Most firms cut the people and keep the legacy operating model. That sequence is backwards. Layoffs first, redesign later. You cut the institutional knowledge before you've figured out what the new operating model actually needs. The judgment that took years to build is the first thing the cuts hit. This isn't an argument against cuts. AI is shifting where leverage sits in every function. Some roles consolidate. Some grow. Some look unrecognizable a year out. The argument is about sequence.
Cuts come before the operating model is rebuilt.
Savings now.
The cost lands later.
Operating model is rebuilt before the cuts.
Comparable savings.
The multiple expands.
Same target on the board. One path accelerates the savings into the current quarter. The other lands comparable savings and compounds long-term margin upside through to exit.
A common pattern in AI-driven restructures: layoffs first, redesign later. Four reasons it tends to happen. None are about bad faith. All are about which cost is easiest to see on the way in. Not every restructure shows all four, but most show some.
Cost-cut framing tends to dominate. In a margin-compression cycle, headcount is often the first lever pulled, and senior operators sit in that line. The cut comes off the spreadsheet rather than the operating model. By the time someone asks what the team should look like next, the people who would have shaped that answer are sometimes already out.
Redesign can feel longer than the layoff. Cutting often lands inside a quarter. Redesigning the operating model, retraining the team, and rewiring the hiring scorecard usually takes two or three. It's natural to optimize for the action that closes this quarter. The cost of that choice tends to show up in the next two.
AI fluency sometimes gets miscoded as a new-hire problem. A common working assumption is that the legacy operator can't become the new one. In practice, senior leaders across product, engineering, and revenue ops with strong judgment often pick up AI workflow architecture inside 90 days when they choose to. The more frequent constraint tends to be institutional inertia about who owns the new shape, not the people themselves.
Savings can get spent twice. Severance, rehiring, and ramp often absorb a meaningful share of the payroll savings inside six months. What's left can be a team built for the workload AI is absorbing rather than the workload AI creates. Lean on paper. Less so in practice. A real drag on the value-creation thesis at exit.
What changes by function is what the new operating model actually needs. Here it is for product. Three tracks, each a different bet on the person's aptitude, career intent, and where the organization's immediate leverage actually sits. They are not interchangeable. The work is naming which track fits which person, deliberately, before the spreadsheet forces the call.
High-judgment product managers develop AI capability through a structured plan. You keep the institutional context while upgrading technical proficiency. The judgment was already there. The technical surface is teachable. Those who lean in become your strongest Product Operators because operator instinct compounds with new technical fluency.
A legacy product manager pairs with a builder who has the AI surface but not the customer judgment. They operate as a single unit until the product manager absorbs enough proficiency to consolidate the role. Fastest way to get a working team without starting from scratch.
Shift highly commercial product managers into value creation, partner, or strategy roles. Their strengths find more leverage there than inside a technical builder profile. Customer empathy and strategic communication are high-leverage assets. Buried inside the builder profile, they tend to flatten. Outside it, they compound.
A PE-backed SaaS company cut 40 percent of its product team in Q1 to hit a margin target. The math was right on the spreadsheet. The operating model behind the cut hadn't been rebuilt. There was no defined Product Operator profile to hire into, no track assigned per person, and no plan for the institutional knowledge walking out the door.
DiscoveryBy Q3, the new team, hired against the same legacy JD with younger candidates at lower comp, was producing roadmap drafts that didn't account for which customers had which contractual carve-outs. The CRO escalated three times. Two enterprise renewals churned over decisions the new team didn't know they were making. Severance, recruiting, and ramp absorbed most of the payroll savings inside six months. The margin play held for a quarter. The capability collapse hit the next four.
OutcomeThe fix wasn't rehiring. It was admitting the sequence had run ahead of the redesign and rebuilding the operating model and the institutional knowledge over 18 months at a cost higher than the original savings. The firms that run the sequence the other way around tend to land comparable savings and protect the multiple at exit. That's the version of this restructure that compounds.
Yes or no for the product organization on your bench. Take it as a CPO answering for your team, or as an operating partner pressure-testing a portfolio company before headcount decisions land. The questions are blunt by design. If the honest answer is yes-but, it's a no.
A real redesign isn't a memo. It's a set of named decisions, captured in a way that survives the next cycle. Three things define the work, and they all exist on paper before any name moves from the bench to a list.
What the team does, what AI does, where leverage sits a year out. The cuts are derived from the new model, not the other way around. No more cutting first and discovering the gaps in Q3.
Each senior operator gets one of three explicit tracks: retrain, pair, or reposition. The conversation has happened with each individual. Nobody is in a fourth, undefined track that resolves to severance by default.
The next hire is hired for the Product Operator role, not the legacy product manager. Comp band reflects the new shape. Institutional knowledge from anyone exiting is captured before the last day. The cycle stops compounding.
If a restructure is scheduled and the operating model hasn't been redrawn yet, I'd be glad to talk.