Key Takeaway

Debates about artificial intelligence frequently start with job counts. Yet the transformation visible inside firms is often more granular: employees spend less time preparing initial drafts, more time reviewing outputs, and managers must define who is accountable when software acts. An AI tool becoming available does not establish a measured productivity improvement or a change in headcount.

In the Financial Times of 26 October 2026, Sarah O’Connor and John Burn-Murdoch describe how AI agents are changing workflows even at Anthropic, a company developing such technology. The account highlights a paradox: systems automate technical tasks, yet managers may spend more time checking outputs. It is a company-specific case, not evidence that an identical pattern already exists across Georgian workplaces.

Tasks Change Before Occupations Disappear

A profession is a bundle of activities, not a single job description. A financial analyst gathers evidence, validates assumptions and explains findings. A marketing specialist writes copy, studies customers and makes brand decisions. A lawyer searches documents and assesses legal risk. AI may accelerate stages of these workflows while leaving the most consequential judgements and responsibilities with humans.

An AI agent differs from a text generator because it can carry out a sequence of steps, interact with tools and pursue a defined objective. Greater autonomy, however, creates new questions about data security, unintended actions and error recovery. Firms need to decide when human authorization is mandatory, what the agent may access and how its performance will be audited.

The FT case shows why a simple time-saving narrative is inadequate. A task may take minutes rather than hours to draft, but the total process can still require oversight, correction and coordination. Managers need to measure the complete workflow, not merely the speed of the machine.

International Evidence and Its Limits

The International Labour Organization’s 2025 study of generative AI measures exposure at the task level. High exposure does not automatically translate into unemployment; occupations can be transformed as their component tasks are reorganized. Demand, regulation, organizational choices and workers’ ability to learn all influence the eventual outcome.

The World Economic Forum’s Future of Jobs Report 2025 reports that surveyed employers expect, on average, 39% of workers’ existing core skills to change or become outdated between 2025 and 2030. This is an expectation from an international employer survey, not an observed Georgian result or a forecast for Georgian employment. AI and big data, cybersecurity and technological literacy stand out among areas where employers anticipate rising demand.

The distinction between exposure and prediction is crucial. Task-level research tells us which activities technology may affect. Employer surveys report planned responses and expectations. Neither provides an empirically established number of Georgian positions that will disappear because of AI.

What Georgian Evidence Shows

Geostat’s survey of enterprise information and communication technologies reported that 94.9% of Georgian enterprises had internet access in 2025. Internet connectivity is a prerequisite for many applications but not proof of effective artificial-intelligence adoption. Official data discussed in BTUAI.ge’s September 2026 analysis suggest that approximately 4% of surveyed enterprises used AI in some form in 2025. The statistic applies to the survey’s defined enterprise population, not to every self-employed person or employee.

The BTUAI.ge analysis also examines a July 2026 survey of 37 small and medium service businesses. It highlights resource and capability constraints, while its narrow sample cannot support economy-wide estimates. Large enterprises appear better placed to deploy AI, but adoption alone cannot establish causal effects on staffing, wages or efficiency.

Georgia’s unemployment rate stood at 13.8% in the second quarter of 2026 according to Geostat. That is an important contextual statistic, but it does not isolate AI-related labour-market changes. A national employment series should never be converted into a claim about technology-driven job losses without additional evidence.

The Challenge for Junior Employees

Entry-level work often includes collecting information, preparing first drafts, processing routine requests and checking records. These are among the activities that AI tools can assist with. If a firm delegates every introductory task to software, it must still find a way to train the next generation of experienced analysts, engineers and managers.

One response is to redesign junior roles around verification, explaining results, interacting with clients and improving processes. The employee learns how to use AI while also developing the professional judgement required to challenge its answers. This is different from removing junior opportunities altogether.

Management Becomes More Important, Not Less

As routine output becomes easier to produce, leadership must spend more effort on reliability and accountability. An AI-generated customer reply still needs rules about confidentiality and escalation. A financial summary requires evidence and clear assumptions. The organization must assign responsibility for validation and identify when a system should stop and ask for human review.

Small Georgian firms face a particular trade-off. A few people may cover sales, operations and administration simultaneously, making routine assistance valuable. Yet if every generated result has to be recreated from scratch, the apparent time saving will shrink. The relevant performance measure is the whole process.

An Illustrative Time-Saving Scenario

Assume five employees each spend six hours per week preparing first versions of repetitive reports: 30 hours in total. Suppose an AI tool cuts drafting time in half but adds one hour of review per employee. The new total becomes 20 hours: 15 hours drafting plus five hours reviewing. The net saving is 10 hours a week, one-third of the original workload.

This arithmetic is hypothetical, not a measured result from a Georgian company. It does not amount to cash savings or layoffs. Software subscriptions, implementation, error correction and information security would also need to be assessed before an organization could calculate an actual return.

Table 1. Illustrative weekly workflow

Opportunities and Risks in Georgia

A tourism provider may use AI to prepare initial responses in several languages. A retailer may draft product descriptions. A smaller manufacturer may automate the first stage of inventory reporting. These are potential applications, not evidence of realized gains. Each should be tested against accuracy, cost, customer experience and employee time.

Language matters. Tools optimized for English may struggle with specialized Georgian terminology, local rules or communication norms. Verification in Georgian should be designed into deployment rather than treated as an afterthought. Universities and vocational providers can respond by emphasizing source evaluation, data reasoning, domain judgement and accountability alongside tool operation.

The analytical assessment in this article is that Georgia should prioritize work redesign over simplistic predictions of disappearing occupations. Pilot projects should compare baseline time, errors and service quality with results after implementation. Saved hours and fewer employees are different outcomes. The former does not establish the latter.

Conclusion

Artificial intelligence is reshaping professions, but not according to a single predetermined path. Technology reorganizes tasks; firms decide how to use the capacity it creates. A gain can become better service, stronger skills or lower costs, depending on management choices.

For Georgia, competitive advantage will depend on more than access to AI software. Training, responsible oversight, Georgian-language reliability and clear rules for human decisions are essential. The decisive question may not be which enterprise uses more tools, but which enterprise organizes human and machine work more intelligently.

Data and Main Sources

Financial Times – Sarah O’Connor and John Burn-Murdoch, “How are AI agents changing the way their own creators work?”, 26 October 2026, p. 14 – მომხმარებლის მოწოდებული ბეჭდური ნომერი

National Statistics Office of Georgia, Use of Information-Communication Technologies in Enterprises – 2025 – https://www.geostat.ge/media/79412/Use-of-Information-Communication-Technologies-in-Enterprises—2025.pdf

National Statistics Office of Georgia, Indicators of Labour Force – II Quarter 2026 – https://www.geostat.ge/en/single-news/3832/indicators-of-the-labour-force-employment-and-unemployment-ii-quarter-2026

International Labour Organization, Generative AI and Jobs: A Refined Global Index of Occupational Exposure (2025) – https://www.ilo.org/publications/generative-ai-and-jobs-refined-global-index-occupational-exposure

World Economic Forum, The Future of Jobs Report 2025 – https://www.weforum.org/publications/the-future-of-jobs-report-2025

BTUAI.ge, AI Integration in the Business Sector in Georgia, 30 September 2026 – https://btuai.ge/en/ai-integration-in-the-business-sector-in-georgia/

Prepared by the academic team of Business and Technology University and the BTUAI Research Team, Tbilisi, Georgia.