Georgia’s computer and information services exports are expanding rapidly, but aggregate growth does not answer the strategic question of what the country is actually selling: complex knowledge and engineering outcomes, or standardized digital tasks whose global price is falling as AI improves. In the first quarter of 2026, computer and information services exports reached USD 441.3 million, up 65.7% year on year. The official balance-of-payments category, however, does not reveal the detailed task mix behind that revenue.
According to an assessment by BTU researchers, Georgia should not optimize for the largest possible number of remote digital jobs. It should move toward export services in which AI raises the productivity of Georgian specialists while clients still pay for domain context, integration, accountability, security and ownership of outcomes.
Kenya’s warning: when cheap digital labour stops being an advantage
The New York Times reported on September 9, 2026 how Kenya’s digital gig economy was hollowed out at its lower rungs. After high-speed internet expanded, online outsourcing became a path to income for educated workers doing transcription, data entry, graphic design, website development, basic coding and other remote tasks. Kenya’s government also promoted online work as part of its digital development strategy.
AI attacked precisely the most standardized parts of that model. Transcription work began to shrink as automated tools improved, while generative AI later reduced demand for routine text and other repeatable tasks. The lesson is not that all digital work disappears. It is that a national advantage based mainly on lower labour costs can erode quickly when the output is standardized, remotely deliverable and easy to verify.
For Georgia, this is a business-model warning rather than a direct forecast. If digital exports are built mainly on wage arbitrage and routine execution, AI competes directly with the country’s cost advantage.
| Work type | AI pressure | Human value | Strategic response |
| Transcription / data entry | High | Quality control and exceptions | Do not base export advantage on low price |
| Generic text / simple content | High | Brand and domain context | Move to specialized content and accountability |
| Basic programming | Rising | Architecture, integration, client process | Use AI for productivity; move up complexity |
| Domain-specific software services | Medium and rising | Domain expertise, accountability | Combine domain knowledge with technology |
| Software product | AI changes production, not the asset itself | IP, customers, data, brand | Build ownership of product/IP |
High-value work is exposed too – so searching for an “AI-proof” profession is the wrong strategy
The International Labour Organization’s 2025 refined global index shows that generative AI exposure is no longer limited to clerical work. Clerical occupations remain the most exposed, but exposure has increased in highly digitized professional roles, including financial analysts, web and multimedia developers and application programmers. The ILO’s central conclusion is that transformation of tasks is more likely than wholesale elimination of most occupations.
That distinction matters for Georgia. High-value software is not protected from AI. In fact, competition can intensify as clients obtain the same output from smaller teams. The strategic shift should therefore be from selling hours to selling outcomes.
Simple coding may become cheaper, but clients still pay heavily for system architecture, integration of legacy and new systems, cybersecurity, data quality, migration to cloud infrastructure, responsible deployment in regulated sectors and products that actually work inside real organizations. In these areas AI is both competitor and productivity tool.
Georgia’s digital exports are already large enough for the composition to matter
The National Bank of Georgia reported total services exports of USD 1.8 billion in the first quarter of 2026. Computer and information services accounted for USD 441.3 million, rising 65.7% year on year. This is no longer a marginal export category.
Georgia’s National Statistics Office, Geostat, reports that the information and communication sector employed 53.0 thousand people in the second quarter of 2026, with average monthly remuneration of GEL 4,463.9. The category includes the broader information and communication sector, not only export-oriented software firms, so it cannot be used to infer the exact composition of export jobs. It does show that Georgia already has meaningful human and economic capacity in the field.
| Indicator | Official value | Period | What it tells us |
| Total services exports | USD 1.8bn | Q1 2026 | Large service-export base |
| Computer & information services exports | USD 441.3m; +65.7% YoY | Q1 2026 | Rapid export growth |
| Information & communication employed | 53.0 thousand | Q2 2026 | Meaningful human-capital base |
| Average monthly remuneration | GEL 4,463.9 | Q2 2026 | Broad sector; not export-only |
The central data limitation is important. The balance-of-payments category does not separate basic digital processing, software products, complex engineering services, cybersecurity, enterprise implementation or other subcategories. We therefore cannot say what share of Georgia’s current USD 441.3 million is more or less exposed to AI-driven price compression.
That is itself a policy lesson. If Georgia wants to manage the quality of digital exports, measuring only the dollar value is no longer enough. The country needs a better view of what is being sold, which skills create it and where the intellectual property resides.
Sell difficult outcomes, not cheap digital hours
According to an assessment by BTU researchers, the most vulnerable export services are those whose output is standardized, instructions are short, the work is fully digital and quality can be checked cheaply. In those markets a global client will increasingly ask why a person should be paid for a task that AI can perform at acceptable quality.
A more durable model sells a complex result rather than an isolated task: secure software for a bank, integration of industrial data, cybersecurity architecture, cloud migration, responsible AI implementation in regulated sectors, or a software product that is continuously improved for international customers. None of these are immune to AI, but the client is paying for context, accountability, integration and ownership of the outcome.
The distinction between services and software products also matters. Hourly labour is continuously benchmarked against workers in other countries and against AI. A software product can accumulate value in code, customer relationships, data, brand and intellectual property. For a small country, that matters because scale cannot come only from adding people.
| Criterion | More exposed model | More resilient model | Business question |
| What is sold | Hour / standardized task | Complex outcome | What is the client paying for? |
| Context | Low context | High domain context | Can a short prompt describe the work? |
| Accountability | Low | High | Who is responsible for the outcome? |
| Intellectual property | IP stays with client | IP partly or fully retained | Does the firm accumulate an asset? |
| Revenue model | One-off / hourly | Product or recurring revenue | Does value compound with the client? |
Education needs a different target
If routine digital tasks are becoming cheaper globally, education policy should not stop at producing more people who can write code. Georgia needs specialists who combine technical capability with finance, logistics, manufacturing, energy, healthcare, cybersecurity or other domain knowledge and can deploy technology inside real organizations.
Foundational programming remains essential. But market value increasingly comes from combining it with domain expertise, communication, English, systems thinking and accountability for outcomes. AI fluency becomes part of that bundle-not a separate profession, but a general productivity layer.
What business and policy should watch
Support for digital exports should not be judged only by the number of jobs created quickly. If those jobs are low-value and highly automatable, their growth may be short-lived. More important indicators are complexity of exported services, depth of client relationships, ownership of intellectual property, recurring revenue and how easily the service can be substituted by another low-cost provider or an AI system.
For companies, the practical question is direct: what share of current revenue comes from work that clients may soon obtain much more cheaply through AI? Where exposure is high, firms should use AI to raise their own productivity while moving their offer toward harder outcomes. Competitors cannot be prevented from using AI; the realistic strategy is to increase the value created with it.
Conclusion
Georgia’s digital-export growth is a major opportunity, but the country should not build its strategy around work whose main advantage is low price and easy remote execution. Kenya’s experience shows how quickly the lower rungs of digital outsourcing can shrink when AI reaches sufficient quality.
A more resilient path is to move toward complex software and engineering outcomes, cybersecurity, data systems, domain-specific technology integration, responsible AI deployment and productized software. AI will change these fields too, but there its best use is to multiply the productivity of Georgian specialists rather than simply compete with them. Georgia should ultimately export not cheap digital hours, but knowledge and outcomes that are difficult to replace.
Data and Main Sources
- National Bank of Georgia – Balance of Payments of Georgia, Q1 2026, June 30, 2026.
- Georgia’s National Statistics Office, Geostat – Information and Communication, Q2 2026 business-sector data.
- International Labour Organization – Generative AI and Jobs: A Refined Global Index of Occupational Exposure, May 20, 2025.
- The New York Times International – They Wrote Essays for a Living. Then A.I. Came., September 9, 2026.
Prepared by the academic team of Business and Technology University and the BTUAI Research Team, Tbilisi, Georgia.