The Next Generation of Startups Will Be Built Around Intelligent Decision Engines
The most valuable software of the next decade will not just show people information. It will tell them, and eventually do for them, what to do next. Here is why we think that changes everything about how startups should be built.
For most of the history of business software, the core promise was visibility. Give a person better information, presented more clearly, and they will make better decisions. Dashboards, reports, and analytics tools are all built on this premise, and they have created enormous value over the past two decades. But we think this era of software is reaching its natural limit, not because visibility stopped mattering, but because the volume and complexity of data most businesses now generate has outpaced what a human looking at a dashboard can reasonably act on in a timely way.
The next generation of valuable startups, across nearly every industry we look at, is being built around a different premise: not just showing information, but continuously recommending, and increasingly executing, the next best decision. We call this a decision engine, and we think it represents a genuinely different category of software from what came before it, even though it often looks similar on the surface, sitting inside a familiar dashboard or application.
In retail, this looks like a system that does not just report that inventory is running low, but automatically recommends the optimal reorder quantity based on demand forecasting, supplier lead times, and cash flow constraints, and in more mature implementations, places the order automatically within limits a human has approved in advance. The value here is not the reporting. Reporting has existed for decades. The value is the system doing the analytical work that would otherwise require a skilled operator to do manually, and doing it continuously rather than in a periodic review.
In logistics, decision engines are showing up as systems that do not just track where a shipment is, but continuously recalculate the optimal route as conditions change, weather, traffic, fuel prices, delivery windows, and automatically adjust plans without waiting for a dispatcher to notice a problem and intervene manually. The best logistics companies we have looked at are not the ones with the most detailed tracking dashboards. They are the ones whose systems are quietly making thousands of small routing and scheduling decisions every day that a human team simply could not make at that speed or that scale.
Healthcare is moving in the same direction, though more cautiously and appropriately so, given the stakes involved. Decision engines here look like systems that do not just present a physician with a patient's full history and test results, but actively flag which combination of factors suggests a specific risk or a specific next step, drawing on patterns across far more data than any individual physician could hold in their head during a short consultation. The physician remains firmly in control of the final decision, as they should, but the system is doing real analytical work that meaningfully improves the quality of the decision being made.
Finance got here earliest. Algorithmic trading has made autonomous decisions within defined parameters for years, and that same logic is now reaching consumer and small business finance: systems that do not just show a business its cash flow, but recommend, and sometimes automatically execute, moving money between accounts to avoid a fee, or adjusting a credit line based on real-time changes in financial health rather than a static, infrequently updated score.
Real estate and trade, two sectors that have historically lagged in technology adoption, are seeing the same pattern. In real estate, decision engines recommend listing prices from live market conditions instead of a comparable analysis done once and left stale for weeks. In trade, they recommend which supplier to use for a specific order based on continuously updated pricing, reliability, and shipping data, rather than a once-a-year vendor list that is often outdated by the time anyone uses it.
What ties these examples together is a common architecture. Each needs a continuous, reliable stream of real-time data. Each needs a model sophisticated enough to weigh multiple factors the way a skilled expert would, but faster and more consistently across a large volume of decisions. And each needs a clearly defined boundary for when the system can act on its own versus when a human needs to approve a recommendation first. Getting that boundary right is often the single most important design decision a founder in this space will make.
For founders, this is a real opportunity, but it raises the bar for what a compelling company needs to show. It is not enough to have an interesting model. A founder needs reliable access to the right data, a specific understanding of what decision is currently being made poorly or too slowly, and a thoughtful answer for where human oversight belongs. Founders who can answer all three with specifics, not generalities, are the ones we get genuinely excited about, because they are describing systems that will still matter five years from now, long after the current wave of AI hype has moved on.
This is also, fundamentally, why we exist as an investor. We do not believe the future belongs to companies that make information look nicer on a screen. We believe it belongs to companies that make better decisions on behalf of the people and businesses depending on them, continuously and at a scale no human team could match alone. That thesis runs underneath nearly every investment we make, and we expect it to only grow more true from here.
