Kyndryl's first global Modernization Report, published this month, contains one number that should retire a decade of consulting orthodoxy: only 9% of the 2,000 leaders surveyed fully understand dependencies across their own applications. A quarter admit to large numbers of undocumented applications. Nearly half are behind schedule on modernization — and 98% say the delays have already hurt the business.
The Culture Excuse
For years we've been told transformations fail because people resist change. So we commissioned culture surveys, hired change managers, and ran town halls. Meanwhile the estate we claim to be transforming became a black box.
That isn't a soft-skills failure. It's a cartographic one. You cannot modernize what you cannot map. Every "rip and replace" on an undocumented stack is a blind bet; every migration is a surprise outage waiting for a long weekend.
AI Is Making It Worse
The report found AI has now overtaken legacy replacement as the top driver of modernization, and every surveyed mainframe user plans to deploy AI on that platform.
We're bolting the most complex technology we've ever operated onto systems whose dependencies we don't understand.
Adding complexity to an unmapped estate isn't transformation — it's decoration.
The Fix Isn't a Workshop
There's a signal worth watching. Agentic AI is being used for dependency mapping, code conversion, and documentation — the boring work — and early adopters report being ahead of schedule 67% of the time versus 51% for everyone else. But only 10% have it in production.
The takeaway: before the next culture initiative or AI pilot, fund the unglamorous project. Inventory the estate. Map the dependencies. Document the undocumented. Transformation is an inventory problem first and a people problem second — and pretending otherwise is the most expensive story we keep telling ourselves.