Как не потерять работу и бизнес в эпоху ИИ? Лекция Эвана Голованова в Финансовом Университете.

Summary of Key Arguments and Points

• AI will disrupt “office” and humanities-type careers fastest. The lecturer argues that many people in Russia study and work in humanities/social sciences—areas where AI impact is expanding (management, business/finance, law, parts of social sciences, media/art). He suggests that education pipelines producing large numbers of such degrees (e.g., paid psychology programs) will become increasingly vulnerable.

• AI capabilities are accelerating exponentially, making entry-level roles increasingly replaceable. He cites industry activity (e.g., Anthropic/Claude) and claims that model capability roughly “doubles” every few months. As a result, tasks previously done by juniors or entry-level specialists—especially coding/IT-like work, testing, design, and similar—can be automated or handled by more experienced systems using “vibe coding” and agentic workflows.

• Western layoffs and hiring trends demonstrate replacement pressure. He points to layoffs at major corporations (Amazon, Oracle, etc.) and to research showing a decline in the share of young (22–25) hires in IT. His argument: AI-driven “agents” may replace junior pipelines rather than create new jobs at the bottom.

• Russian labor-market data shows resume overload in “intellectual” roles. Using HeadHunter-style metrics (resumes per vacancy), he claims there are far more applicants for strategy/consulting/management-adjacent roles (e.g., “94 for one vacancy” type examples). By contrast, he notes labor shortages in physical roles (cashiers, bakers, welders, janitors). His conclusion: office work will face stronger competition and restructuring.

• Recommended response: retrain and rebuild efficiency via AI, not nostalgia. He frames the economic logic as follows: the state and businesses want increased efficiency, and AI will be a main tool. People whose work doesn’t fit those efficiency gains will be pressured toward retraining into roles with shortages. Practical takeaway: adapt now or risk being squeezed out.

• “Vibe coding/agents” are real, but most people don’t actually use them effectively. He argues there’s limited real competition because most people only consume AI content/courses, while a minority truly implement workflows. He uses personal metrics (install counts and usage examples) to claim most people “chase narratives” rather than build functional systems.

Core Technical/Practical Thesis

• The winning skill is managing context with AI tools. He emphasizes that LLMs are not like humans—they rely on explicit context (tokens) and can hallucinate or mix previous text when overloaded.

He advocates a workflow hierarchy:

1. **Chat**
2. **Projects** (persistent context: files/knowledge, writing style, requirements)
3. **Skills** (automated repeatable tasks; avoids rewriting prompts)
4. **Agents** (autonomous performers with error-handling; can include sub-agents)

Personal Implementation Examples

He describes using tools like Claude and cloud-code (and GPT for entertainment/aux tasks) to build:

• Skills for recurring content checks (e.g., proofreading posts)
• Projects for topic-specific writing and analysis
• Agents to automate information collection and personalized news summaries (e.g., daily digests to reduce doomscrolling “costs”)

Quant Analysis as Proof-of-Value

He claims he used these tools to perform quantitative content analysis of his own output and observed shifts—for example, geopolitics-related spikes tied to Donald Trump-related news cycles. His argument: AI-based quantitative workflows can reveal patterns that manual work or consultants might miss.

Career Advice Going Forward

• Move toward AI-augmented roles and decision-making. He argues that a “classic” path (e.g., learn Python → do modeling manually) won’t be enough. Employers will want people who can deploy AI tools in context and make decisions where automation is not complete. Example: in finance, he frames tasks like bond pricing/bond curve work as “skills” that can be packaged and delegated to agents.

Meta Claims

• Soft skills become more valuable later. Beyond technical capability, interpersonal/psychological development (“soft skills”) will matter. He hints that empathy/communication/psychology training may increase in value after 2026.

• Endurance depends on genuine interest, not just adaptation. He responds to “who benefits / job loss” fears by arguing that success under disruption requires motivation and genuine interest, not only adjustment. He frames “learned helplessness” as a failure mode—blaming others instead of fixing workflow/installation issues personally.

Ethics Q&A: Adaptation vs. Slowing AI

When the audience challenges the ethics of AI acceleration (job displacement, inequality), the lecturer responds that:

• Historically, attempts to slow progress failed (he references restriction advocates and “effective accelerationism” perspectives).
• At the individual level, he chooses adaptation rather than trying to stop progress (he argues the probability of “winning” is low).
• He positions AI use as potentially beneficial if applied ethically (e.g., a story where AI helped identify a dog’s cancer; he also references medicine research).
• He argues misuse should be avoided and that the best possible “good” contribution is to use AI for positive outcomes.

Presenters / Contributors

• Evan Golovanov — speaker/lecturer
• Department of Psychology and Human Capital Development (Financial University) — thanked for organizing/supporting soft-skills-related aspects
• Audience questioner(s) — unnamed attendee who raised the ethics/AI-acceleration question