No twelve-month transformation roadmap, no generic "AI readiness" framework. Four phases, scoped to one workflow at a time — whether the fix ends up being a process, a knowledge system, or an automation.
Not the process doc — the real thing. Who touches a ticket, claim, or job before it's resolved, what tools they use, and where it waits on a person instead of a system. This usually surfaces two or three handoffs nobody had written down.
Most workflows don't fail everywhere — they fail at one or two specific handoffs. This phase isolates those points and separates what's a process problem from what's a tooling problem, so the fix targets the right one.
Your CRM, helpdesk, ERP, or storefront — whatever's already load-bearing. The fix gets configured on top of that first; a new platform only gets proposed if the current one genuinely can't do the job. And occasionally the fix is fewer tools, not more — cutting a redundant subscription instead of adding one.
The build runs with a real team on real work before anyone talks about rolling it out wider. If it doesn't hold up under normal Tuesday conditions, it gets fixed or dropped — not defended.
The workflow changes. The discipline of mapping before building doesn't.