Article illustration 1

For decades, mid-sized companies faced a brutal choice in software strategy: endure ill-fitting off-the-shelf solutions or bankrupt themselves funding custom development. That binary is collapsing under AI's weight. As artificial intelligence dramatically reduces coding time and costs, it's triggering an economic phenomenon first observed in 19th-century coal markets—the Jevons Paradox—where efficiency gains paradoxically increase total demand. The result? A projected tripling of the custom software market to $146 billion by 2030.

The Efficiency Explosion

When Klarna replaced 1,200 SaaS tools with custom AI systems in 2024, the fintech giant didn't just save $40 million annually. It demonstrated a fundamental power shift:

"AI has collapsed the cost of building what you used to have to buy," notes Appunite CEO Karol Wojtaszek.


alt="Article illustration 2"
loading="lazy">

Klarna's AI assistant Kiki handles 2,000 internal inquiries daily while its customer service AI performs work equivalent to 700 employees—all running on proprietary infrastructure. This exemplifies AI's dual impact: **slashing development costs** while creating **self-reinforcing data moats**. Systems built in-house generate unique operational data, which trains better models, creating competitive advantages competitors can't purchase.

Mid-Market Liberation

Historically, companies between $50M-$500M in revenue were trapped:

  • Too complex for generic SaaS
  • Too resource-constrained for custom builds

AI-assisted engineering dissolves this constraint. Projects requiring $500K and 18 months now cost $200K with 6-month deliveries. As Wojtaszek observes: *"When development costs drop by 40-60%, the threshold for viable custom projects moves dramatically."* This unlocks pent-up demand—SME adoption of custom solutions is growing at 21% annually.

The New Development Calculus

Three structural shifts are emerging:

  1. Death of Hourly Billing: Compressed development timelines make time-and-materials pricing obsolete. Forward firms now sell outcome-based value—tying fees to business metrics rather than coding hours.

  2. Runtime Over Language: When LLMs generate competent code in any language, framework loyalty becomes irrelevant. What matters is the underlying runtime's resilience. As Wojtaszek notes: "Elixir's BEAM ecosystem delivers fault tolerance at scale—exactly what AI-generated systems need."*

  3. Proprietary Data Dominance: Custom systems' real value lies in generating unique training data. SaaS vendors dilute this advantage by aggregating insights across clients.

Strategic Imperative

The new build-vs-buy decision matrix favors custom development when:

  • Solutions require deep business context
  • Operations generate valuable proprietary data
  • Competitive differentiation is critical

Mid-sized companies now face a generational opportunity: Build unique systems that compound advantages or remain tethered to off-the-shelf tools while competitors pull ahead. As development economics tilt toward ownership, the question shifts from "Can we afford to build?" to "Can we afford not to?"

Source: Analysis based on Appunite CEO Karol Wojtaszek's blog post