When “Right-Sizing” Meets Real Risk
Are Organizations Undermining Their Own Resilience?
Across industries, "right-sizing" has become a common way to describe a bigger shift. Companies are reducing headcount while leaning more heavily on AI. The appeal is obvious. Lower costs, faster delivery, and fewer layers of management.
But there is a question that does not get enough attention:
What happens to resilience when human oversight is reduced?
A recent situation puts this into focus.
A Real-World Stress Test
In early May 2026, Coinbase announced a workforce reduction of about 14 percent, roughly 700 employees. The move was part of a push toward an AI-native, "no manager" model. The company expected to save between $120 and $150 million annually and highlighted gains in productivity, noting that non-technical teams were now shipping production code.
Then, within days, things went sideways.
The platform experienced a major outage that lasted somewhere between five and seven hours. The root cause? A cooling failure at an AWS data center in Virginia. Multiple chillers failed, and some systems were affected despite built-in redundancy. The CEO described the outage as "unacceptable".
To be clear, the outage was tied to infrastructure, but the timing sparked a broader conversation:
When organizations reduce human oversight, they also reduce their ability to respond when things go wrong.
The Hidden Gap Between Plans and Reality
Most organizations already have resilience frameworks in place. Business continuity plans, disaster recovery, cyber response, emergency operations, and crisis communications are all familiar territory.
The issue is that these plans are often built for an organization that no longer exists.
When staff levels change or responsibilities shift to AI, key assumptions break down:
- Who is actually available to respond during an incident?
- Are there enough trained people to sustain operations during extended disruptions?
- Are decision roles still clear in flatter structures?
- Can AI systems recognize when something is outside their scope and escalate effectively?
- Is there still enough experience in the room to handle something unexpected?
If these questions are not addressed, even well-written plans can fall apart in practice.
AI Is Powerful - But It Isn’t Judgment
AI is excellent at speed, scale, and pattern recognition. It can automate workflows, flag anomalies, and even help predict issues before they happen.
What it doesn’t do well (at least not yet) is exercise judgment under pressure.
During complex incidents, people still provide what technology cannot fully replicate. They connect dots across systems, make decisions with incomplete information, and adapt when playbooks do not quite fit the situation.
If organizations reduce headcount without rethinking how resilience works, they risk creating environments that run smoothly most of the time but struggle when conditions change.
What the Coinbase Incident Really Highlights
The outage itself wasn't about layoffs. It was about infrastructure failure. But it exposed something more fundamental:
Resilience isn’t just about systems. It’s about people.
Even the best architectures will fail at some point. When they do, recovery depends on how quickly issues are detected, how effectively teams respond, and how well coordination happens under pressure.
If those human capabilities have been reduced, recovery becomes slower and more difficult.
Rethinking Resilience in an AI-Driven World
For companies moving toward AI-driven operations, resilience planning can't be an afterthought. It has to evolve alongside the operating model.
That means:
- Updating plans to reflect current staffing levels and skill sets
- Running simulations that assume fewer responders, especially during off-hours
- Clarifying decision-making authority in flatter structures
- Defining where AI supports response—and where humans must step in
- Identifying critical expertise and ensuring it isn't lost
- Reassessing dependencies on third parties and internal response capacity
In other words, resilience needs to be redesigned, not assumed.
A New Standard for Resilience
Ultimately, the question organizations must answer is no longer just "Are we insured to value?" It is "Are we prepared to recover?"
In today's environment, understanding how quickly a facility can realistically return to operation following a loss may be just as important as understanding what it would cost to rebuild. As equipment markets tighten, infrastructure demand grows, and supply chain pressures persist, lead times are becoming a defining component of risk.
Those who integrate this reality into their valuation and risk management strategies will be better equipped to navigate disruption and build true operational resilience.
It’s whether you still have the capacity to recover on a bad one.
About the Author
As a key member of the Risk Solutions practice, Jennifer Posten helps organizations anticipate disruption, respond with confidence, and recover with resilience.
With a comprehensive focus spanning operational resilience, continuity planning, IT disaster recovery, crisis management, and exercise design, Jennifer partners with both public- and private-sector organizations. They specialize in translating complex, modern risks into practical, actionable programs that align perfectly with regulatory expectations, day-to-day operational realities, and strategic leadership decision-making.