Càng ngày càng bận - Nếu bạn còn dùng AI theo cách cũ

Overview

The video argues that while AI tools (e.g., GPT-style chat, agents, and generative systems) can dramatically speed up work—such as summarizing long documents, drafting emails/articles, generating images/videos, and producing plans—people often feel more busy, not freer.

The core explanation is that AI’s speed increases throughput and expectations, but it doesn’t automatically reduce the overall workload or the human effort needed to complete real projects.

Key Points

• Work expands instead of shrinking (expectations rise). Tasks may take less time to produce, but bosses/clients/teams then expect more output—more versions, more options, more competitor analysis, more content volume. Net workload increases or shifts rather than disappears.

• Employees don’t necessarily work less; they work faster and longer. A cited UC Perkele H field study (presented as 2026) claims AI doesn’t reduce workload. Instead, workers take on more task types and may extend working hours, driven by the ability to do more.

• AI introduces “hidden time”: waiting and interruptions. Even when AI completes tasks quickly, humans must wait for responses during multi-step work (research, coding, projects). These waiting periods fragment the day into smaller intervals that disrupt deep focus. People may check social media/drama while periodically re-checking the AI.

• AI shifts the bottleneck from generation to review/approval. Previously, the slow part was creating content. Now AI drafts quickly, but the remaining work—human review, verification, corrections, decision-making, formatting, and saving into the correct systems—often takes comparable or more time. A marketing example highlights that AI increases posting quantity, yet reviewers still have limited time.

• A large share of “saved time” is lost fixing AI output. A claimed 2026 study (by a national organization) says nearly 40% of AI-saved time is wasted on checking, correcting, and refining results.

• Demo success doesn’t translate to real-world reliability. Tools that look impressive in demos can be unusable or require substantial ongoing supervision due to the “last 20%” of reliability work: debugging, constant updates, correcting errors, and managing changing models/interfaces.

• Using AI can add a managerial/distraction burden. Instead of directly executing tasks, workers become coordinators and moderators of AI output. The video also warns about FOMO and distraction from rapidly changing AI technologies, which can cause people to neglect high-value responsibilities.

Practical Principles to Reduce “Busyness”

1. Question whether the task should exist at all If a report doesn’t drive decisions or create value, don’t accelerate an unnecessary process.

2. Map the whole process and identify true bottlenecks In the AI era, bottlenecks are usually review/decision—not generation.

3. Design how you’ll handle AI waiting time Set check times, do other tasks that fit within the waiting window, and avoid constantly interrupting work.

4. Start small and define uncertainty handling Clarify data sources, define “good enough” criteria, specify when humans must approve, and document responses for unusual cases.

5. Plan in advance what you’ll do with saved time Otherwise, new tasks will simply refill the time, preserving the feeling of busyness (e.g., allocate time to client work, strategy, product development, or rest).

Overall Conclusion

AI doesn’t automatically create freedom; it mainly increases output speed within the same time budget. The video reframes the goal as reducing low-value manual work and protecting time for high-value tasks that AI can’t replace.

Presenters / Contributors

• The speaker/presenter is not named in the subtitles (no contributor names provided).