Кои професии ще изчезнат и кои ще оцелеят? Светлин Наков

Summary

AI is automating tasks, not necessarily eliminating whole occupations. A profession changes substantially when AI takes over most of its tasks. Companies may then produce more with the same staff, hire fewer people, or gradually reduce headcount. The impact depends both on how much work can be automated and on whether there is a strong business case for adopting the technology.

Business frameworks and operating principles

• Analyze work task by task. Break a role into its component activities, then assess which tasks machines can perform, which AI can assist with, and which still require human judgment or interaction.
• Automation is most likely when work is:
  • Digital and standardized.
  • Repetitive, with clear steps and an expected result.
  • Easy to verify, even if the work itself is complex.
• Use a human-in-the-loop workflow. Let AI draft, research, classify, or execute routine actions. Have a knowledgeable employee review exceptions and approve consequential outputs. This can increase employee throughput without removing the role.
• Evaluate automation as an operating investment. Adoption depends not just on technical capability but also on cost and economic return, reliability and safety, regulation, and whether workflows and workplaces can be redesigned to accommodate it.
• Build capability through practice. Education provides a foundation; real projects and work experience turn knowledge into practical skill.

Company examples and operational applications

• Software development: Programmers increasingly direct AI agents and check whether their work meets requirements, follows good practices, and is correct. AI can generate standard code, tests, and documentation, but humans remain responsible for specifying, reviewing, and iterating on the work.
• Customer support at SoftUni: An agent reads an incoming email, checks the customer’s profile and payment or course information, searches for an answer, and drafts a reply. A person reviews and sends the response rather than leaving it fully automated.
• School administration: An agent can generate a teacher’s class schedule and enter the information into the school system. The speaker describes a process involving roughly 60 schedule rows and about 80 system entries, completed in around 20 minutes while the employee does other work.
• SEO analysis: An agent connected to Google Search Console, Google Analytics, and other sources gathers and analyzes site data. One cited result was a course moving up two positions in search rankings; the speaker does not attribute that result solely to the agent.
• Accounting: AI can classify invoices, suggest project or asset coding, and prepare routine reports and reconciliations. Accountants still handle unusual cases, investigate errors, and deal with authorities or complex decisions.
• Other automatable office work: First-line customer support, standard documents and correspondence, content and translation, routine design and video tasks, legal research and document review, and simple data or marketing analysis.
• Physical operations: Production, sorting, packaging, warehouse movement, cleaning, and some agricultural operations are already being automated. More variable work—such as repairs, plumbing, electrical work, or picking delicate crops—is harder to automate reliably.

Metrics and claims cited

• The SoftUni speaker says AI replaced the work of 35 people over the previous two years. He describes the change as often gradual, including not replacing people who leave, rather than always involving abrupt layoffs.
• In an illustrative programming example, a team’s queue grows from 100 to 400 tasks, while monthly throughput rises from 10 to 300 tasks, with salaries unchanged. This is an example of increased capacity, not a general benchmark.
• In the support workflow, the AI-generated answer is said to be valid in 95% of cases, with a person reviewing and approving it.
• The school scheduling example involves approximately 60 rows and 80 system entries, taking about 20 minutes.
• The speaker believes a B2B salesperson will remain difficult to replace even in 10 years, because the work depends on human meetings, discussion, and trust. By contrast, he expects routine calls to large banks to be increasingly handled by AI agents.

Implications for roles and skills

• More exposed: Standardized digital and administrative work, routine content production, basic coding and testing, first-line support, simple accounting tasks, and repetitive processes with clear instructions.
• More resilient for now: Skilled trades and repair work requiring dexterity and on-site adaptation; roles involving complex exceptions and accountability; and work that depends heavily on trust, communication, care, or negotiation.
• Skills likely to retain value: Professional expertise, critical thinking, problem definition, setting goals and success criteria, checking data and sources, evaluating AI output, managing exceptions, communication, negotiation, teamwork, and the ability to keep learning.
• AI literacy is becoming a basic workplace capability: choosing tools, providing clear context and instructions, and checking results—not just knowing how to write prompts.

Actionable recommendations

• Map roles into tasks and identify a contained, repeatable process for an AI-assisted pilot.
• Keep human review where errors carry meaningful operational, financial, legal, or safety risks.
• Measure outcomes such as throughput, review time, error rates, service quality, and total cost—not just the amount of work automated.
• Develop domain expertise alongside AI skills; employees need enough knowledge to spot omissions and incorrect outputs.
• Build experience early through real projects, internships, volunteering, and practical work. These also develop reliability, communication, discipline, and familiarity with how organizations operate.
• Treat automation as a redesign of work and processes, not merely a software installation.

Presenters and sources

• Presenter: Svetlin Nakov.
• Source: Excerpt from a live SoftUni Community seminar; video published by the Stefcheva channel.