I've spent the last three posts describing Logic Entombment as an architecture problem. Technically, it is. But the reason it persists, the reason organizations keep deferring the work, isn't technical. It's financial. More specifically, it's that the cost is invisible until it isn't.
I wrote a few weeks ago about the hidden headcount in your AI bill, the idea that companies that replaced junior engineers with AI are now quietly hiring those engineers back to manage the AI, and that the labor cost didn't disappear, it just moved off the org chart and onto the infrastructure bill. This is the same dynamic, applied to architecture.
When you choose not to modernize your legacy systems, you're not avoiding a cost. You're deferring it. And deferred costs compound.
Here's what the math actually looks like. Every AI initiative you layer on top of entombed logic requires workarounds: manual data exports, screen scraping, brittle middleware, and custom integrations built for one use case that won't generalize to the next. Each of those workarounds has a build cost, a maintenance cost, and a fragility cost when it breaks. They also have an opportunity cost — every AI project that stalls because it hit an entombed wall is a business outcome that didn't happen.
None of this shows up as a line item labeled "cost of not modernizing." It shows up as slightly longer AI project timelines, slightly higher integration budgets, and slightly more engineering time on maintenance. The kind of costs that are easy to accept one at a time and hard to recognize as a pattern.
The comparison I keep coming back to is the monolith tax I wrote about in 2015. The argument then was that the industry was calling microservices expensive, which they are, while ignoring that the monolith has its own tax that compounds over time. The same thing is happening now with legacy modernization. Everyone talks about how expensive it is to change. Nobody is accounting for how expensive it is not to.
The organizations that are seeing real return from AI in 2026 are consistently the ones that did the infrastructure work first. Not because they're more technically sophisticated, but because they made the cost of entombed logic visible, put it on the balance sheet where it could be compared to the cost of fixing it, and made the investment decision honestly.
That's what good management of this problem actually looks like. Not a three-year architectural transformation program. Just an honest accounting of what the status quo is actually costing.
What does that look like in practice? A few things.
First, when you're scoping an AI initiative, account for the integration cost separately and explicitly. Don't let it hide inside "engineering time" as a rounding error. If your AI agent needs to call your pricing engine and that requires four months of integration work before the agent can do anything, that integration cost belongs next to the agent cost when you're presenting the business case. Make the debt visible.
Second, every time you build a workaround instead of a proper interface, write it down. Not as technical debt in a Jira ticket nobody reads — as a line in whatever financial tracking you do for your AI program. You're taking on a liability. Treat it like one.
Third, the modernization work pays compound returns. An API you build to support one AI initiative doesn't just serve that initiative — it serves every AI initiative that comes after it. The second agent that needs your pricing logic costs almost nothing to integrate because the door is already open. The math improves with each use case, which is the opposite of how the debt math works.
The door problem I described on Monday is real. But it persists because the people who own the budget don't see the full cost of leaving it closed. When you make that cost visible — when you put the real price of workarounds, stalled projects, and compounding integration debt next to the cost of proper modernization — the decision usually gets easier.
The work isn't glamorous. Building API wrappers around legacy systems doesn't make for a compelling board slide. But it's the difference between AI initiatives that work in a demo and AI initiatives that work in production.
Make the cost visible. Then make the investment.