Does AI really create time?
When AI saves ten hours of manual drafting, those hours rarely return to the business as clean capacity. They are usually consumed by new coordination tasks unless explicitly redeployed.
Jean Bernier
Founder, BTCF Originator
What we are trying to understand
If a business adopts technology that dramatically reduces the time required for a specific task (like drafting a report, writing code, or summarizing calls), does the business actually become faster? Or does the newly created time disappear into the system?
Why it matters
Companies are currently investing heavily in AI under the assumption of a direct 1:1 ROI: If we save 1,000 hours, we can reduce headcount or double output. But if that time is quietly absorbed by other friction, the investment yields no systemic operating leverage.
What we reviewed
We reviewed contemporary studies on AI productivity (such as the NBER study on customer support agents), historical literature on the productivity paradox of IT deployment, and our own longitudinal evidence from three growth-stage service organizations over a 12-month period of AI adoption.
What the evidence supports
Studies on technology deployment consistently show that time saved in one node of a system often creates bottlenecks in another.
- In customer support settings, AI can reduce resolution time per issue by 14% on average (NBER, 2023).
- However, in broader knowledge work, reducing the time to generate content (like code or documents) often increases the total volume of content produced.
- As generation speed increases, the burden shifts to the reviewers and approvers, who now face a 10x increase in their review queue.
- Overall cycle time (Flow Time) often remains identical or worsens, despite the initial task being completed faster.
What it does not support
This evidence does not mean AI is useless. It does not support the argument that AI fails to improve individual efficiency. It simply proves that individual efficiency does not automatically equal systemic speed.
Bernier Interpretation
Productivity is the wrong goal when the work shouldn't exist, and saving time is the wrong metric if that time isn't captured.
If you automate a process, you reduce Human Time (effort). But unless you change the system, you do not improve Flow Time (speed of value).
To actually capture the value of AI, companies must establish what we call Verified Time Created (VTC). If you save ten hours, you must explicitly redeploy those ten hours into Opportunity Time, otherwise they will be absorbed by new meetings, more emails, and increased coordination.
What this changes
Before deploying AI to "save time," an operator must:
- Measure the current baseline cycle time.
- Define exactly what the person will do with the saved hours (Time Redeployment).
- Ensure the downstream reviewers have the capacity to handle increased output, or change the decision rights so review isn't required.
Open Questions
If generation becomes nearly free, where does the operating burden move? How do we measure the cost of review?
Sources
- NBER Working Paper: Generative AI at Work (Brynjolfsson et al., 2023)
- HBR: The Productivity Paradox of New Technology
- Anonymized longitudinal evidence from 3 growth-stage operating models
Follow the Idea
What question should this raise?
If the logic in this piece is true, what current operating assumption in your business must be false?