Every senior team I work with wants to start in the same place. The loudest broken process. The review cycle everyone is fighting about, the planning process that just failed publicly, the hiring loop that lost them a candidate they wanted. The instinct is completely understandable, and I almost always say no to it.
Not because the loud process does not matter. Because a redesign that lands on top of unstable foundations does not hold. It snaps back to baseline within two quarters, and then everyone concludes that redesign does not work, when what actually happened is that we built on sand.
So here is the sequence I run, in this order, in every engagement, whether it is three months or twelve.
First, diagnose. No engagement begins with a recommendation. It begins with a structured diagnosis using the same instruments and the same questions every time. What is the AI strategy actually trying to do. Where are the high-cost processes running unchanged. Where is the senior team genuinely aligned, versus where the org chart says they are aligned. What is the workforce being asked to absorb without the capability to absorb it. The named instrument I use here is the Workforce Redesign Readiness Scorecard for the AI Era, a red, yellow, green read across the dimensions that decide whether a redesign will survive contact with a real organization.
Second, prioritize. The diagnosis produces a ranked read, not a wish list. Not everything that could be improved, the specific things that need to land first.
Third, and this is the one people resist, fix the foundations first. Two foundations show up in almost every engagement before any specific process can be redesigned to hold. One is a senior team that is genuinely aligned on the operating-model decisions ahead, not aligned on a deck and privately disagreeing on the implementation. The other is a clear map of the high-cost processes themselves, because you cannot redesign a black box. Without those two, any process redesign collapses back within two quarters. With them, the redesign compounds.
Fourth, take on the biggest pain. Now we redesign the highest-cost, highest-friction process. This is where the senior team feels the change fastest and where the manager-hour relief is largest. The performance review process is my canonical example, because it touches everyone and because AI in the loop, used well, can genuinely change what it costs and what it produces. I am redesigning exactly this in my own organization right now, with a custom GPT that takes our own source material and produces the drafts a manager then refines. It is one of three redesigns I am running there at the moment. It is not a case from a binder. It is live work.
Fifth, the rest cascade in order. Once the biggest pain is redesigned against solid foundations, the adjacent processes either inherit the pattern, drop in priority, or surface as the next thing to do. The work compounds rather than restarting. This is why a six or twelve-month engagement is not just a longer three-month engagement. The engine runs again, deeper, on foundations that already hold.
The reason I require this order is that I watched the alternative fail, repeatedly, from inside a Fortune Global 500 for a decade. The pattern of failure is always the same. A team goes straight at the loud process, redesigns it well on its own terms, and watches it erode because the senior team was never actually aligned and the foundations underneath were never built. Foundations before flashy fixes is not a slogan. It is the difference between a redesign that holds and a redesign that becomes next year's cautionary tale.
If your AI strategy is live and you are deciding where to start, the honest answer is probably not where it hurts most. It is one layer underneath that. I can show you where, in a conversation.