hn.today

Spending on AI Is Becoming Almost Impossible for Businesses to Budget

wsj.com57 points80 comments
Screenshot of Spending on AI Is Becoming Almost Impossible for Businesses to Budget

Commenters debated how businesses are struggling to budget token-based AI costs and who should bear the burden. Several argued companies are dumping model selection, cost monitoring, and agent management onto employees (hilariously, palmotea), producing confused usage and runaway bills (bunderbunder). Others recommended practical controls: per-person limits or tiered approvals (sajithdilshan, bluGill), clearer SOPs and training (sreekanth850), and choosing cheaper models or tuning prompts and context to reduce token burn (simianwords, awesan, 2OEH8eoCRo0). Some saw the problem as temporary: local models and on-prem hardware could reanchor budgets (thadt), while pessimists said unattended agents are inherently expensive and human-in-the-loop remains the only reliable pattern (rglover).

Opinion split over governance versus decentralization. A faction urged central harnesses and model-routing logic so engineering teams control who can use what and at what cost (FinnLobsien, Tanjreeve), warning vendor incentives and agentic workflows can obscure token drain (bunderbunder). Others countered that worrying about tokens is small compared with traditional personnel unpredictability (01284a7e) or that properly configured tooling already delivers acceptable value (bentt, awesan). Several commenters compared this to cloud adoption, predicting expertise will ultimately gatekeep broad access even as businesses wrestle with short-term chaos (Tanjreeve, solarNet).

Read on wsj.com80 comments on Hacker News

Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.

More in Other

The daily digest

Today's best Hacker News stories, summarized and screenshotted, one email a day.