Imagine I sold you a website today and built it on Adobe Flash. You would look at it, realize it does not work on any modern device, and you would be right to walk. Flash is dead technology.
Now imagine I sell you an AI workflow built on last year's model. It works. It answers questions. It processes data. But in six months, a newer release is the standard. Your competitor is using it. Their system is ten times faster, handles a two million token context window, and costs half as much to run. Your system is slow, dumb, and expensive by comparison.
Technically, no. I delivered exactly what we scoped. I built it using the best tech available at the time. Practically? I locked you into a disadvantage.
This is the Standard Tech Instant Crisis. In traditional software, legacy meant systems that were ten or fifteen years old. Mainframes. On-premise servers. COBOL. In AI, legacy takes ten weeks.
This creates a new ethical obligation for service providers. If your contract does not explicitly state how and when the underlying models will be upgraded, you are selling a lemon.
We have to stop pretending that the build phase is the end. The build is the entry ticket. The real value is in the upgrade.
Clients need to ask one question. When a new flagship model releases, is my system upgraded automatically, or do I have to pay you another twenty thousand to rewrite the prompts?
If the answer is pay me, run. You are buying Flash in 2026.