Companies rein in AI usage as costs strain budgets
Companies that rushed to adopt generative AI are now capping usage and revisiting budgets as token-based pricing and heavy experimentation drive costs into the thousands of dollars per employee per month. Commenters describe a mix of FOMO, executive groupthink and overhyped expectations—often tied to hopes of replacing junior staff—as key drivers of overspending, while real productivity gains remain uneven and hard to link to the bottom line. Many see clear but limited value in AI for specialized tasks like coding, yet argue that misaligned incentives, poor evaluation methods and broader environmental and ethical concerns are pushing the technology far beyond its proven utility.
AI Costs vs Perceived Value
- Many argue AI isn’t yet worth its real cost for most corporate use cases, especially given models are sold below true delivery cost and providers aren’t profitable.
- Some think even $1k–$1.5k/employee/month is trivial if tools boost productivity a few percent; others note that if ROI were clear, budgets wouldn’t be cut.
- There’s concern that usage-based API pricing and coding agents have suddenly exploded costs compared to flat subscriptions.
Executive Incentives, FOMO, and Groupthink
- Several comments frame AI adoption as driven by fear of being “left behind,” peer pressure among executives, and hype similar to crypto, blockchain, metaverse, and VR.
- Some see CEOs as rationally betting on a 10–20 year horizon where AI must be central; others see this as naïve or “psychotic” overconfidence, rewarded regardless of outcomes.
Productivity, Workflow, and ROI
- Mixed reports: some engineers claim huge gains and heavy daily use; others say AI speeds non-critical work, floods backlogs with low-quality features, and creates support burdens.
- There’s a worry that metrics (tickets closed, LOC) show “productivity” without improving the bottom line, and that AI slop creates maintenance drag.
- Some companies have reduced per-engineer AI budgets, imposed audits for heavy users, or banned specific expensive models.
Employment Effects
- One cited study suggests AI adoption reduces junior hiring more than it causes layoffs; senior headcount stays mostly flat.
- Some workers now deliberately avoid over-documenting or over-optimizing, fearing they’re training systems to replace them.
Psychological and Cultural Dynamics
- Several posts describe “AI psychosis”: anthropomorphizing models, over-trusting outputs, and getting swept up in demo-driven hype.
- Others note polarized reactions: zealous AI boosters vs categorical opponents, with nuance drowned out by attention dynamics and marketing.
Governance, Deployment Models, and Future Paths
- Some expect a shift to on-prem or self-hosted models to control costs; smaller firms may stick to subscriptions.
- There’s debate over whether continued scaling will justify current investment or hit a wall, after which AI becomes “just another dev tool.”
Ethical, Environmental, and Class Concerns
- A significant subset opposes current AI largely on moral and environmental grounds: energy use, pollution, and perceived wealth concentration vs unmet social needs.
- Even some who see clear personal productivity gains feel conflicted, citing workplace pressure to adopt AI despite these concerns.