Picture the first Monday of a newly carved-out company. Badges work. Employees log in. Payroll runs, the customer experience is the same as before and somewhere a programme manager breathes in relief.

By every measure on the cutover checklist, the separation succeeded.

Underneath that Monday, the new company is still threaded through its parent at every layer: the systems it runs on, the product it sells, the engineering organisation that builds it and the services business that delivers it to customers. Ownership has transferred. Independence has not.

Two jobs, one clock. We treat a carve-out as two jobs, not one. The first is craft: getting a company through separation seamlessly, on a clock that does not move. The second is opportunity: using the fact that separation must touch every system to rebuild the company around an AI-native core while those systems are already open. Most partners sell the first job and never mention the second. The argument of this piece is that the second is where the larger share of returns lives and that it is only available to teams who have made the first job routine.

The floor: exit the TSA on the date the deal model promised

The economics need no rehearsal. You have negotiated the schedules, you know what extension pricing does when the TSA clock runs out and you know that every month inside is fee paid out, dependency prolonged and management attention spent on the parent’s estate rather than the company’s future.

What separates exits that land from exits that slip is less discussed and it is not effort. It is whether the programme was designed backwards from the exit. Done right has a specific shape. Exit criteria written per TSA service during diligence, with evidence defined, so that “done” is a fact rather than a negotiation. A dependency map that includes the entanglements nobody puts on slides: shared service accounts, interface jobs, the data flows with no owner. Identity first, because everything else hangs off it. Wave gates that hold even when the steering committee wants good news. And a designed sequence, lift and shift now, modernise the moment the clock allows, chosen deliberately rather than arrived at in month 9.

The failure modes are familiar to anyone who has sat through a slipped exit: identity stranded between two estates, security controls lapsing in the gap, employees locked out on a Monday morning, an extension signed at the parent’s price. Avoiding that trouble is a craft and it is learned the only way craft is learned, by repetition.

This is where Persistent starts: the carve-out done well, over many transitions. Assessment before the deal signs, Day 1 stand-up (the first day of independent operation), application and data separation on the TSA clock, managed services for the standing company.

The repetition lives in delivery IP and carve-out templates: Migration Factory for industrialised moves, ExtenSURE for platform continuity, Persistent University so knowledge does not leave with the TSA.

Two recent stand-ups carry the pattern. A security-testing ISV carved out across 23 locations and 2,750 endpoints on a hard clock with zero disruption, roughly 19% efficiency uplift and 8% cumulative savings. A global publisher: 100+ engineers transitioned into a Professional Services CoE, on-time exit, 10% growth supported through the transition, technology debt avoided rather than transferred. Exits landed on or before the dates the deal models promised, with payroll running and customers undisturbed; the first promise the new company keeps, to its investors and to itself.

However, an on-time TSA exit is the floor of a good carve-out, not its ceiling.

The ceiling: an established company’s one founding moment

A carve-out is the only point in an established company’s life when its constraints are removed by contract. Architecture, operating model, technical debt and vendor estate all become decisions rather than inheritances. A NewCo can be designed AI-native from its first day, because separation touches every system anyway.

AI-native here means reinventing the operating model, not a chatbot on the service desk. Operations that self-heal before tickets are raised. Delivery that shifts left by default. A run organisation sized for automation rather than headcount. Data separated once and governed well enough for models to use from day one. This is the layer most separation partners never touch and the same bench builds it: SASVA, Persistent’s AI-led delivery platform, runs the programme itself; PiOps, our automation-led operations framework, runs the standing company.

The estate that had to be touched anyway comes back rebuilt for the AI era, not copied. In business terms, that is a lower run-rate from the first day of independence, savings that season into EBITDA well before exit and an operating model a buyer’s diligence can price.

The commercial logic is short. Run cost sits inside EBITDA, EBITDA is what gets multiplied at exit and the cheapest moment to remove cost is when every system is already open. Scoped as a separation, a carve-out produces an independent copy of the old company. Scoped as a founding, it produces a better one. Cost reduction and exit timing are what engineered independence pays; they are outcomes of the scoping decision, not substitutes for it.

Scope the carve-out as an exit and you will get an exit. Scope it as a founding and you get a company built for the AI-native world.

Next: Part 3, A Carve-Out Can Finish On Time and Still Lose Value