AI has been compared to a lot of transformative technologies, but the internet and

broader ICT boom of the 1990s may be one of the most useful comparisons.

Not because we should expect history to repeat exactly, but because that period

gives us a precedent for how a major technology moves from investment and adoption

into measurable productivity. So the question I want to explore is, if AI is

following a similar path, where are we today, and what might come next?

The clearest thing we can say about AI today is that the build-out is real.

The top chart uses data center investment as one visible measure of that

infrastructure build-out. Since just before ChatGPT was released, real investment

has increased by roughly four times.

But when we look at economy-wide productivity below it, the picture is much less

dramatic. Productivity growth has been reasonably healthy recently, but we've not

seen yet the same kind of unmistakable break from the previous trend.

So I would describe the current moment this way: The build-out regime change is

clear. The productivity regime change is not, at least not yet.

And that gap between investment and realized productivity is where the historical

comparison becomes useful. If we go back to the ICT boom, we can see a similar

sequence. Computer investment began accelerating in the early 1990s and then

continued rising very rapidly. By the end of the decade, it was roughly 15 times

its 1991 level. Productivity was much messier.

There were early bursts of growth, but they did not immediately turn into a

sustained trend. It was several years into the investment boom before a broader

productivity acceleration became much clearer.

That does not mean AI has a fixed three-year or five-year waiting period.

The technologies and the economic environment are different.

The useful lesson is about the sequence.

A technology can be spreading quickly and attracting enormous investment before its

full effect becomes obvious in the aggregate economic data.

History also gives us an idea of where to look for the payoff first.

Before the mid-1990s productivity acceleration, productivity growth in IT-intensive

industries was fairly close to the rest of the economy.

Afterward, the gap became much larger.

The industries using IT most intensively were the ones where stronger productivity

growth became much more visible. That gives us a useful signal to watch with AI.

On the right, industries with high AI exposure have recently begun to outperform

lower exposure industries on productivity.

But this is very new, and these sectors have different underlying characteristics

to begin with. So I would not call this evidence of an AI productivity boom yet.

What becomes interesting is if that gap persists and continues to widen.

That would be an early sign that the productivity effects are becoming more visible

before they necessarily show up across the entire economy.

And that leads to the business implication.

The lesson from ICT is not to sit back and wait for economists to declare that the

productivity boom has arrived. By the time the aggregate data makes it obvious,

some businesses may already be well into capturing the value.

I think there are three stages to think about, and they mirror what we saw in the

ICT era. First, the technology spreads, then

firms learn how to embed it into the way they operate, and only then does the

value become more visible and repeatable. First is access and adoption.

AI usage and AI spending tell you that the technology is spreading, but they're

inputs, not outcomes. Simply buying more tools does not tell you whether they are

creating value. Second is organizational absorption.

This is the work around the technology, redesigning workflows, getting the data

right, developing skills, putting governance in place, and changing how work

actually gets done. And the third is tangible value.

That is where businesses should be looking for repeatable improvements in output,

cost, cycle time, quality, or capacity.

This is also where the link to financial planning becomes important.

The ICT experience is a reminder that productivity gains do not arrive everywhere

at once. So finance teams don't need to make one heroic assumption about how much

AI will improve productivity across the company.

They can build those effects into planning progressively.

Where AI is producing measurable savings or additional capacity, those gains can

begin informing labor, cost, margin, and investment assumptions.

Where the evidence is still uncertain, planners can use scenarios rather than

locking the benefit into the base case.

So the bottom line is this: If AI is really following the ICT path, the macro

payoff may take time to become obvious.

But businesses should be using this build-out period now to develop the

capabilities, measurement, and planning discipline that can turn the technology

into measurable value.