# The Extra Time AI Buys You Is Already Gone

Published: 2026-07-01T13:00:00.000Z · Updated: 2026-08-01T15:06:58.000Z · Author: Rod Amora · Canonical URL: https://rodamora.com/blog/the-extra-time-ai-buys-you-is-already-gone/

> Ask about a task and AI looks transformative: 55% faster coding, 40% less time writing. Ask about the firm and three years of realized productivity gain comes to 0.29%. Freed time doesn't wait to be claimed. It gets absorbed by the tasks around it.

Ask about a single task and AI looks transformative. [GitHub Copilot cut coding time by 55% in a 2023 randomized trial](https://arxiv.org/abs/2302.06590?ref=g.rodamora.com). ChatGPT users on writing tasks [needed 40% less time, per a 2023 Science study](https://doi.org/10.1126/science.adh2586?ref=g.rodamora.com), and customer support agents with an AI assistant [resolved 14% more issues per hour on average, 34% more if they were new to the job, per an NBER study](https://www.nber.org/papers/w31161?ref=g.rodamora.com). Ask about the firm and the number collapses to almost nothing.

In 2026 the NBER published a survey of nearly 6,000 CEOs, CFOs, and senior finance managers across the US, UK, Germany, and Australia. It asked what AI did for their firms over the past three years. [More than 90% report no effect on employment, and 89% report no impact on labor productivity](https://www.nber.org/papers/w34836?ref=g.rodamora.com). Realized productivity gain over three years: **+0.29%**. What those same executives expect over the next three years: **+1.4%**. Same firms, same tools, and the number keeps moving to next year.

That gap is what economists call the AI productivity paradox, and it is the whole question. If the task is faster, where did the time go?

Freed time doesn't wait to be claimed. Left alone, it gets absorbed by the tasks around it, and the firm ends up with a faster team and the same output it had before.

## Why don't the two easiest explanations hold up?

The first answer is that [firms haven't really adopted AI yet](https://rodamora.com/blog/shadow-ai/?ref=g.rodamora.com). That doesn't survive contact with the data. [69% to 78% of firms report using it](https://www.nber.org/papers/w34836?ref=g.rodamora.com). Professional services is among the highest-adopting US sectors, at 33% at the end of 2025 [per Federal Reserve research](https://www.federalreserve.gov/econres/notes/feds-notes/monitoring-ai-adoption-in-the-u-s-economy-20260403.html), against a national rate near 20% [reported by the Census Bureau in May 2026](https://www.census.gov/library/stories/2026/05/ai-use-businesses.html?ref=g.rodamora.com). Your competitors are not sitting this out.

The second answer is the standard one: it takes time to reorganize around a new tool before the gains show up. Reasonable, except the numbers on what firms actually did with that time are small. In an [April 2026 Census Bureau working paper](https://www.census.gov/library/working-papers/2026/adrm/CES-WP-26-25.html?ref=g.rodamora.com), 64% of AI-using firms made no organizational changes at all to use it. Training staff and building new workflows, the two most common moves, each show up at only about 15% of AI-using firms. If reorganizing were the whole story, more firms would be trying it by now, over a year into the adoption curve.

Both explanations assume the time AI frees up sits there, unclaimed, waiting for someone to point it somewhere. That assumption is the thing to test.

![Draftsman's sketch of an empty archway with a large compass-like clock face behind it, one figure pointing into the open space while others stand still](https://rodamora-uploads-347628392068-sa-east-1-an.s3.sa-east-1.amazonaws.com/illustrations/2026-07-30-firm-ai-productivity-paradox/fig-illus-1.jpg "Freed time doesn't point itself anywhere")

## The time doesn't wait. It gets absorbed.

I've watched this pattern across hundreds of service firms through a franchise network's delivery data, and directly at Berry and with clients. Teams that could clearly do more did not pick up more work. Nobody was guiding them toward it. The gap was managerial: nobody had decided in advance what the freed time was for. This is [Stage 1](https://rodamora.com/delivery-model-ladder/?ref=g.rodamora.com) on the model I use to place firms: the tool gets sanctioned, individuals get faster, and the gain leaks away as slack before it ever reaches margin.

I've also seen a stranger twist: the gain from AI is sometimes offset by other tasks that quietly start taking longer. Checking a report takes more time than it used to. Employees stretch the remaining work to fill the same hours. People are more productive, but nobody helped them use the extra time, so it drifts into the void.

This lines up with what that same Census paper shows, without proving the mechanism directly: [66% of AI-using firms use it only to speed up existing tasks](https://www.census.gov/library/working-papers/2026/adrm/CES-WP-26-25.html?ref=g.rodamora.com). And using AI on tasks shows no measurable link to smaller headcount once you account for how deeply the firm [rebuilt its operations around it](https://rodamora.com/blog/what-is-an-ai-native-service-business/?ref=g.rodamora.com). People keep doing the same job, just with a faster tool and a slower neighbor task next to it.

## Why is this worse than doing nothing?

A firm that never touched AI at least knows its team's real capacity. A firm that adopted AI and watched the gain disappear has a team capable of more. How much more, nobody knows, because nobody measured before or after.

![Draftsman's sketch of two identical pavilions side by side: the left fully rendered with a low, defined ceiling, the right open above into unfinished construction lines with figures reaching upward](https://rodamora-uploads-347628392068-sa-east-1-an.s3.sa-east-1.amazonaws.com/illustrations/2026-07-30-firm-ai-productivity-paradox/fig-illus-2.jpg "A team's real ceiling, known versus unmeasured")

Worth noting: a separate 2026 NBER study of corporate executives found only [about 15% of the productivity gain firms report comes from the investment in the tools themselves](https://www.nber.org/papers/w34984?ref=g.rodamora.com). Most of the gain comes from somewhere else. The scarce resource was always management attention, and it still is.

## How do you keep the gain from disappearing?

The fix has to happen before the tool ships. Wait until you notice a gain, and it's already gone, [reabsorbed into checking, reviewing, and rework](https://rodamora.com/production-gap?ref=g.rodamora.com).

**Understand the workflow before you touch it.** Know what the task costs today, in time and in errors. Measure it directly instead of relying on memory.

**Measure productivity before and after, at the task level.** Track the specific task and the specific hours, not just "we rolled out the tool."

**Plan the next initiatives in advance.** Decide where the freed time goes before it's freed. Process improvement, better customer experience, fewer errors and less rework, whatever fits your firm. Put it on paper before the rollout begins.

**Do the enablement and the reallocation together.** Don't ship the faster tool first and figure out the new workload later. Wait, and the adjacent tasks will already have stretched to fill the gap. You'll have no reason to believe there was ever a gain to claim.

**Check the step downstream from the one you just sped up.** After the rollout, look at what happens right after the task AI now handles faster. If checking a report takes longer than it used to, the cost moved one step over. It didn't disappear. The integration only counts once the whole process is faster, start to finish.

Do all five at once and the team barely notices a bigger workload. The faster tool and the new expectations land together, and the team learns a new way of doing the work instead of a faster way of doing the old one.

Before you expand AI into any function, decide in the same breath what the team will do with the time it frees, and measure it. If you can't answer that question for a task you're about to hand to AI, the fastest tool in the world will [leave your P&L exactly where it is today](https://rodamora.com/four-numbers?ref=g.rodamora.com).