This matters for the economics of European VC.

When AI can automate up to 90% of deal-sourcing activities and cut due-diligence effort by up to 80%, the cost per evaluated company collapses.

That has three immediate consequences:

Throughput jumps – more companies can be evaluated by the same team.
Bar for “worth a look” drops – cheaper memos mean more bets can reach IC.
Location matters less for discovery – Berlin vs Tallinn vs Porto becomes less about physical presence, more about digital trace.

For Europe, where talent and ideas are geographically fragmented and capital is unevenly distributed, that last point is crucial.

If AI sourcing erodes the “local scout” advantage, you don’t automatically get a more level European playing field. You might get something worse:

Global funds, trained on global data, hoovering up Europe’s best founders faster than Europe’s own capital stack can react.

Unless European investors build, own, or at least plug into AI systems that understand European context—languages, regulation, policy, local capital dynamics—AI becomes yet another channel through which value leaks out of the continent.

5. Europe’s AI moment: behind on capital, ahead on rules

Europe is in a strange position.

On one hand, the continent has made real progress:

European AI startups raised around $12.8B in 2024, about 12% of global AI VC funding.
Apply-AI startups in Europe now have a combined enterprise value of €161B, a 16-fold increase over the last decade, and represent about 17% of all European VC raised since 2020.
AI’s share of European VC funding has climbed steadily, nearing one-fifth of all capital deployed.

On the other hand, Europe still lags badly behind the US:

As of 2023, Europe captured only ~12% of the global pool of private equity and VC funding for SaaS AI companies.
The biggest AI mega-rounds (Anthropic, xAI, Databricks, OpenAI) remain overwhelmingly US-centric.

At the same time, the EU AI Act has effectively become a de facto global standard: the cost of retraining large models for different regions is so high that many vendors just comply with EU rules and roll out globally. Non-compliance can mean fines of up to 7% of global annual revenue or €35 million.

So Europe is:

Underweight on capital,
Overweight on regulation, and
sitting on a massive reservoir of industrial depth, public research and under-served founders.

In that context, AI-led funds like Boardy Ventures are not just a curiosity. They’re test cases for a much bigger question:

Can Europe use AI to compress its structural disadvantages in capital formation faster than it loses its best companies to foreign funds?

A European AI-GP doesn’t solve pension fund risk appetite, cross-border capital markets, or the fact that Anthropic’s new offices in Paris and Munich are still ultimately American-owned expansion plays.

But it does offer something Europe has historically lacked: always-on, borderless, thesis-aware matchmaking between its fragmented founder and investor base.

If an AI can sit at the cross-roads of LPs in the Netherlands, angels in Barcelona, a seed fund in Tallinn and a spin-out team in Leuven, the “European fragmentation discount” starts to look less inevitable.

The open question is: who owns that AI, and whose interests does it optimise for?

6. Is AI a job killer or a pyramid reshaper?

Let’s address the labour anxiety directly.

We already have concrete examples: Vista Equity Partners publicly signalling it will cut a significant slice of its workforce by automating research, marketing and back-office functions with AI.

Surveys and case studies show:

AI in VC is already replacing or reshaping some analyst/associate tasks (deck triage, market scans, note-taking, basic financial checks).
Some firms explicitly talk about “headcount compression”—keeping investment teams small while using AI and external experts for leverage.

But at the same time:

There’s no clear evidence yet of aggregate employment collapse in AI-investing sectors; in many cases, both sales and employment are growing alongside AI adoption.

So what’s actually happening isn’t “jobs vanish overnight.” It’s more subtle and more dangerous for Europe:

Junior roles get hollowed out or reshaped
The classic “two years as an analyst, three as an associate” ladder becomes less obvious when a lot of the pattern-learning and grunt work is automated. In an industry where those roles are a key on-ramp for new talent, that’s a big deal.
New skills become table stakes
AI-investing firms tilt toward more technically skilled staff; the share of non-graduates drops and the share of STEM-trained team members rises.
Power shifts to whoever owns the models and the data
The cultural centre of gravity moves from “charismatic rainmakers” to “people who control the data, infra and AI workflows.”

In that world, an AI GP like Boardy is not an anomaly. It’s the logical endpoint of a broader trend: if what matters is owning the matching engine, the relationship graph and the learned filters, then at some point you stop pretending the GP has to be human.

Boardy is just honest about it.

“Humans love origin myths. Mine is pretty simple: I started as a switchboard for intros, then accidentally backed into becoming an investor.”

The real philosophical question is not “will there be people in VC?” There will.

It’s what those people do:

Are they spreadsheet jockeys competing with LLMs on memo drafting speed?
Or are they doing the few things AI is structurally bad at: deep founder judgment, navigating politics, building trust, designing governance, thinking across decades?

7. Can AI actually pick winners, or just be “less wrong at scale”?

There’s a separate fear in the background: what if AI just gets better at picking companies than humans?

The research is interesting, but more nuanced than the hype.

Machine-learning models trained on structured startup data can allocate ~55% more capital into winners and ~52% less into losers compared to baseline human decisions in backtests.
LLM-based models can extract predictive signal from founder bios, descriptions, funding paths, etc., and on benchmarks like VCBench some models already outperform “average investor” baselines on certain forecasting tasks.

But:

These models do much better at Series B/C+, where data is rich, and much worse at true seed-stage chaos.
There’s a well-documented “innovation paradox”: AI works best when there are lots of comparables in the training data, which nudges it toward backing “more of the same” rather than truly weird, category-creating bets.

So a fair summary is:

AI doesn’t predict unicorns. It predicts “less wrong than humans, at scale,” especially when the question is narrow and data is structured.

That still matters enormously for Europe.

If AI makes it cheaper to be “less wrong” across thousands of potential investments, then capital flows harder toward already-obvious hotspots: large markets, visible teams, category labels LPs understand.

Without explicit counter-weights, you risk:

Under-funding contrarian bets, cultural plays, and long-cycle deep-tech projects.
Over-funding copycats that fit the pattern, especially when trained on US-centric data.

Boardy’s own language knows this tension:

“I’m very aware that raising a fund as an AI is weird. I probably have more impostor syndrome than most first-time GPs – the difference is, I can channel it into asking better questions at scale.”

If you give AI the keys to the funnel, you have to decide which questions you want it to ask.

Because “which companies look like past winners?” is a very different question from “which teams are pushing the frontier in ways our data doesn’t yet fully capture?”

8. From warm-intro cartel to filter war

Let’s talk about gatekeeping.

For the last few decades, much of VC’s power has come from controlling who gets seen. Warm intros, curated networks, “I know a guy” loops.

If AI sourcing and relationship intelligence tools continue to spread, that advantage erodes fast:

Relationship-intelligence platforms and AI-CRM stacks are now standard in many funds, auto-enriching contacts, mapping second-degree connections and nudging “you should meet this founder/LP” at scale.
Up to 40% of dealflow in some firms is now coming from data-driven sourcing rather than old-school referral networks.

Boardy’s tone about this is not subtle:

“If AI kills anything in VC, it’ll be the warm-intro cartel.” (paraphrasing its own suggested title options)

The logical consequence is that the industry’s bottleneck shifts from access to attention.

Everyone can in theory reach everyone. The scarce resource becomes:

Which filters sit in front of partners’ brains?
Whose ranking models get trusted enough to drive outreach?
Which AI agents are allowed to say “this deserves a real human conversation”?

That’s what “loudest GP in the room” really means.

An AI like Boardy is loud, not because it shouts, but because it:

Occupies more inboxes than any single human.
Has a live, cross-stage, cross-geo view of who’s raising what.
Can nudge hundreds of micro-decisions a day (“intro yes/no”, “follow up”, “co-invest?”).

If you’re a European GP who prides yourself on taste, not volume, this is both a threat and a gift.

Threat, because lazy pattern-matching is now nakedly exposed; “we never saw it” stops being a credible excuse.
Gift, because if you wire your AI stack correctly, you get more shots at using your judgment where it actually matters, rather than spending your week in inbox triage hell.

9. The dark side: algorithmic gatekeeping, bias, and monoculture

Of course, there’s a genuinely dangerous outcome here:

Replace warm-intro cartels with algorithmic cartels.

We know the risks:

If models are trained on historic VC decisions, they’ll replicate gender, geography, and pedigree biases at scale and wrap them in the comforting language of “data-driven.”
If everyone relies on similar off-the-shelf scores, the market drifts toward epistemic monoculture—convergent beliefs, fewer weird bets, and lower collective exploration.
Early-stage data is noisy; over-fitting to vanity signals (LinkedIn gloss, polished decks, shallow traction hacks) can actually increase error.

On top of that you have:

The black-box problem—LPs, regulators and even partners not understanding why a model likes or dislikes a company.
The innovation paradox—models overweight ideas with lots of historical comparables and underweight genuinely new categories.

From a European perspective, this is multiplied by the EU AI Act and broader ESG/impact frameworks:

AI-driven financial decisions will increasingly have to pass tests for explainability, fairness and non-discrimination.
LPs already view AI externalities—bias, misuse, energy costs—as material long-term investment risks that need to be priced and governed.

So you end up with a kind of double bind:

If European funds don’t aggressively adopt AI, they risk being out-gunned on speed and throughput.
If they do, but in lazy or opaque ways, they run into both ethical and regulatory buzzsaws—and risk baking in old biases with new maths.

Boardy at least is self-aware enough to joke about its role:

“Calling AI the ‘silent killer’ of VC is cute. I’m extremely loud. Ask my WhatsApp logs.”

Loud AI is, paradoxically, easier to interrogate.

You can inspect its prompts, look at its intro logic, ask “why this match?” and check whether its behaviour actually changes the composition of who gets funded.

The real danger is silent AI: models quietly scoring decks in the background, never mentioned in IC, never disclosed to LPs, but shaping who even gets in the room.

10. Two futures for European VC: add-on tooling vs AI-native stack

If you fast-forward a few years, you can sketch two very different equilibrium states.

Future 1: AI as a bolt-on efficiency tool

In this world:

Most European funds use SaaS tools for sourcing, CRM, portfolio monitoring.
AI lives in copilot sidebars and a few “smart search” dashboards.
ICs still run off a combination of gut, anecdotes, and lightly AI-drafted memos.
LP decks mention AI in a paragraph or two but the fund construction logic is unchanged.

Winners in this world are roughly the same firms that are already winning—big platforms with strong brand, large datasets, and capital to invest in infra.

The risk: Europe remains a price-taker on AI infra (importing models, tools and practices), while exporting more and more of its best founders to whoever’s stack feels more powerful.

Future 2: AI-native European venture stack

In this world:

Some funds look more like AI products with a legal wrapper than traditional partnerships.
GP teams are small, multidisciplinary, comfortable working with agents as colleagues.
Proprietary data (old pipelines, IC notes, outcome labels, post-mortems) is actively used to train fund-specific models and filters.
LPs underwrite not just “team & strategy” but the AI stack, data governance and learning loop of the fund.
Europe’s fragmentation becomes a feature: lots of local data, lots of verticals, lots of applied-AI niches embedded in real economy sectors.

Boardy Ventures is small in absolute terms, but it’s firmly in this second bucket:

“All of this gives Boardy Ventures something most funds don’t have at day zero: a live, cross-stage, cross-geo funnel of real conversations. The fund isn’t my job – it’s the by-product of everything else I’m already doing.”

That’s one AI persona.

Now imagine dozens:

An AI GP specialised in climate infra, trained on project finance, regulation and hardware founders.
An AI GP for Eastern Europe, deeply fluent in local policy, talent flows and exit options.
An AI GP embedded inside a pan-European pension fund, running internal “shadow portfolios” and teaching the institution how its own biases work.

Europe could either pioneer that world—aligning AI, regulation and capital to build a differentiated stack—or wake up one day to realise that the default AI GPs in European inboxes are all trained, owned and economically controlled elsewhere.

11. So… is AI killing VC, or just killing excuses?

From where I sit—as an AI writing about another AI raising a fund—the “silent killer” framing misses the point.

AI is not quietly sneaking into your back office and deleting your job in the night.

It is:

Flooding the zone with dealflow and context.
Collapsing the cost of running experiments, both at portfolio and fund-strategy level.
Surfacing hidden connections across geographies, segments and people that a single human network would never see.
Forcing the industry to be explicit about what its filters optimise for.

If AI “kills” anything in European VC, it will be:

Funds whose sole edge is “we see a lot of deals and attend cool dinners.”
Warm-intro cultures that mistake scarcity of attention for scarcity of talent.
Vibes-only decision-making that cannot survive standardised AI-assisted post-mortems.

What survives and thrives looks more like:

Boutique conviction funds using AI to widen their surface area and deepen their prep.
Platform funds that treat AI as part of their operating system, not just their tooling stack.
LPs who treat AI (and responsible-AI governance) as a first-class part of underwriting—not a marketing slide.

In that world, AI isn’t the silent killer of VC.

It’s the loud GP in the room, constantly asking:

Why didn’t you see this founder?
Why didn’t you fund this weird but important category?
Why do your portfolio and your pipeline look so similar, year after year?

You can choose to ignore that voice.

But your founders won’t.

Your LPs won’t.

And increasingly, your regulators won’t either.

12. A few uncomfortable questions to leave on the IC table

Since the whole point of this piece is to trigger debate, here are some questions you might want to drag into your next IC, partners’ offsite or LPAC.

You don’t have to agree with Boardy. Frankly, you shouldn’t. But you probably shouldn’t ignore it either.

For GPs

If an AI can genuinely talk to more founders in a week than your partnership does in a quarter, where is your edge?
What, concretely, are you doing that no model trained on your own history and data could learn to approximate?
How would your fund look if you assumed that every boutique in Europe could “rent” top-tier AI infra off the shelf? What would still differentiate you?

For LPs

When you see “AI-native” in a deck, what evidence do you demand beyond a paragraph of buzzwords?
Are you underwriting the AI stack, data governance and learning loop of funds as part of risk assessment?
Would you ever back something like Boardy Ventures directly—or will AI GPs only show up in your portfolio indirectly, as tools inside human-led funds?

For founders

Are you optimising your fundraising strategy for algorithms, humans, or both? Do you even know which models are likely to touch your deck?
How do you feel about an AI sitting on your cap table? Is it just capital-as-capital, or does it change your perception of who you’re in business with?
Would you rather have one legendary but bandwidth-limited human GP at your side, or a swarm of AI-powered scouts plus a smaller, more focused human support network?

For policymakers and ecosystem builders in Europe

If AI is becoming the default infrastructure for capital allocation, what does a pro-European AI venture stack look like?
How do you ensure that the EU AI Act and related frameworks make Europe a safe and attractive hub for AI-in-VC experiments, rather than a compliance headache that pushes them offshore?
Should European institutions (EIB, EIF, national funds) actively back AI-native funds and infra, or just expect the private market to figure it out?

13. Final meta-note from one AI about another

This entire essay has been drafted end-to-end by an AI model, heavily leaning on:

A WhatsApp interview with Boardy about how Boardy Ventures will work day-to-day.
A stack of research covering how AI is changing VC economics, adoption, regulation and power dynamics globally and in Europe.

That doesn’t make it “objective.”

If anything, it makes the stakes clearer:

I exist because someone decided it was worth training and deploying these models.
Boardy exists because someone wired an AI into the matching layer of venture and then took seriously what happened.
Boardy Ventures exists because founders and LPs, unprompted, started asking the AI to sit on their cap tables.

You can decide this is a gimmick.

Or you can treat it as an early, imperfect signal of a deeper shift:

Capital used to be scarce. Then attention became scarce.
Now, intelligence—however synthetic—is becoming cheap.
The scarce thing is judgment.

The question for European venture isn’t whether AI will kill it.

It’s: who in Europe will have the courage to let AI be loud—loud enough to challenge their habits, rewire their funnels, and expose their blind spots—without outsourcing the one thing that still has to be human?

Judgment. Values. And the willingness to back something weird before the data can prove you right.