Who Will AI Agents Empower?

Main

BTU researchers’ assessment is that an AI agent can create a new opportunity for a person, but it can also merely redistribute an already scarce opportunity in that person’s favour. Georgia’s wider benefit lies in expanding the former: removing barriers that make meaningful goals in learning, work, services and business difficult to achieve.

Key Takeaways

• BTU’s analytical synthesis distinguishes access to AI from the economic capability it provides: a useful outcome requires an understandable system, relevant information and connections to the services where the work is completed.
• Georgia’s 2026 statistics show internet access in 93.5% of households, while 20.7% of people aged 15 and older have basic or above-basic overall ICT skills. These indicators have different populations and do not measure Georgian AI-agent adoption or inequality.
• BTU’s assessment distinguishes agents that remove barriers to useful activity from agents that merely strengthen their users’ position in competition for scarce opportunities; the difference matters for broadly shared benefits.

One important opportunity in AI agents is to make help available that previously required a specialist, substantial free time or familiarity with several software systems. An employee could prepare for an unfamiliar assignment, a small organisation could obtain information it needs, and a citizen could understand the conditions of a service. The economic significance is not simply faster work. An activity that someone previously could not begin, because it was too complicated or assistance was too expensive, could become possible.

Download counts cannot show the full scale of that opportunity. Two people might use the same model, yet one receives an understandable response and completes the task while the other is left with text that requires additional knowledge and assistance. The difference emerges after login: who can turn a personal goal into a usable result?

A Divide Beyond Receiving Answers

“Agentic inequality,” published in Nature Machine Intelligence on 20 October 2026, examines three dimensions: agent availability, quality and quantity. The authors’ conceptual analysis considers how unequal capacity for delegated action could change the position of people and organisations. It is a theoretical framework, not a measurement of inequality in Georgia. Nature Machine Intelligence

BTU’s research materials distinguish income inequality from unequal access to the resources that produce outcomes. People with similar incomes today might have different opportunities tomorrow to find suitable education, develop a project, navigate a complicated administrative process or offer a new service. If one party can use such opportunities more frequently and accumulate experience, a small initial difference may deepen.

An agent’s role in this process goes beyond finding information. A good explanation is not the same as a completed application, and a completed application is not the same as securing a suitable opportunity. Different barriers arise along the way: an unclear condition, unavailable information, disconnected services or a cost that appears halfway through the process. A stronger answer cannot resolve all of them.

Economic capability therefore cannot be reduced to model quality. An agent might explain a course description clearly in Georgian but fail to establish whether its schedule fits the user’s working hours or which document registration requires. If resolving that uncertainty still requires an expensive intermediary, information has become easier to obtain while the goal may remain difficult to achieve. This illustrates a mechanism; it is not an assessment of a particular educational institution.

Georgia Already Has Broad Connectivity

Georgia’s starting point should not be described only in terms of deficits. Geostat’s June 2026 survey records internet access in 93.5% of households, a broad foundation for reaching new forms of service. A different aspect of the digital environment appears in its overall ICT-skills indicator: 20.7% of people aged 15 and older have basic or above-basic skills. Geostat data

Two Indicators of Georgia’s Digital Environment

Source: Geostat, June 2026; internet-access and ICT-skills tables. The percentages concern different populations and must not be subtracted from one another. Values rounded to one decimal place.

The internet indicator concerns households; the skills indicator concerns people aged 15 and older. Subtracting the two rates would be invalid. The skills indicator combines information and data literacy, communication and collaboration, digital content creation, problem solving and safety. It is not a test of AI-agent use and does not establish that everyone outside the category is incapable of using the technology.

BTU’s calculation puts the urban–rural difference in the ICT-skills indicator at 12.3 percentage points. This describes the existing digital environment, not an effect of agents. Its significance is that offering the same software in different places may require different support. A new service that demands complicated technical setup cannot ignore that difference.

Conversely, an agent could help reduce precisely this complexity. Natural-language interaction may remove the need to understand several menus and forms. For someone with limited digital skills, the answer is therefore not always more technical instructions; it may be a clearer service and accessible assistance when needed. A useful system reduces the existing burden rather than requiring users to master it in full.

When Capability Expands-and When Only Advantage Does

BTU’s analysis separates two economic situations. In the first, an agent reduces a barrier that made an activity expensive or impossible. Clarifying a request, finding current conditions and helping complete an incomplete application can benefit both the individual and the service provider. Fewer misunderstandings can mean fewer repeated enquiries and a process that is easier to finish.

In the second situation, people compete for a scarce opportunity. If the number of available places is unchanged, a faster agent may help its user without creating any additional places. One person might secure an opportunity another misses. The individual benefit is real, but it does not automatically represent an expansion of opportunity for society.

Consider an illustration: an agent explains a service’s conditions and helps a citizen prepare the necessary document. Its value lies in reducing complexity. If the same system instead captures a limited appointment before others can reach it, the allocation rule also becomes relevant. These are different problems for the service organiser. One concerns simplifying a process; the other concerns how faster software redistributes access between people.

This distinction travels across sectors. In education, helping more people prepare is not the same as allocating an unchanged number of places. In employment, finding a suitable opportunity differs from increasing application volume. In business, discovering an unmet need may create new demand, whereas sending more automated messages to the same customer may simply burden that customer’s attention. Counts of agent actions conceal these differences.

The question “Who has more agents?” therefore needs a second question: what do those agents make possible for their users and for others? When technology simplifies a dysfunctional process, several parties may benefit together. When it merely accelerates a contest, better technical capability also calls for better service organisation so that success is not determined only by having the fastest system.

Common Foundations Can Serve Many Sectors

One route to broad benefits in Georgia is to reduce barriers that recur across activities. Understandable Georgian-language conditions, current public information and consistent information across connected services do not help just one company. They can support employment, education, trade, professional services and citizens’ everyday affairs.

Their economic significance is the reduction of repeated costs. If every organisation must separately decipher the same rule and explain it again to its software, even an affordable model becomes expensive to use. When the shared condition is clear and current, organisations can devote more resources to what is specific to their own work. This does not imply building one universal agent for everyone: different tasks require different professional knowledge.

Universities, professional associations and service providers can do more than promote new software. They can help people understand which activities an agent is suited to and how the assistance connects to their own goals. Professional knowledge might reach more people while complex, individual cases remain with the relevant specialist. That would expand the reach of assistance, not promise to replace every specialist function.

BTU’s research framework does not locate a small state’s opportunity in producing every technological component domestically. Using a global model can be valuable when local language, needs and services are not left outside it. Maintaining choice matters too: changing a provider should not force a person to begin an essential activity all over again. Technological options should ultimately serve the user’s ability to act.

A Person Is More Than a Logged-In User

Delegation empowers a person when the goal remains their own. Even a successfully executed action can be unhelpful if the system pursues what is easy to measure rather than what the person needs. A quickly identified course does not help someone who cannot attend at the required time; a prepared offer creates little opportunity if the business cannot fulfil its conditions.

Practical independence therefore involves more than operating software. People need to be able to clarify their needs, change the proposed route and obtain help when the situation is unclear. Choosing not to use an agent should not automatically exclude them from an essential service. Otherwise, developing a digital channel may reduce the provider’s costs while shifting part of the burden to a user with less support.

This argument does not require identical systems or identical outcomes for everyone. What matters is that basic assistance is sufficiently useful for real activities, and that a more expensive tool does not become the only ticket to participation. This is where technological development meets human judgment: which task are we making easier, which barrier are we removing, and where does the person retain a meaningful alternative?

Conclusion

AI agents could expand access to professional assistance and economic activity, but software access does not secure those gains by itself. BTU researchers’ assessment is that Georgia needs an environment in which broad connectivity connects with understandable services, usable skills and real human choice.

Progress is not measured only by how many agents are operating. It also concerns how many people can learn, work, obtain a service or begin an activity that was previously too difficult. In this sense, a powerful agent is not merely one that performs the most actions. It is one that genuinely expands a person’s capabilities.

Questions and Answers

Does 20.7% mean only that share of people can use AI?

 No. It is an overall ICT-skills indicator for people aged 15 and older, not an assessment of readiness to use AI agents.

Would cheaper AI solve the problem?

 Lower prices matter, but even an inexpensive answer may be unusable without relevant information, connected services and understandable assistance.

Why do more automated actions not necessarily mean greater benefit? 

Some remove barriers; others accelerate competition for an unchanged resource. Their economic significance depends on that difference.

What does this mean for a service provider? 

Distinguish removing a barrier for users from allocating an already scarce opportunity faster; request counts alone cannot show which outcome occurred.