Enterprise IT + startup pilots (2023–2026 pattern) · 2025 · cost index 3/5
AI codebase rewrite pilots that shipped nothing
Six-figure LLM API and SI costs; production cutover: zero
What happened
Exec decks promised autonomous COBOL-to-microservices migration "by Q3," then rebranded as agentic modernization. Teams fed repos to copilots, generated plausible diffs, and discovered the model hallucinated APIs their mainframe never had. Pilots looked great in demos with 500-line samples; production needed tests nobody had.
Failure mode
Treating probabilistic codegen as accountability-free migration — whether branded "AI rewrite" or "agentic migration." No strangler plan, no parity tests, no owner for edge cases in the legacy module the model never saw. Skipping the hard organizational work with fresher slides.
Lesson
LLMs accelerate typing, not ownership. Use them inside a migration strategy you would trust with interns: small slices, verified behavior, humans on the hook for prod. Autopilot slides lie; diffs do not deploy themselves.
Sources
- Gartner — AI-assisted modernization cautions
- Industry pilot reports — LLM migration POC abandonment
- Joel Spolsky — Things You Should Never Do (still relevant)
Related tech
Practice · 2023–now
AI will rewrite the codebase
Big-bang rewrite, now with a chatbot — then with agents. Same failure mode; prettier slides.
$ Exec decks promised autonomous migrations "by next quarter," then rebranded as "agentic modernization." Teams that skipped tests, ownership, and incremental strangler patterns bought expensive demos and fragile diffs.
Practice · 2021–now
AI pair programming
Autocomplete that understands the file — now table stakes. Generation outran review capacity; ownership is the bottleneck.
Framework · 2023–now
LLM app frameworks
LangChain-class glue gave way to vendor SDKs and thin wrappers. The durable pieces are boring: evals, retrieval, and product UX.
Practice · 1995–now
Big-bang rewrite
Replace the working system instead of strangling it. A perennial money pit with new branding each decade.
$ "We'll freeze features and rewrite in the hot language." Multi-year projects shipped late, incomplete, or not at all — while competitors iterated on the ugly system that made money.
Architecture · 2012–2023
Microservices Everywhere
Independent deployability as religion. Correct for some; ruinous cargo-cult for most. Consolidation and monolith return are the 2020s epilogue.
$ Teams split 4-person products into 40 services, then hired SREs to babysit the mesh. Many later consolidated — after years of distributed tracing invoices and on-call trauma. See case studies on premature microservices and big-bang cutovers for named failure modes.
Language · 1995–now
Java
Write once, run everywhere — then enterprise forever. Still the backbone of serious backend estates.
Practice · 2024–now
Vibe coding
Accept AI output on feel, skip the diff, hope tests exist. Fast demos; CVEs and nobody owns the middleware.
$ Ship-to-prod without reading the diff became a flex until incidents revealed nobody knew which prompt wrote the auth middleware. Review debt arrived in one weekend demo.
Framework · 2023–now
LLM orchestration frameworks
Chains, tools, memory, and agents as a framework — LangChain-class glue sold as architecture. By 2026: LangChain-fatigue and thinner SDKs.
$ Teams adopted chain/agent frameworks before they had a single reliable tool call. Abstraction layers multiplied while prompt quality stayed flat.
Framework · 2023–2025
Multi-agent orchestration hype
LangGraph, CrewAI, AutoGPT cosplay — agents delegating to agents for tasks one function call could do. Hype peaked; backlash stuck.
$ Autonomous agent swarms burned API budgets on coordination loops that a single prompt and a cron job would have handled. Debugging "which agent lied" became the new on-call sport.