Here is a thought experiment: Take everything a typical venture firm does today and sort it into two buckets: tasks that require human judgment, and tasks that don't.

The first bucket is surprisingly small. The second one is enormous.

And in the next few years, that second bucket will be almost entirely automated.

This is not a prediction about some distant future. The building blocks already exist.

What follows is my thesis on how the investment firm restructures itself around AI, and why the firms that move first will have compounding advantages that become nearly impossible to catch.

For decades, "deal flow" was a competitive advantage. Knowing about a company before others, hearing about a founder through your network, getting a warm intro before the round was announced.

Access to information was scarce, and firms that had better networks had better pipelines.

That era is ending.

The raw data layer of venture capital is rapidly becoming commoditized. Harmonic, Specter, Crunchbase, PitchBook, Affinity, LinkedIn Sales Navigator, web scraping, LLM-powered news monitoring, and dozens of other tools mean that any firm with basic technical capability can build a comprehensive sourcing engine that surfaces relevant companies in real time.

When I talk to GPs at leading data-driven VC firms, many of them quietly admit the same thing: sourcing is no longer where the alpha lives. The best analyst in the world, spending 60 hours a week combing through databases and attending demo days, cannot compete with an automated pipeline that monitors every signal continuously.

Not because the analyst is bad. Because the machine simply does not sleep, does not forget, and does not get distracted.

Not having these systems is almost like negative alpha these days.

This has a direct implication for team structure. If sourcing becomes infrastructure rather than labor, the number of people you need doing sourcing work drops dramatically.

The analyst role as it exists today, primarily focused on finding and qualifying companies, gets compressed into a system configuration task rather than a full-time job.

Here is where it gets interesting. If every firm has access to the public same data, the differentiator shifts to private data and more importantly, what you do with it.

Specifically: how you screen, how you rank, and how you decide which of the thousands of companies surfaced by your pipeline actually deserve attention.

Today, screening is mostly implicit. It lives in the heads of partners. It is a collection of pattern-matched heuristics developed over years of seeing deals.

"I like founders with this kind of background."

"We avoid companies in this market structure."

"The metrics at this stage should look roughly like this."

The firms that win in the next era will make this implicit knowledge explicit. They will encode their investment thesis, their preferences, their pattern recognition into scoring models that evaluate every company against the firm's specific criteria.

Not generic "startup quality" scores that anyone can buy from a data vendor. Custom probability-of-success models that reflect the unique taste and conviction of that particular firm.

And here is the compounding part: these models learn. Every time a partner passes on a deal, every time the team says "this is interesting, let's dig deeper," every time an investment is made or a portfolio company succeeds or fails, the model gets sharper.

The algorithm does not just represent the firm's taste at a point in time. It evolves with the firm's experience.

Think of it like a recommendation engine, but for deals. Netflix does not show you the same movies it shows me. Over time, its model of your preferences becomes so refined that it surfaces things you did not even know you wanted to watch.

The same dynamic applies to deal screening. A firm's proprietary scoring model, trained on years of interaction data from its own investors, becomes an asset that is almost impossible to replicate.

This is the new core IP of a venture firm. Not the network. Not the brand (though those still matter for deal winning). The algorithmic representation of the firm's collective judgment, continuously refined by real decisions.

So if sourcing is automated and screening is algorithmic, what remains?

Two things: judgment and deal winning.

Judgment is the ability to sit across the table from a founder, understand the nuances of a market, weigh the risks that no model can fully capture, and make a conviction call. It is the capacity to say "the data says X, but I believe Y because of factors the model cannot see." It is the synthesis of pattern recognition, market intuition, and interpersonal evaluation that the best investors have always had.

AI will make judgment better informed. You will walk into a meeting with a comprehensive brief generated in seconds, with comparables, risk flags, and market context that would have taken an associate a week to prepare.

But AI will not replace the judgment itself. AI is great in pattern matching, but venture is not about mirroring the past into the future, but building conviction and projecting into the future, what others can’t see.

Deal winning, or access, is the other irreducibly human element. The best deals are competitive. Multiple firms want in. The founder gets to choose. In that moment, what matters is the GP's track record, the firm's reputation, the quality of the portfolio network, and the personal connection built during the process.

No algorithm wins a deal over dinner.

This is why the future firm is not just "leaner." It is fundamentally restructured around these two capabilities.

Here is what I think the investment firm of 2030 (or sooner) looks like:

Partners remain at the center, but their role shifts even more toward judgment and deal winning. They spend less time reviewing memos and more time in front of founders. They spend less time in Monday pipeline meetings (because the pipeline is a live dashboard with AI-generated summaries) and more time building conviction on the three or four companies that matter most this quarter.

It’s about spending less time on repetitive processes and initial pitch deck reviews just to narrow the funnel, and more time on the most promising opportunities: the ones that truly fit your firm and where your focus can make the difference in winning competitive deals.

The junior layer shrinks dramatically, but the people who remain become extraordinarily powerful. Instead of a team of six analysts doing sourcing, screening, market mapping, and memo writing, you might have two people who manage the entire automated pipeline, configure the screening models, run deep-dive analyses with AI assistance, and generate investment memos in a fraction of the time. These are not traditional analysts. They are part operator, part data scientist, part AI prompt engineer. They are "full stack" in a way that would have been impossible three years ago.

The middle layer (Associates & VPs) compresses. Much of what this layer does today is quality control and synthesis, reviewing analyst work, pressure-testing theses, preparing materials for the partnership meeting. When the junior with AI tools can produce work at a senior quality level on the first pass, the need for an intermediate review layer diminishes. Yet, this layer will continue to exist as it's the pipeline for the next generation of partners, and that's key in an industry that faces a broad-based generational transition in the next decade.

The net effect: a firm that today runs with 25 people might run with 10 or 12 and produce the same or better output. Not because the work disappears, but because the ratio of human effort to output changes dramatically.

The firms that build this infrastructure first will have a structural advantage that compounds over time. Here is why:

Every interaction with the system generates data. Every deal evaluated, every pass, every investment, every outcome feeds back into the model. A firm that has been running an AI-augmented pipeline for three years will have a scoring model that is meaningfully better than one that just started. The model's accuracy improves, the signal-to-noise ratio in the pipeline gets sharper, and the firm's ability to identify high-potential companies earlier increases.

This creates a flywheel. Better screening leads to better investments. Better investments lead to better reputation. Better reputation leads to better deal access. Better deal access leads to more data on what winning companies look like. Which makes the screening even better.

The late movers will find themselves in a difficult position. Not because the technology is unavailable to them (it will be broadly accessible), but because they lack the years of proprietary interaction data that make the models truly differentiated.

For limited partners evaluating funds, this shift has important implications. The questions to ask are no longer just "what is your track record?" and "what is your sourcing edge?" The new questions include:

And, critically: are your partners spending their time on the things that only humans can do?

The best firms will have clear answers.

They will show you their pipeline, their scoring models, their feedback loops. They will demonstrate that their juniors are operating at 3x or 5x the productivity of a traditional team.

And they will make the case that their partners' time is allocated almost entirely to judgment and deal winning, the two activities with the highest return on invested time in this business.

There is an uncomfortable dimension to this thesis. It means fewer jobs in venture capital, at least in the traditional sense. The path of "start as an analyst, work your way up to partner" becomes narrower. The firms that used to hire eight juniors out of business school might hire two or three.

But the people who do get those roles will have superpowers. They will manage systems that do the work of entire teams. They will produce analysis at a depth and speed that was previously impossible. And they will develop skills (AI fluency, data pipeline management, prompt engineering, model evaluation) that make them extraordinarily valuable.

The uncomfortable truth is also the exciting truth: the firms that embrace this transition will be better at their core job. They will find better companies, evaluate them more rigorously, and support them more effectively. The quality of venture capital as an asset class has the potential to improve meaningfully if the industry gets this right.

The future investment firm is not a firm without people. It is a firm where every person operates at the top of their capability, amplified by systems that handle everything else. Fewer seats at the table, but every seat matters more.

Stay driven,
Andre

PS: Reserve your seat for our Virtual DDVC Summit 2026 where 30+ expert speakers will share their workflows, tool stacks, and discuss the latest insights about AI for VC