
Technology Outages: Have We Learnt the Lessons?
The technology industry has become reasonably good at investigating major failures and explaining what went wrong. But the question is whether those lessons consistently translate into better resilience across the organizations that depend on its systems.
More than two years after CrowdStrike’s faulty software update disrupted airlines, hospitals and businesses worldwide, Microsoft’s global outage on 12 and 13 November 2026 left organizations and public services improvising once again. The technical causes differed, but the questions about shared dependencies, change control and business continuity were disconcertingly familiar.
In my own organization, email was unavailable for much of Friday, but Teams, phones, documents and internet access continued working. In parts of the UK National Health Service (NHS), difficulties communicating between clinical and administrative services presented more serious challenges.
The significance of a technology failure depends on the activities it interrupts, the consequences for those affected and the alternatives available until systems are restored.
The Warning Signs Were Already There
In FDM CCS Insight’s Senior Leadership IT Investment Survey, 2023, 53% of respondents indicated that their organizations would be critically affected immediately or within an hour of losing all workplace digital connectivity. Healthcare and financial services were among the sectors particularly exposed.
That scenario differs from the Microsoft outage, which affected specific services rather than connectivity itself. Nevertheless, the research highlighted how quickly disruption could move beyond IT operations into business activity.
Following the July 2024 CrowdStrike incident, my commentary raised concerns about insufficient visibility into third-party dependencies, software validation, change management and recovery. I argued that technology leaders needed to understand which business workflows were genuinely critical, how systems interacted and whether contingency arrangements could sustain operations when essential technology failed.
Our subsequent research reinforced these concerns. The 2025 study showed organizations reassessing cloud placement and repatriating some workloads for cost, compliance or operational reasons, without abandoning public cloud services. This was a considered adjustment to cloud strategies rather than a rejection of the public cloud.
By 2026, resilience and sovereignty were influencing more explicit operational decisions. Among the 1,001 senior decision-makers surveyed, 46% reported increasing cybersecurity, resilience or incident-response readiness in response to geopolitical, sovereignty or regulatory uncertainty. Another 37% were diversifying suppliers or hosting arrangements, while 36% were reviewing contingency, exit or switching plans.
These findings preceded November’s outage and show greater attention to dependency and control. Whether contingency reviews translate into tested operational readiness remains less clear. Two years after CrowdStrike, that distinction deserves considerably more scrutiny.
The Dependencies We Don’t Always See
Microsoft’s preliminary investigation attributed its outage to a software defect introduced in July, compounded by increased traffic following a routine Microsoft Teams failover and reduced server capacity. Automated recovery mechanisms subsequently contributed to cascading failures across infrastructure supporting Microsoft 365.
The troubling aspect is that measures intended to maintain availability could not contain the disruption. Shared infrastructure and software dependencies meant that additional capacity and failover arrangements could encounter the same underlying weakness.
This raises questions about the effectiveness of testing, fault isolation and recovery mechanisms within highly integrated technology platforms.
For customers, the concern extends beyond Microsoft’s engineering. Organizations may purchase applications without fully appreciating the shared infrastructure, services and operational dependencies that support them.
A recent discussion on The Fourcast podcast offered a telling example. A former NHS data engineering leader described how clinical teams frequently rely on locally developed spreadsheets and documents for activities such as discharge planning, bed management and waiting lists. These informal systems meet genuine operational needs but may remain poorly understood or documented at board level.
There is no suggestion that all such systems failed during Microsoft’s outage. More concerning is that it exposed how poorly understood essential workflows can remain. They depend on technology arrangements whose importance becomes apparent only when something goes wrong.
As AI and automation extend across enterprise systems, this problem could become harder to manage. Our research has found that automation is becoming more integrated while observability and assurance are not necessarily keeping pace. Organizations may become increasingly capable of automating work without gaining equivalent visibility into how failures propagate through those processes.
Sovereignty Raises the Stakes
Digital sovereignty introduces another dimension to this dependence.
In our July 2026 report, Digital Sovereignty: Risk-Based Control and Competitiveness, we argued that sovereignty should be treated as a proportionate operating discipline. Organizations must understand which assets and services require greater control, where dependency is acceptable and what happens if access or operational authority is weakened.
Microsoft’s outage demonstrates the technical dimension of that argument. Recent scrutiny of Nvidia, exports of advanced chips and US policymaking raises separate questions about geopolitical and institutional dependencies. The allegations surrounding Nvidia remain contested, and the two incidents are unrelated. Nevertheless, they illustrate different circumstances in which access to important technology can be affected by decisions or failures outside customers’ control.
Sovereignty therefore extends beyond where data is stored. It concerns who can administer, interrupt or restrict a service, which dependencies remain and what practical alternatives customers retain.
Yet greater sovereignty does not automatically guarantee better resilience. A locally operated platform can fail, while a global provider may offer security expertise, geographic redundancy and engineering resources that smaller organizations cannot realistically reproduce.
The commercial judgement is deciding where greater independence is justified, rather than assuming that eliminating foreign suppliers or duplicating every technology platform represents sound risk management.
Resilience Requires Engineering Discipline
My background in chemical and advanced control engineering makes one consequence of digital transformation uncomfortably familiar: the more connected an enterprise becomes, the further a failure can travel through workflows, suppliers and customers. Aviation, healthcare and other safety-critical industries have long relied on detailed investigations, tested safeguards and recovery procedures proportionate to the consequences of failure.
Serious incidents prompt detailed investigations into the sequence of events, the effectiveness of safeguards and the conditions that allowed a failure to spread. The consequences can be catastrophic, which explains the considerable investment in prevention, redundancy, containment and recovery. Digital enterprises increasingly need comparable discipline, even where the principal risks are interrupted services and lost income rather than physical harm.
CrowdStrike provides a useful historical comparison. Its July 2024 incident resulted from a defective security-content update that escaped validation and caused affected Windows computers to crash. CrowdStrike withdrew the problematic update relatively quickly, but restoring individual systems proved considerably more difficult.
After the incident, the company documented improvements to testing, staged deployment, customer control over update timing, and independent reviews. These were appropriate corrective measures and evidence of supplier-level learning.
Microsoft’s incident last week involved different technical weaknesses, but similarly exposed limitations in safeguards intended to prevent widespread disruption. Its preliminary investigation and subsequent corrective actions deserve examination against the same standard.
Supplier improvements address only part of the problem. Customers must understand the consequences of losing essential services, identify credible alternatives and test whether those alternatives can operate independently of the affected technology.
That might involve separate communications channels, accessible copies of critical information, alternative connectivity, manual procedures or additional infrastructure. The appropriate response depends on business criticality, cost and the likely consequences of interruption.
Routine applications need not carry the safeguards of bank transaction systems or critical hospital services. But a backup — inexpensive or otherwise — that depends on the failed platform is an odd form of reassurance.
Recovery Is Only Part of the Lesson
Restoring technology matters, but business continuity also requires clear communication about what has happened, which services remain available and how employees, customers and partners can continue essential activities.
A clinical system may remain functional while disrupted referrals or communications delay patient care. A financial institution may retain its core processing capabilities but struggle to complete transactions when supporting communications fail. In both cases, technical availability offers an incomplete measure of operational continuity.
It also raises questions about how suppliers demonstrate accountability. Post-incident reports should establish the causes, explain why safeguards failed and provide evidence of corrective action. Customers should revisit their own dependencies, recovery procedures and supplier assurance.
Our 2026 survey found that almost a third of respondents still relied on supplier self-assessment without independent audit when evaluating resilience. Recognizing the importance of resilience is one thing; demonstrating that it works under pressure is considerably more demanding.
There is a danger that the industry becomes better at producing incident reports than at converting their lessons into sustained changes in operating practice.
CrowdStrike’s corrective measures demonstrate supplier-level learning; Microsoft’s response must be assessed against its own engineering failures. The harder test is whether lessons from both incidents are changing customer practices and strengthening resilience beyond the individual supplier.
The Real Test Comes after Recovery
Technology outages will continue. Complex, interconnected systems cannot be made infallible, and attempting to eliminate every dependency would introduce substantial cost and complexity. That does not excuse avoidable weaknesses and inadequate preparation.
The question is whether organizations understand the consequences of disruption well enough to make informed decisions about where to invest, which dependencies to accept and which alternatives must work when called upon.
For technology suppliers, the responsibility is to demonstrate resilience through effective engineering, transparent investigation and independently credible assurance. For customers, the responsibility is to understand the practical consequences of failure and ensure continuity arrangements extend beyond the assurances of their principal providers.
The industry has had repeated reminders of these responsibilities. The test now is whether the lessons survive beyond service restoration.
After all, a successful recovery may end an incident. It does not necessarily mean an organization was adequately prepared for it.
