The Most Dangerous Part of an Upgrade
What Lockheed Skunk Works, Microsoft, and the FAA can teach ABA providers about keeping critical systems running.
In 1988, I started my career at Lockheed's Skunk Works, working on the SR-71 and U-2. These planes were the fastest and highest-flying aircraft in the world, yet much of their underlying technology was already decades old.
What kept them flying wasn't newness. It was engineering discipline. We built redundancy into critical systems, maintained what worked, and upgraded components without taking the aircraft out of service any longer than necessary.
I've carried those lessons through nearly four decades in technology, including more than 25 years in healthcare.
And I thought about them again this week.
When New Technology Meets Old Infrastructure
On Thursday, Microsoft 365 experienced a major disruption affecting services including Outlook, Teams, and SharePoint. Microsoft's investigation pointed to infrastructure health and configuration issues, although the precise cause was still being investigated.
On Friday, the FAA halted departures nationwide for approximately five hours. Officials pointed to aging telecommunications equipment that had become overwhelmed as new technology was being introduced.
Two very different environments. Both dependent on enormous networks of interconnected systems. Both disrupted while their operators were managing infrastructure changes.
The FAA's preliminary explanation points directly to the risks of introducing new technology alongside old. Microsoft's disruption raises a related question about how changes propagate through systems that depend on one another.
Neither incident establishes that older technology is inherently unreliable. Nor does either suggest that modernization is a mistake.
But both illustrate something I've seen repeatedly throughout my career.
The most dangerous part of an upgrade may be the connection between what you're replacing and what you intend to keep.
What Didn't Fail Matters Too
The FAA's ground stop was enormously disruptive. But it was also an example of a system choosing a safer operating state when the infrastructure supporting normal operations became unreliable.
Flights didn't continue departing on the assumption that everything would probably work out.
Microsoft's disruption was different. Organizations found ways to continue some work through phone calls, text messages, and systems outside the affected environment. Other work had to wait.
Neither response eliminated the disruption. But they illustrate an important distinction between keeping a system available and keeping an organization operational.
That distinction matters in healthcare, where technology supports activities that cannot simply stop because a vendor is having a bad day.
ABA Has Already Lived This
When Change Healthcare was attacked in 2024, providers across healthcare lost access to critical claims processing infrastructure. For some, the resulting cash flow disruption lasted weeks.
ABA providers were not responsible for the failed infrastructure, but they still had to manage the consequences.
Some found alternative clearinghouse connections or other ways to keep claims moving. Others discovered just how dependent their operations had become on a single external service.
The lesson wasn't simply that providers needed a backup clearinghouse.
It was that a provider doesn't have to own the technology to own the consequences when it fails.
That applies to practice management systems, clinical documentation, scheduling, authorizations, claims, and the integrations connecting them.
And as providers adopt more specialized applications and AI-enabled workflows, the number of dependencies is likely to grow.
A modern technology stack can offer far more flexibility than the monolithic platforms many organizations have relied on for years. But every new connection also introduces another potential point of failure.
The answer isn't to avoid those connections.
It's to understand and manage them.
Redundancy Isn't the Same as Recoverability
At Skunk Works, redundancy was part of the engineering, not something we figured out after a component failed.
That principle applies to technology operations, but with an important distinction.
Two applications don't necessarily provide redundancy if both depend on the same cloud service, clearinghouse, or identity provider.
And having a backup isn't enough if nobody knows how to activate it, how long the transition will take, or what happens to the data created while the primary system is unavailable.
For an ABA provider, the questions are practical.
If the clearinghouse goes down, can claims move through another channel?
If the practice management platform is unavailable, can sessions still be documented and reconciled afterward?
If an integration fails, can the affected workflow be isolated without bringing down everything connected to it?
And when a vendor introduces a major upgrade, does the provider understand which operational dependencies could be affected?
These aren't necessarily reasons to build more technology internally. In many cases, they're reasons to ask better questions of the vendors already supporting the organization.
Modernization Without Losing Control
I've written recently about the capital flowing into ABA technology and the choices providers face as they build, rent, or outsource different capabilities.
I believe providers should take advantage of that innovation. Many have spent too long working around the limitations of older platforms.
And some of the innovation now emerging is too important to ignore.
We're beginning to see technologies that can passively capture clinical information, identify behavioral events, and potentially change how treatment is documented and measured. Digital phenotyping is one example. These capabilities could improve what providers know about the care they're delivering, reducing reliance on manual data collection and creating a richer longitudinal record.
But they also introduce new dependencies. Where does that information go? How does it connect to the clinical record? What happens when the technology is unavailable? And how does the organization distinguish information that has been validated from information that still requires clinical review?
These are not arguments against adopting new technology. They're reasons to think carefully about the architecture supporting it.
Modernization shouldn't mean exchanging one set of dependencies for another that nobody fully understands.
Data portability gives a provider options when it wants to change vendors.
Operational resilience gives it options when a vendor fails.
Both depend on understanding how information and work move across the organization, where critical dependencies exist, and what happens when one of those connections becomes unavailable.
That's why I would make availability and recovery planning part of every significant technology investment for 2027, not an exercise reserved for the IT department after implementation.
A vendor's feature roadmap matters. So do its integration capabilities, data access policies, change management practices, and recovery procedures.
And providers need their own operating plans for what happens when something stops working.
Engineer for Both
Nearly 40 years after I started working on aircraft designed in an earlier generation, I still think there's something valuable in that engineering philosophy.
We didn't stop innovating because the aircraft needed to remain operational. We engineered the changes so they could.
ABA providers face a similar challenge today. The next generation of technology promises capabilities that weren't imaginable even a few years ago. But adopting those capabilities shouldn't require putting the operation at risk.
The goal isn't to choose between innovation and reliability. It's to engineer for both.
If this analysis is useful, you can subscribe to receive related insights through the MissionViewpoint newsletter.
Prefer LinkedIn? Subscribe here.