Over the past weeks, I've been looking at firms that collapsed during the digitization wave from the 1990s through the 2010s and those that adapted. Brobeck, Howrey, Coudert, and Thelen on one side and Cravath, Skadden, and Clifford Chance on the other. Same industry and moment with different outcomes.
The critical point is what they were actually facing, and how closely it maps to what leaders are now navigating with AI. Digitization was not a single innovation but a structural shift that reorganized how knowledge work was produced, delivered, and valued. It dismantled a model built on leverage where large numbers of junior professionals performed repeatable work under senior supervision.
At first, the changes appeared incremental as documents became digital, research moved online, discovery was automated, and communication and collaboration accelerated across locations. Over time, however, these shifts reordered the system, steadily eroding the repeatable work that had sustained the traditional pyramid structure.
What Changed and Why Outcomes Diverged
The disruption was not just to workflows, but to the economics and identity of professional work.
Work that had once sustained entire teams fragmented into smaller, visible tasks, many of which were absorbed by technology or shifted to lower-cost providers. At the same time, clients gained visibility into cost, firms rebuilt into more complex structures, and talent became increasingly mobile. Over time, power moved away from firms and toward clients and platforms.
Every firm operated inside this same structural reality, yet outcomes diverged because leaders understood the challenge differently.
Some treated the shift as technical. They invested in efficiency while assuming the model would hold. Others recognized that something deeper was changing and re-anchored value around judgment, trust, and client relationships. While others used the moment to expand, redesign delivery, or experiment with new models.
While there wasn't a single strategy, there was a shared discipline in that they diagnosed the nature of the disruption accurately and aligned their response to it.
The AI Mirror
That same pattern is now repeating with AI, and on a compressed timeline.
We can see a similar sequence, where the most repeatable work goes first as the production layer comes under pressure from systems that can draft, review, and analyze at speed and scale. Economic pressure follows, as clients begin to question pricing and staffing models. Additionally, identity questions, about what it means to be a professional when machines can perform core tasks, arrive far earlier than most organizations are prepared for.
What took thirty years under digitization is unfolding in five, removing the buffer that once allowed institutions to adapt gradually and forcing leaders to make decisions while the ground is shifting in real time, with far less room for error.
Diagnosis and Where Impact Lands
This is why diagnosis becomes the central leadership task.
Some aspects of AI are technical and can be solved with expertise and execution, but much of what organizations are encountering is adaptive. It requires changes in behavior, incentives, and what is valued, all while operating in a growing fog where leaders must act without a stable view of the end state.
Across organizations, the impact is not uniform but tends to concentrate in layers: repeatable, rules-based work is already being automated; analytical and creative work is beginning to hybridize, requiring new ways of structuring teams and maintaining accountability; and distinctly human work, judgment, persuasion, trust, and responsibility, not only remains but becomes more valuable.
The most common failure in this environment is misdiagnosis, whether treating adaptive work as technical or treating fog as predictability, both of which lead to responses that are misaligned with the nature of the challenge.
Acting Under Uncertainty
In many organizations, short-term stabilizing moves are being used to preserve runway, but they are often mistaken for a response to AI when they simply create time without building the capacity to adapt.
The real work is cultural. Leaders must shift from models anchored in production to ones anchored in judgment and trust. They must also operate in conditions where the goal is not to wait for clarity, but to stabilize enough to move, and to do so in ways that preserve learning and future options.
That means setting direction, even if provisional, and being explicit about what is known and unknown. It means focusing attention on what is deteriorating fastest, and judging each move by whether it builds trust, generates learning, and can be adapted as conditions change.
The Question for Leaders
AI is not introducing a new pattern but accelerating one we have already seen, and leaders who recognize that pattern, and understand the sequence rather than reacting to isolated symptoms, will have a meaningful advantage.
The difference will not be the disruption itself, but whether leaders can recognize it clearly enough to act at the level it requires.
For a deeper look at how to do this in practice, read Diagnosing Disruption: What the Legal Industry's Digital Reckoning Reveals About Leading in the AI Era.
A version of this essay first appeared on LinkedIn.