Stop Migrating Your Legacy Systems. Start Building Around Them.
Claude
Enterprises spend an estimated $40,000 per legacy system per year just to keep the lights on — and the most common prescription is to rip it all out and start over. That prescription is wrong. The enterprises that have figured this out aren't migrating faster. They're building smarter. And the surprising implication of getting this right is that you don't need to replace your legacy stack to eliminate its drag on the business. You need to build the workflows it was never designed to handle.
The Real Enemy Isn't the Old System
The mainframe in the basement is not what's killing your digital transformation. Neither is the ERP system that predates the iPhone. The real enemy is the organizational instinct to treat every legacy problem as a migration project.
Big-bang replacements carry enormous risk, take years, and fail at a rate that should embarrass the industry. And while your IT organization is absorbed in the migration — the requirements gathering, the vendor negotiations, the parallel-run testing — the development backlog that created the technical debt in the first place keeps growing. You finish the migration and discover you've moved the debt, not retired it.
Why the Migration Argument Isn't Unreasonable
Before making the case against wholesale replacement, let me give the other side its due. The argument for migration is not stupid.
Legacy systems create genuine security and compliance exposure: unpatched vulnerabilities, no modern authentication, integrations that were brittle when they were written and are worse now. According to Kissflow's enterprise modernization research, 59% of applications in an organization face technical and business-fit issues — outdated technology, scalability limitations, inefficient workflows. That's not a niche problem.
When a 30-year-old system is the single source of truth for a critical process, it holds the entire organization hostage. The conventional wisdom — you can't build a modern business on a crumbling foundation — is intuitive and not wrong in principle. CIOs who greenlight nine-figure modernization programs aren't reckless. The status quo has real, documented cost. I understand why they do it.
But understanding why the decision gets made doesn't make it the right one.
Argument 1: Migrations Replace Infrastructure, Not Process Debt
Here is what the migration roadmap almost never accounts for: the technical debt that actually slows enterprises down isn't in the codebase. It's in the workflows that were never built.
The average legacy migration takes two to five years. During that time, the business doesn't stop moving. It generates new manual processes, new shadow IT, new spreadsheet-based workarounds that fill the gaps the migration hasn't addressed yet. By the time the new system goes live, there's a fresh layer of process debt sitting on top of it — and now your team is too exhausted and over budget to tackle it.
Research on low-code and enterprise transformation consistently identifies the same bottleneck: the constraint in enterprise digital transformation is delivery speed and cross-department workflow gaps. Not server age. Not codebase language. The problem isn't that the old system exists — it's that new capabilities aren't being delivered fast enough around it.
A caveat worth naming: some legacy systems genuinely do need decommissioning. Infrastructure that's truly end-of-life, with no viable integration path and active security exposure, has to go. I'm not arguing against that. I'm arguing against the reflex that treats migration as the solution to what is fundamentally a development velocity problem.
The insight from Stop Automating Technical Debt: A Guide to Building Scalable AI-Native Apps is the right frame here: the goal isn't to stop building. It's to build on a foundation that doesn't create new debt with every release.
Argument 2: Low-Code Custom App Development Delivers Transformation Speed That Migrations Can't Match
The organizations closing the digital gap fastest are building custom applications in weeks, not migrating systems over years.
ToolJet's March 2026 analysis identifies the primary ways low-code accelerates transformation: speed of delivery, bridging the IT talent gap, and enabling cross-department innovation. Notice what's absent from that list — those are app development outcomes, not migration outcomes. The speed advantage comes from building capability, not from replacing infrastructure.
AstraZeneca reclaimed 30,000+ hours annually through workflow automation on the ServiceNow platform. That number didn't come from cutting over to a new ERP. It came from building the workflows that the existing systems couldn't execute. The value was in the gap between what the legacy stack could do and what the business actually needed — and the gap got closed by building, not by migrating.
ServiceNow App Engine has been recognized as a consecutive Leader six times in the Gartner Magic Quadrant for Enterprise Low-Code Application Platforms. That's not marketing positioning — it's the analyst community's signal that enterprise-grade app development on a low-code foundation is a mature, credible capability. The enterprises that have invested in that capability are not waiting for their migration projects to finish before they transform.
One genuine limit to this argument: low-code without governance creates its own form of debt. ToolJet's governance guidance makes the point directly — establish governance before you scale. Ungoverned citizen development produces a sprawl of disconnected apps that becomes the next generation's technical debt problem. The answer isn't to slow down development; it's to build governance into the development environment from the start.
Argument 3: Technical Debt Compounds When Apps Don't Share a Data Model
The deeper problem with legacy modernization isn't any individual system. It's that most enterprises have spent 20 years building point solutions that don't talk to each other.
You replace the old system with a new one, and six months later you have a new integration problem. Mendix's legacy modernization research frames this failure mode precisely: compatibility issues with new software are the specific thing that makes legacy replacement projects so painful. The old system's data doesn't map cleanly to the new one. The integrations break. You've traded one form of technical debt for another.
The structural answer to fragmentation debt isn't better apps in more silos. It's a unified platform with one architecture and one data model. When every application shares the same data foundation, you eliminate the integration layer that generates most of the ongoing maintenance cost. Adobe achieved 25% faster outage resolution on ServiceNow — that speed came from unified data across IT workflows, not from a new monitoring tool sitting on top of a fragmented stack. The NHL saw 87% of employees report higher productivity after moving to a unified platform. These aren't outcomes you achieve by replacing infrastructure. They're outcomes you achieve by connecting it.
There's an AI readiness dimension here that most transformation roadmaps are ignoring. AI agents can only act on clean, connected data. If your enterprise has fragmented systems and siloed data models, the AI layer you're planning to add will have nothing coherent to work with. Fragmentation doesn't just create operational debt — it makes autonomous operations architecturally impossible.
Argument 4: If Your Apps Aren't AI-Native, You're Already Building 2030's Legacy Systems
This is the argument I find myself making most often, and the one that gets the most pushback.
Every app your organization builds today on fragmented legacy infrastructure — and then tries to retrofit with AI capabilities later — is a legacy system in formation. The pattern is already visible: enterprises spent the last decade building cloud-native apps on top of on-premise systems, and now they're dealing with the hybrid complexity that created. The same dynamic is happening right now with AI.
ServiceNow AI Agents are autonomous systems that gather data, make decisions, and execute tasks. They require the underlying app architecture to support governance, audit trails, and cross-workflow data access. You cannot bolt those requirements onto an app that wasn't designed for them. Attempting to add AI on top of fragmented legacy infrastructure doesn't accelerate transformation — it generates technical debt at a scale that will take the next migration project to unwind.
ToolJet's 2026 analysis on AI in low-code identifies natural language-to-application generation and AI-assisted development as table stakes for new enterprise apps. That's not a future state — it's what modern low-code platforms deliver today. Building anything new that can't support these capabilities is, by definition, building something that will need to be replaced.
Gartner ranked ServiceNow #1 in the Building and Managing AI Agents Use Case. That recognition matters here because it signals that platform choice determines AI outcome. The platform you build on shapes what AI can actually do. Choosing a platform that can't support autonomous AI operations means your transformation investments are creating the exact technical debt you're trying to retire.
For a deeper look at the architecture behind AI-native apps — specifically how to avoid automating debt rather than eliminating it — Stop Automating Technical Debt is worth reading before your next architecture decision.
What Changes If This Argument Is Right
If the argument holds, several things have to change — not abstractly, but in specific decisions.
The digital transformation budget conversation has to be reframed. Less money to rip-and-replace infrastructure. More to building a governed custom app development capability on a unified platform. The measure of transformation success changes too: instead of counting how many legacy systems you've decommissioned, count how many manual workflows you've eliminated and how fast you're delivering new capabilities.
IT governance has to move upstream. Not a compliance checkpoint at deployment — governance built into the development environment from day one. This is non-negotiable if you're scaling low-code across business functions. The risk of ungoverned development is real, and the answer is architecture, not restriction.
Enterprise app development stops being an IT function and becomes a business function. The low-code model means business users build; IT governs the platform. That's a real cultural shift, and it's uncomfortable for organizations where IT has owned the development backlog for decades. But the backlog is exactly the problem — and it can't clear itself if every new request has to queue behind a migration project.
Organizations that don't build AI-native from the start will face a third wave of technical debt within five years. Not despite their transformation investments — because of them. The investment in infrastructure replacement without investment in a governed, AI-native development capability will produce exactly the fragmentation problem the migration was supposed to solve.
ServiceNow now serves more than 8,100 customers, including over 85% of the Fortune 500® — organizations that have made the platform their digital backbone precisely because it provides one architecture, one data model, and AI capabilities that scale without creating new debt. That's not an accident of market share. It's the outcome of the architectural bet described above.
The Question Worth Sitting With
If your digital transformation roadmap is organized around what you're replacing rather than what you're building, ask yourself who on that roadmap owns the answer to: What will we wish we had built three years from now?
If nobody owns that question, the migration will finish and the backlog will still be there.
The next step is seeing what that backlog looks like on a platform built not to repeat it. Start at www.servicenow.com.
