When New York Governor DeWitt Clinton championed the Erie Canal in 1817, critics dismissed the proposed 363-mile waterway as an expensive ditch leading nowhere. They were looking at the infrastructure rather than what would flow through it.

Construction took eight years. For most of that period, workers were not merely digging a trench; they were building the infrastructure through which future commerce would move.

When the canal opened in 1825, freight costs between Buffalo and Albany fell by roughly 90%, travel time dropped sharply, cities along the route expanded and New York City emerged as America’s commercial center, according to the Erie Canalway National Heritage Corridor .

The canal’s value was not the ditch itself. Its value came from the commerce it enabled. That distinction offers an important lesson for the investment industry’s fascination with artificial intelligence.

AI may be the canal. But data is the water. And most firms are filling their new canals with the same water everyone else already has.

Faster Through the Same Crowded Channel

Asset managers, hedge funds and investment platforms are adopting AI to summarize earnings calls, analyze filings and accelerate research. McKinsey has noted that generative AI could deliver meaningful efficiency gains in investment management. But efficiency and investment advantage are not the same thing.

When firms apply similar models to the same earnings releases, transaction records, analyst estimates, news feeds and market prices, the result may be faster consensus; not differentiated foresight.

Lasting alpha rarely comes from processing widely available information more efficiently. Any advantage built on common data will narrow as competitors acquire comparable technology.

The greater opportunity is not simply to build a faster canal. It is to find a different source of water.

Moving Upstream From Transactions

Most financial and alternative data describes activity that has already occurred. Credit card data, web traffic, store visits, shipping activity, earnings reports, and market prices are useful, but they primarily observe downstream consequences.

Forward-looking zero-party data operates farther upstream.

It is information people intentionally provide about expectations, priorities, planned purchases, financial outlook and motivations. When collected regularly from representative populations, it can identify the forces forming demand before they appear in transactions or reported results.

The Federal Reserve Bank of New York uses its monthly Survey of Consumer Expectations to measure views on inflation, spending, earnings, employment, credit, and housing. Its research underscores that expectations influence behavior and have implications for macroeconomic activity and monetary policy.

That is the central investment insight: Human expectations are not merely commentary about the economy. In aggregate, they help create it.

Before a consumer buys a vehicle, changes retailers, cuts spending or trades down, confidence, financial pressure or purchase intentions often change first. The transaction comes later.

The Eight-Year Data Advantage

Creating a differentiated time-series asset requires the same kind of commitment that built the Erie Canal.

A firm cannot manufacture eight years of forward-looking consumer expectations overnight. Those observations must have been collected month after month using stable methodology, representative samples, and rigorous quality controls.

But investment leaders do not need to wait eight years. The immediate opportunity is to identify the relatively few data owners that have already made that investment and connect those historical assets to today’s AI systems.

A longitudinal dataset captures what a snapshot cannot: how shifts in confidence, purchase intentions and financial pressure later appear in spending, company revenues, and the broader economy. That accumulated history is what turns data into signal.

AI software may be the canal, current zero-party data the water, and the multiyear time series the reservoir that makes the system dependable. That reservoir already exists in some places. The challenge is finding, validating, and accessing it before competitors do.

AI software can be acquired quickly. Historical observations cannot be reconstructed after the fact. You can buy the AI model tomorrow. You cannot buy yesterday’s missing observations unless someone had the foresight to collect them. That makes established longitudinal data assets strategic infrastructure.

Tim Geannopulos , Partner at Broadhaven Capital and former CEO and Chairman of Trading Technologies, explains, “In the Fintech world, everyone relentlessly seeks to fortify their ‘Technology Moat’ so that their software product can remain unique and competitive. This is especially the case in the growing fintech vertical of alternative data, where the alpha-hungry trading industry is perpetually searching for the scarcest of data to be a differentiating ingredient in their proprietary trading model soup.”

Demand-Formation Data Provides a 41-Day Head Start

Geannopulos continues, “One illustrative use case of such alternative data is consumer intent survey data, which can predict key macroeconomic indicators—such as PCE and CPI—several weeks to months in advance and thus provide actionable signals for trading fixed income, FX, equities, commodities, crypto and especially prediction markets. The trading firm that owns eight, but especially 25 years of such survey data has a moat that is as impenetrable as Fort Knox—unless someone can build a time machine to go back in history and conduct such surveys themselves.”

This observation gets to the heart of the issue. Technology can be copied, licensed, or developed. A continuously collected historical record of human expectations cannot.

What New Water Makes Possible

Applied to continuously refreshed, representative demand-formation data, AI can produce outcomes conventional research systems struggle to generate.

Models can translate changing intentions into advanced forecasts of retail sales, personal consumption, inflation, employment, housing, and automobile demand.

At the company level, shifts in category intentions, retailer preference, financial confidence, and planned spending can improve revenue forecasts before quarterly results.

At the portfolio level, investors can identify companies positioned to benefit from emerging demand—and those likely to face pressure—before changes become fully visible in analyst estimates.

At the risk level, probability models can detect deteriorating household conditions, weakening purchase intentions or rising caution before these forces appear in traditional economic releases.

The strongest systems will connect upstream intentions with transactions, fundamentals, and market data.

But the upstream signal may provide something scarce: time. And in investing, time is often the raw material from which alpha is created.

The Data Becomes the Moat

CFA Institute has observed that alternative data and natural-language processing can help investors develop a clearer, quicker picture of the world, while warning that relevance, reliability, and data quality remain critical.

That warning will matter more as AI models proliferate. Algorithms will improve. Computing costs will decline. Sophisticated software will become accessible to nearly every credible investment organization. The models themselves may become less differentiated.

The durable advantage will increasingly come from proprietary, timely and difficult-to-replicate data flowing through them. Instead of asking only, “Which AI platform should we use?” investment leaders should also ask: “What do we know that the rest of the market does not yet know?”

The Erie Canal did not merely make an existing journey faster. It created a new route that changed where commerce flowed, which cities prospered and who could compete.

AI powered only by familiar, backward-looking data may make existing research faster.

AI powered by forward-looking demand-formation data could create a different route to knowledge—one connecting human expectations to economic outcomes before those outcomes become obvious.

The market sees AI software. The more perceptive investor should be looking for the new water, the reservoir behind it—and the organizations that had the foresight to start filling that reservoir years ago.

The consumer sentiment study referenced above was conducted by my company, Prosper Insights & Analytics . This is the same dataset used by the National Retail Federation, and available from Amazon Web Services, Bloomberg, and the London Stock Exchange Group for economic benchmarking.