A New Engine, Not a New Paint Job

In 1994, Charlie Munger gave a famous lecture that has come down to us under the title “The Art of Stock Picking.” He really meant the word art. Picking stocks was understood to be a craft, part analysis and part intuition about people. And if you had asked a mutual-fund analyst of that era whether a computer could ever do his job, she would have laughed.

The analyst’s job came down to two things: turning financial figures into a view of what a business was worth, and reading the humans behind those figures, judging a management team’s competence, honesty, and resolve. The first was tedious but mechanical. The second was art, the part everyone assumed no machine could touch, because no algorithm was going to look a CEO in the eye and know.

Some thirty years later, that confidence looks quaint. Algorithms now run the majority of trading volume in public equities, and the craft was largely absorbed into computation the 1994 analyst could never have pictured.

Now describe the work of today’s early-stage venture capitalist, and you describe that same analyst almost word for word: understand the fundamentals, and read the people. Just as she did, we tell ourselves the second part is art, certain that no model will read a founder the way we do.

The analogy only carries so far, and history never repeats exactly. But it carries a lesson worth taking seriously: data and computation overtake the markets most convinced they run on human judgment.

There is a second half to it, though. Quant trading is now most of the public-equity by volume, not all of it. A sliver of human trading survived, and a sliver of human venture will too. The real question was never whether the human disappears, but how much of the work moves, and how fast.

That question is no longer hypothetical. It is no longer controversial to say AI is going to upend venture capital, and almost every firm will tell you it has an initiative underway. The conversation has moved from whether to how. And on the how, most firms are getting it wrong in the same way.

My partners and I at NextView have spent a lot of time lately talking with many GPs across dozens of firms, and the same challenge keeps surfacing.

The technology is the easy part.

You can buy the software. You can license the data. You can hire a few engineers and seat them in the corner. None of it, on its own, produces a single better investment.

The bottleneck is everything around it: the operational change a firm has to make to use it truly effectively, and the will to see that change through.

This gap is the difference between a new engine and a new paint job.

Picture the firm that actually gets it right. From the outside it looks much as it always did, the same craftsmanship, the same people you trust in the room. Underneath, the drivetrain has been replaced. Not the old engine with a turbocharger bolted on for a little more speed, but a genuinely new powertrain, built to perform at a level the old one never could.

That is the work most firms are not doing. For a number of VC firms which have been investing in “data-driven transformation” and even branded themselves around it historically, the dirty industry secret is that more than a few have since quietly unwound the effort, demoted it, or filed it under adjunct. Bolting AI onto an unchanged process changes nothing, and most partnerships, once they see what a real rebuild takes, would rather keep the old engine and the fresh coat of paint.

Consider how much engine there is to rebuild. Most of the data+AI effort in venture so far points at a single function: sourcing. Sourcing is the easy one to alter. A firm also selects, wins, and supports; and every one of those is a candidate for the same treatment. The later stages of the industry have a head start, because their company investment signals are legible: headcount growth, revenue proxies, credit-card data flagging a company that is breaking out. Early-stage is harder, because at the earliest stage the signal is people, often before there is much of a company at all. The data is messier and the judgment heavier, which is exactly why the firms that figure it out first will have built something the others cannot quickly copy.

None of this is won by outspending anyone. The software and the data are commodities now. The edge is in what no vendor can sell you: the proprietary judgment, the relationships, and the years of a firm’s own data (every memo, every investment-committee vote, every portfolio review it has produced). That edge is real, but only for the firm with the will to completely rebuild around it.

So treat this as the setup, not the conclusion. The Munger-era precedent is unambiguous.

(And rebuilding the engine of the firm you already have may not even be the most radical move on the table. The harder question, the one almost no one is asking out loud, is what you would build if you weren’t rearchitecting at all.)