AI Is Making Everyone Faster. Why Isn’t the Business Performing Better?

On this page

1. Production sped up. Review and absorption did not.
2. Which ledger are you looking at?
3. Follow the business workflow and find the bottleneck
4. Management Advice

A few days ago, I came across a question on Zhihu, a Chinese Q&A platform often compared to Quora: “Do meetings now feel like everyone is just firing AI-generated documents at one another?”

Many of the replies came from people who had clearly had enough. Reading through them, I was left with one thought: when AI meets corporate formalism, “productivity” can become little more than polishing a bad idea—only faster.

That may sound harsh, but it captures a very real frustration for business leaders. Employees are visibly using AI. They are producing more. Documents are getting longer, presentations look better, and daily and weekly reports arrive more often. Yet revenue has not increased, delivery has not accelerated, and outcomes have not improved.

Everyone is moving faster, but the company is still standing still. That gap is what business leaders should actually be watching.

Production sped up. Review and absorption did not.

What AI improves most directly is the efficiency of production.

Writing a document, building a spreadsheet, replying to an email, or drafting a proposal—AI can dramatically reduce the unit cost of these individual tasks. An employee who once needed three days to write one report may now produce three reports in an hour.

But an organization has never needed more reports for their own sake.

Management makes a useful distinction: efficiency is doing things right; effectiveness is doing the right things. Business leaders are accountable for the latter. Efficiency is only a means. AI can complete a defined task faster, but it cannot decide whether the task is worth doing or who should own it. It can summarize and help coordinate, but it cannot draw the lines of authority, resolve conflicts of interest, or take responsibility for the result.

Production accelerates, while review and absorption lag behind. An employee produces three reports in an hour, but people still need to evaluate them, make the call, and move the work forward. Those steps continue at human speed. Output doubles, absorption does not, and the result is a growing inventory of content that nobody has time to use.

Individual productivity may genuinely improve at a single point in the workflow. The organization, however, cares about whether the entire chain moves faster. In content-heavy work, the bottleneck used to sit at the production stage. Once AI takes over much of that work, the bottleneck moves downstream—to review, coordination, and decision-making. Speed up production alone, and everything after it may become even more congested.

Which ledger are you looking at?

Many business leaders start with the wrong ledger. When you invest in AI, what you ultimately want is better organizational effectiveness, not gains that stop at individual productivity. Yet the first numbers most companies put on the table come from an “adoption-and-output ledger”: how many people used AI, how many documents they produced, and by what percentage their efficiency increased.

Those numbers look good and are easy to report. That is precisely why they so easily turn into a competition over who can produce the most AI-assisted work.

The “process-and-outcome ledger” is much harder to explain. From the moment a piece of work is proposed to the moment it is actually delivered, did the process become faster, better, and less prone to rework? Are decisions being made more quickly? Did cost, revenue, gross margin, or customer outcomes improve?

The first ledger measures individual productivity. The second measures organizational effectiveness. Look only at the first, and you can easily mistake activity for progress.

Try following one piece of work from beginning to end. Once the volume of content increases, who is waiting? Who is reading? Who is reworking? Who still cannot make the decision? When a report is submitted, how long does it take a leader to absorb it? Are the same decisions still stuck with the same people and in the same meetings? At the end of the chain, did delivery and business outcomes actually improve?

If the answer is no at too many points, then you did not buy an improvement in organizational effectiveness. You only enabled everyone to run faster inside the same old process—without getting anywhere.

It is unrealistic to expect AI alone to solve organizational-effectiveness problems. AI is valuable for processing information and shortening parts of a workflow. But setting goals, redrawing accountability, changing incentives, and owning outcomes remain the work of business leaders and managers. Can those responsibilities really be handed over wholesale to AI? The deeper problem is that many people in senior positions plug AI into an old process and then assume the organization has already been transformed.

AI is an amplifier. It amplifies the structures and processes your organization already has. If the structure is sound, AI helps you move faster. If the structure is broken, AI helps the disorder spread faster.

Follow the business workflow and find the bottleneck

To see the real constraint, trace one business workflow from beginning to end.

If employees can produce ten reports a day but the reports look polished and say very little, stop pushing the team to write faster. If the reports are genuinely useful but everything still stalls because decisions take too long, examine the decision mechanism first. Can decision criteria be defined earlier? Once the workflow is clear and functioning, then decide where AI belongs: drafting materials, reviewing them, supporting decisions, replacing manual work, or helping redesign roles and processes.

Measure people against outcomes as well. Stop rewarding whoever burns the most tokens or produces the most documents.

Management Advice

For business leaders

Do not reassure yourself with AI adoption and output numbers. Whether AI is worth the investment comes down to one question: did end-to-end delivery, decision-making, quality, and business outcomes actually improve? If not, the numbers are window dressing. Identify the organization’s bottlenecks and eliminate low-value output before deciding where AI belongs.

For managers

When the business leader asks, “Is AI actually working?”, do not submit a report that says, “The entire company now uses AI, and output is up 300%.” Submit a process-and-outcome ledger instead: which unused outputs were stopped, which waits were shortened, how much rework was reduced, and what business result ultimately changed. That helps the business leader understand the investment. Individual-productivity metrics alone only dress up a false sense of progress.

Next time, we will use management thinking to crack open another AI question.