Last time, I wrote that enterprise software was built to tell people how to do their jobs. That's still true — companies still buy software to instruct their humans. But it's less true every quarter, because the humans doing the instructed tasks are increasingly not humans.
Which brings me to the forklift.
For decades, the playbook for aging systems was the forklift upgrade: rip out the old system, drop in a newer version of the same kind of system. A trillion-dollar IT services industry — that's roughly the size of the global market — got paid handsomely to move you from old ERP to new ERP, old CRM to new CRM. Same workflows, new plumbing. SAP's own deadline tells the story: mainstream ECC support ends in 2027, and the default migration path is literally called "lift and shift." Most of these projects still blow their budgets and timelines. You paid tens of millions to end up where you started, with fresher paint.
That era is ending, and not gently.
Here's the new test for your workhorse systems: either they perform in a world where AI does the idiotic tasks — the rekeying, the reconciling, the swivel-chair work between screens — or they don't. There is no cosmetic option. You cannot put lipstick on a system of record whose entire design assumption was an army of people typing into it.
Private equity sees this first because it has to. The average PE portfolio company has now been held 4.4 years — a twenty-year high — and roughly half of all buyout-backed companies have passed the four-year mark, the highest share on record. At the current exit pace, the industry needs about a decade to clear its inventory. Meanwhile, only about one in five portfolio companies has turned AI into measurable returns. That's thousands of sleepy companies, sitting on tired systems, losing value by the hour while their owners need them exit-ready.
So the migration path changes shape. In the old world, you hired an integrator to move you to a like system. In the new world, the winning move is different: look at the shape of your data across all your systems — the ERP, the CRM, the spreadsheets duct-taping them together — build a spine, and move the workflows on top of it. You're not migrating from System A to System B. You're migrating from systems to shape.
Is this the biggest, fastest software and data migration in history? I think it will be. The cloud migration took fifteen years and mostly moved workloads without changing them. This one changes what the workloads are — and the 2027-style deadlines, the PE exit backlog, and the economics of AI are all pushing in the same direction at once.
The companies that treated their workhorse systems as a problem for next year's budget are about to find out what "left in the dust" feels like. The ones that start with the shape of their data — not the logo on the software — get to skip the last forklift upgrade ever sold.