Most migrations rot because they're sold as "let's see how it goes." Hardcut doesn't. Every phase has a definition of done. The cutover happens on a date you can put on a calendar.
Two weeks reading your infrastructure like a coroner.
Every workload, every dependency, every undocumented thing the last DevOps lead carried in their head. Out the other side: a written plan and a fixed price. The number comes after discovery, not before. Hardcut does not quote a migration it has not seen.
The new estate exists in a repo before it exists in production.
Terraform for the cloud surface. Ansible for the systems on top. Both versioned, peer-reviewed, and runnable by anyone with the credentials. No clickops, no console drift, no "the staging box was hand-built and nobody remembers how."
Failures move from the live event to the rehearsal, where they belong.
We rehearse the cutover against production data clones until the runbook is uninteresting. Every step timed. Every rollback verified. Every "what if X fails" answered before the night of.
One defined window. New system live. Old system quiet.
No parallel run, no eternal hybrid, no "we'll finish next quarter." A clean cut. Two weeks of post-cutover support follow: the runbook handed over, the team trained, the lights left on.
Thirty-minute call by default. No deck, no sales engineer, no follow-up sequence. Either there's a Hardcut-shaped engagement here or there isn't, and you'll know which one by the end of the call.