Apollo says AI has delivered zero financial payoff outside tech
An independent exercise, not a statement by the video creator or verification of every claim in the video.
The value of time saved differs from cash left in the bank
Assume a monthly example. Saving 100 hours at an assumed labor cost of KRW 30,000 per hour produces potential time value of KRW 3,000,000. If payroll or outsourcing payments do not actually fall and no additional revenue appears, this is not an equal cash saving. Monthly model costs of KRW 1,000,000 and additional review costs of KRW 500,000 make incremental cash flow −1,500,000 won. In a separate case where actual outsourcing spending of KRW 2,000,000 can be removed, subtracting the same costs improves monthly net cash by KRW 500,000. Quality, speed or revenue benefits from redeploying time need separate validation; no cash saving also does not mean no value of any kind.
Check for yourself
- Did saved time become lower actual payments or additional contribution profit?
- Were review, error repair, integration and training included?
- Was the pre-adoption alternative compared over the same period without counting savings twice?
Limits of application
100 hours, KRW 30,000, 3,000,000, 1,000,000, 500,000, −1,500,000, 2,000,000 and 500,000 are assumptions. Existing payroll is unchanged and review costs are assumed to be additional actual cash payments. Initial build costs, taxes, working capital and long-term effects are excluded; this does not estimate actual company or industry ROI.
- What to look for
- Examine results for enterprises paying for AI and for infrastructure suppliers using different metrics and time horizons.
- Why it matters
- Separate productivity, reported satisfaction and cash savings, and assess company-level payback through costs, review work and realized changes in returns.
Timeline
4 segmentsApollo compares the Magnificent 7 with the S&P 493; the latter still includes technology firms and is not identical to everything outside technology. Revenue per worker also changes with prices, demand and staffing, so it does not isolate AI’s causal productivity effect. Social-opinion polling likewise does not directly measure actual usage or corporate margins and must be treated as a different question.
