Leading and Lagging Indicators in AI.

Most people read the AI industry through the wrong instruments. The fix is an old idea from strategy and economics.

By Eric McLean · 3 June 2026 · Part of Learning

A lagging indicator confirms what has already happened. Revenue, market share, adoption surveys, a model topping a public benchmark — all are real, all are concrete, and all are rear-view. By the time they move, the event that caused them is finished.

A leading indicator is an early signal of what's coming. It's noisier, more ambiguous, easier to dismiss — and far more valuable, because it gives you time to act before the rest of the market has noticed.

Why this matters more in AI than almost anywhere

In a slow market, lagging indicators are fine; the world doesn't move much between the event and its confirmation. AI is the opposite. The gap between "a capability appears" and "it shows up in the revenue figures" can be a year — and the window to position yourself closes long before the lagging number confirms the shift. Run your AI strategy on lagging indicators alone and you will always be a year late.

The trap is that lagging indicators are comfortable. They're published, quantified, defensible in a board meeting. Leading indicators require you to read ambiguous signals and back a judgement. That discomfort is exactly why most people avoid them — and exactly where the edge is.

What to actually watch

Leading indicators

Read these to see what's coming

  • What the frontier models can just barely do today — the ragged edge is next year's default.
  • What the labs are hiring for, and where the capital and compute are being committed.
  • Research direction — what's on arXiv, what the serious people are excited by.
  • Developer behaviour — what's being adopted, forked and built on, before it's monetised.
  • Capability demos and the limits people are probing.

Lagging indicators

Useful for confirmation, dangerous as your only guide

  • Revenue, ARR, market share.
  • Public benchmark leaderboards — by the time a model tops one, its successor is already training.
  • Enterprise adoption surveys and headcount changes.

How GrabStack is built around this

This isn't abstract — it's the design of the site. The Landscape's "signal to watch" line on every debate is, deliberately, a leading indicator: the thing to monitor now that will tell you which way the debate breaks. Frontier shows the ragged edge — a leading indicator of what's about to go mainstream. The Wire reports events, which are often lagging — the launch is the confirmation of a shift that the leading signals flagged months earlier.

The heuristic

For any AI question — a tool to adopt, a bet to make, a debate to call — ask two things: what's the leading indicator here, and what's the lagging one? Then weight the leading one more heavily the faster the domain is moving. In AI, that means weight it heavily.

Lagging indicators tell you the race you already lost or won. Leading indicators tell you the one you can still enter.

3 June 2026 · Review by 3 September 2026.