Rip and Replace Was Always a Fantasy. Here Is What Actually Works

Every few years our industry rediscovers the same bad idea. Tear out the legacy systems, drop in shiny new ones and emerge transformed. It almost never goes that way. The programs run over schedule, the budgets expand and the legacy estate systems you set out to replace are still quietly running half your operations three years later.

We are not going to sell you that fantasy again. I want to talk about what actually changed, because something genuinely did.

Agentic AI raised the bar and handed you a ladder

The AI you piloted last year primarily generated and summarized content. The AI showing up now acts. Agentic systems can reason across EHR, claims, billing and administrative systems and execute multi-step workflows autonomously. This represents the shift from AI-assisted to autonomous operations. CMS is openly pushing it for Medicare.

Here’s the catch and the opening. Agents are only as capable as the systems they can reach and most healthcare platforms were never built to let one in. But the same technology straining your legacy estate can help take it apart. Agents can sit between old and new systems as a translation layer, pull scattered data into one view and orchestrate the migration itself. IDC calls this the shift from modernization as a project to modernization that never really stops. The thing pressuring your old systems is also your fastest way out of them.

The playbook that works

Move away from the binary thinking neither full replacement nor stagnation works. The teams getting this right mix approaches:

• Migrate in phases, hardest-working systems first, so the business keeps running.
• Use middleware so old and new talk now, letting portals, EMR and RCM share data today.
• Move to cloud where the economics are obvious, not as a religion.
• Aim AI at the engineering: documentation, code conversion, testing, deployment.
• Rip and replace only where a modern package clearly wins.
• Keep a tight bench of partners who bring real scale and engineering muscle.

And if you’re a payer, stop treating FHIR as a box to check. CMS-0057-F forces those APIs live by 2027 regardless. Build them as genuine interoperability infrastructure and a mandate becomes an asset you reuse for years.

Where real productivity gains come from

Generic coding assistants are not a modernizing strategy. Tools like Copilot and CodeWhisperer are valuable but their effectiveness declines in the complex, poorly documented environments where modernization is hardest. This is where purpose-built platforms deliver impact.

Our SASVA™ platform combines generative and deterministic AI across the entire lifecycle, delivering:

• ~20-30% gains in early stages
• ~25-35% improvements mid-stage
• ~40% efficiency gains in late-stage modernization

These gains come from critical but often overlooked work: reverse-engineering undocumented systems, converting code, generating test cases and managing deployment.

Governance is what scales AI

One hard-earned lesson. The reason most agentic pilots never reach production isn’t capability. It’s trust. The moment an agent acts across clinical and administrative systems, your board, your clinicians and your regulators will want to know what it did and why. Build that in from the start: observability, auditability, traceability, explainability and real human control where it counts, backed by an enterprise architecture and governance function that sets the standard. Governance isn’t the brake on autonomy. It’s what lets you take your foot off it.

So here’s the honest ask. Pick your most painful legacy system, the one nobody wants to touch and make it your proof point this quarter. Modernize it in phases, let AI do the heavy lifting, build governance in. Prove it once and you’ll have the pattern and the credibility, to repeat it everywhere. The agents are already at the door. The only question is whether your systems will let them in.

If you want a partner that’s run this play across 125-plus healthcare and life sciences clients, that’s what we do.

Sources

• BCG and BCG X, How AI Agents and Tech Will Transform Health Care in 2026.
• Microsoft Industry Blog citing IDC research, modernizing regulated industries with cloud and agentic AI, 2026.
• CMS, Interoperability and Prior Authorization Final Rule (CMS-0057-F), FHIR APIs required live by January 1, 2027; and CMS leadership remarks promoting agentic AI for Medicare beneficiaries, HIMSS 2026.
• https://www.hl7.org/fhir/
• Persistent, Re(AI)magining Legacy Software in Healthcare with an AI-Driven Approach (2025), for SASVA capabilities and stage-by-stage productivity figures.