Remember 2010? Everyone needed a mobile app - I was a student at the time, and I was getting at some point 2-3 project requests per week. The Apple App Store doubled to 300,000 apps in a single year, while the Android Market grew by 544%. Development costs ranged from $10,000 to $400,000+, but businesses justified these investments with optimistic projections.

Then reality hit. According to Gartner research, less than 0.01% of consumer mobile apps were considered financially successful. About 80-90% were abandoned after a single use, and two-thirds failed to reach 1,000 downloads in their first year.

2025-6: Same story, different technology

Today's AI agent market is following an eerily similar pattern. The market is growing from $5.4 billion in 2024 to $7.8 billion in 2025—with 85% of organisations integrating AI agents and 51% actively exploring implementation. Development costs? $10,000 to $500,000+. Sound familiar?

But here's the problem: AI agents have even more hidden costs than apps did. A mid-sized product with 1,000 daily users can burn through 5-10 million tokens monthly. Add infrastructure ($500-$2,000/month), maintenance (20-30% of build cost annually), and the ongoing challenge of managing model drift and prompt engineering, and costs spiral quickly.

The warning signs are already visible. Over 40% of agentic AI projects are at risk of cancellation by 2027 due to unclear ROI. While 78% of companies have adopted generative AI, only a fraction have scaled beyond pilot phases. Technical debt concerns are rising, with 75% of technology leaders anticipating significant debt by 2026 from AI integration.

The lesson is NOT avoid AI agents, but understand the real costs and have a concrete value proposition - a monetisation strategy for it. To save yourself from a crippling financial situation here are some things you need to consider: 

  • What's the total three-year cost, including tokens, infrastructure, and maintenance?
  • What problem are we solving that existing solutions can't? Can AI actually fix it, or is it just a marketing title?
  • Do you have expertise to manage ongoing optimisation and retraining? If not, how much will it cost to train or hire it?
  • What is the optimal application of AI to the problem being solved?
  • Can or are they willing users to pay the price? 

The businesses that succeeded in 2010 weren't those with the biggest budgets, but the ones that had clear strategies and realistic expectations - and some lucky exceptions. Enthusiasm is free, but custom development is expensive, and maintenance is forever. 

Sources for the numbers above:

  1. Gartner (2014): "Less Than 0.01 Percent of Consumer Mobile Apps Will Be Considered a Financial Success" - Mobile app failure rateshttps://www.gartner.com/en/newsroom/press-releases/2014-01-13-gartner-says-less-than-one-tenth-percent-of-consumer-mobile-apps-will-be-considered-a-financial-success-by-their-developers-through-2018
  2. Grand View Research: "AI Agents Market Size And Share | Industry Report, 2033" - AI agent market sizing and adoption statisticshttps://www.grandviewresearch.com/industry-analysis/ai-agents-market-report
  3. MarketsandMarkets (2025): "AI Agents Market by Agent Role, Offering, Agent System – Global Forecast to 2030" - Project risk analysis and market forecastshttps://www.marketsandmarkets.com/Market-Reports/ai-agents-market-15761548.html