AI is in the plan. Is the work left behind accounted for?
“Reduce next year’s payroll. Introduce AI as well.” The planner adjusts hiring, trims departmental costs and adds software subscriptions. This is an illustrative scene, not evidence of a practice shared by most Korean companies. If that instruction arrives, one question matters: who will do the work previously assigned to the people leaving?
An agent may take it over. Existing staff may absorb it. A contractor may invoice for it. Or the work may no longer be needed. Those possibilities require different budgets. A 2027 workforce plan needs evidence that distinguishes them.
International planning signals show a funding and operating gap.
Gartner’s August 2026 commentary on its 2027 CIO and Technology Executive Survey reports expected average IT budget growth of 3.7%, alongside expected average growth of 31.8% in funding for agentic AI. These measure different budget categories. They are respondent expectations, not spending already delivered. Gartner highlights disconnects between AI ambitions and budgets, governance and workforce planning. [1]
Investment growth is not a replacement rate. A larger agent budget does not establish how many developers or planners can be removed. Integration, review and exception handling may add operating costs. Nor does one CIO survey describe every international company or industry.
Korean evidence points to automation and gaps in preparation.
In the Federation of Korean Industries’ survey of second-half 2026 investment plans, 43.7% of 106 responding companies from Korea’s 500 largest firms by revenue selected increased investment in task and production-process automation as a change in strategy associated with AI diffusion. AI-assisted R&D accounted for 20.8%, and entry into AI-based new businesses for 3.8%. The results point toward productivity and existing operations. They do not establish Korean companies’ final 2027 plans or intended workforce reductions. [2]
A Korea Chamber of Commerce and Industry study released in 2026 surveyed about 3,000 employees. Respondents reporting that their company had no generative AI adoption roadmap accounted for 54.4% in large companies and 70.4% in SMEs. These are employees’ reports, not an audit of official planning documents. Individual use of generative AI is also different from deploying agents across a business. The findings nevertheless raise a useful question about organizational preparation. [3]
| Observed signal | Question for the 2027 plan | What it does not establish |
|---|---|---|
| International CIO expectations of higher agent funding | Are operating costs, staffing and review arrangements funded? | A workforce replacement rate for all international companies |
| Automation priorities in Korean investment responses | Which tasks and processes will change? | Final Korean business plans for 2027 |
| Employee reports of missing AI roadmaps in Korea | Does tool use translate into redesigned workflows? | A causal link between missing roadmaps and poor performance |
These are different samples, periods and questions. They cannot rank countries or support a claim that foreign companies plan strategically while Korean companies simply follow orders. Manufacturing processes, software operations and financial review also have different constraints. This table translates the evidence into planning questions; it is the author’s interpretation.
International companies also confuse fewer staff with better returns.
A separate Gartner release in May 2026 reports workforce reductions at approximately 80% of organizations piloting or deploying autonomous business capabilities. Reduction rates were nearly equal among respondents reporting higher ROI and those reporting modest or negative outcomes. The underlying survey covered 350 global executives in the third quarter of 2025, at qualifying enterprises with at least USD 1 billion in annual revenue. It included agents, intelligent automation and other autonomous technologies. [4]
The findings do not prove that AI caused the reductions or that reductions caused failure. They do show why a lower headcount alone cannot establish better AI returns. Payroll and the organization’s ability to deliver need separate examination.
A payroll reduction target is an instruction. It is not evidence that an agent can do the work.
Specify when the plan must change.
When evidence is assembled to satisfy a reduction target already chosen, review work and failures can disappear from the presentation without disappearing from the business. This is the author’s critique of a planning practice, not a statistical claim about Korean companies. More rework, contractor spending or customer problems should trigger a review of staffing assumptions. The executives setting the target should also authorize those revision conditions.
An AI agent executes multiple steps toward a goal. Its operating scope needs validation in real work. The cited evidence does not establish the arrival date of AGI or superintelligence, or justify treating their workforce replacement capacity as a settled premise for a 2027 budget. Broader capability advances belong in separate scenarios. The base plan should reflect demonstrated capability.
Part II compares development, marketing, planning, HR and finance, then proposes evidence for decisions about hiring, redeployment and workforce budgets.
Sources and limitations
Evidence reviewed as of September 30, 2026. Source [1] concerns 2027 expectations; [2] concerns Korean investment plans for the second half of 2026; [3] surveys Korean employees; [4] is a 2026 release based on a 2025 survey. They are not a matched country comparison. A representative account of all Korean companies’ 2027 business plans was not established. The scene, table and planning questions are author proposals. No departmental replacement rate or causal employment effect is estimated.