One hundred and fifty years ago, over half of the American laborers worked in agriculture. Today, that number is about 1.5%; a mere fraction despite the fact that our production is exponentially higher. As technology advanced, efficiency gains required fewer farmers and this transition enabled labor force to migrate from the soil to the assembly line, the laboratory, and eventually the abstract architecture of the information age.

Legal professionals (and most service providers, which now account for over ¾ of the U.S. workforce) now find themselves in a remarkably similar position. However, unlike previous technological shifts, the AI revolution is occurring at a comparatively unheard-of pace. As a result, the prevailing narrative in most service sector jobs is one of catastrophic AI cannibalization; SaaS is dead; margins will compress; talent eroded. But this cycle of expansion, efficiency development, and contraction is not a new phenomenon; it is a historical constant. As we’ve seen, what we consider the “tech sector” today didn’t exist in 1960, yet it now comprises approximately 6% of the U.S. workforce. There seems to be a tendency to focus on the destructive component of technological advancement rather than the fact that it catalyzes workforce expansion and diversification.

The fundamental difference today is the perceived threat to the law firm business model itself. How can law firms survive if productivity (measured in hours) declines? However, in the age of AI, we must differentiate between the volume of labor and the value of the output. Humans remain the bottleneck in high-stakes professional services. As I argued in Deep Moats, while the scope of transactions that will be handled by humans may be narrowing, the value proposition of that work is only increasing. If technology enables a lawyer to produce the same result in fewer hours, the value of that work is not diminished. We are entering the era of the High-Density Hour.

The Meter

The billable hour has prevailed as the primary measurement of value and the internal meter stick for productivity. I’ve had many conversations around whether AI will finally break that model and force a shift to project-based billing; one where clients will order services off a price sheet, like ordering an entrée and side dish. While I would certainly appreciate the freedom from reporting how I spend my day, this dream overlooks the primary function of the hourly rate: it is a risk-sharing mechanism. More importantly, I don’t believe either the firms or their clients have any true desire to abandon that.

For any service, unknown complexity makes fixed, value-based billing inherently risky for both parties. If a matter proves more complicated than anticipated, the provider carries the burden of uncompensated costs. If it is simpler, the client overpays. The “billable hour” persists because it allows both parties to share the risk of unintended complexities and efficiencies of a bespoke fact pattern. For commodified work (basic incorporations, reviewing NDAs, document review or diligence) where the scope is definitionally confined and predictable, the value has already started to erode. These are situations where templates are sufficient and research and summarization currently performed by humans should be automated. The scope of commodified work is expanding due to AI (hence the narrower moats), but we should embrace this transition since it accentuates the value proposition for human involvement in high stakes matters. In those situations, measurement by the hour will continue.

Yield Consistency

If a lawyer or firm can now spend 5 hours on a task that previously required 50 hours, does that make the product 90% less valuable? This would be the outcome only if the legal service derives its value from the volume of labor required to produce it. But risk and exposure are not divisible by time. No one would hire a lawyer that can quickly deliver advice if that advice turns out to be incorrect. Technology enables increased efficiency, but it does not inherently decrease the value of the service provided; to suggest otherwise implies that a lawyer is selling simply their time rather than their professional judgment and responsibility. Instead, what shifts is the density of value contained in that hour.

If a farmer produces 100 bushels of corn, the market pays for the 100 bushels regardless of whether the farmer employed ten farmhands or one automated tractor for the harvest. And the market is agnostic whether that harvest took ten hours or fifty. Similarly, if AI allows every software engineer to write lines of code twice as fast, the baseline expectation for productivity simply shifts upward. However, the “baseline value” of a functioning software feature does not drop by half; rather, the human required to guide the model and debug the output becomes a more critical, high-value resource given the inherent risk of a faulty product. The value of the yield remains constant.

AI is doing what technology has always done; it is reducing the “manual labor” component of law. Hours spent by a human professional are no longer “filler” necessary to achieve the objective. They are the high-concentration hours of strategic judgment and liability backstopping. Because the total value of the “legal yield” remains constant, and the hours required to produce it have contracted, the cost per hour will increase to restore an economic equilibrium.

The Training Paradox

The most significant impact of this technological shift will not be the death of the billable hour, but the massive shift in the leverage model. Traditionally, value of a service was often approximated by the number of bodies assigned to it since hours are constrained by each day and each person able to work during that period. Just as it no longer takes ten farmers to produce 100 bushels of corn, we will soon no longer need multiple attorneys to “produce” 100 hours of time on a matter. Instead, the math inverts. Fewer attorneys will be able to spend 50 hours to generate the same 100-hour output previously delivered by a larger team.

This causes a talent pipeline problem. Historically, junior attorneys were trained by performing the very pattern-recognition tasks that are now being commodified or automated by AI. Total leverage (the ratio of junior to senior attorneys) should decline because firms will no longer need a broad base of juniors to perform “review” work. The remaining humans capable of providing high-level intermediation will obviously remain a constrained and highly valuable resource. While reducing leverage necessary to produce a service would be accretive from a margin perspective, it also inhibits the development of the types of attorneys that are more valuable because of the higher-density hour.

If leverage can decrease from three- or four-to one to one-to-one, you could predict an environment where junior associates are paid one-third or one-fourth of their current salary until they can provide value commensurate with their cost. I doubt that is the outcome given that competitive nature of recruiting top-tier talent and the potential talent drain that would result. Instead, I suspect that the law firms that will thrive will accelerate learning programs to shorten the training window. The value of the high-density hour will incentivize firms to develop as many attorneys are possible to capitalize on it.

Technological Recursion

We should not expect the legal profession to be the sole exception to the historical pattern of technological evolution. While AI will inevitably commodify the routine labor of law it will also expand the market by lowering the friction of commerce. As the volume of complex activity increases, the demand for counsel capable of negotiating bespoke situations will intensify.

In this landscape, the “hour” is no longer a measure of effort, but a measure of risk managed and judgment applied. We should stop comparing the hourly output of tomorrow to the labor of yesterday; as the work becomes more concentrated, the value of that time will rise to reflect the weight of the responsibility being sold.

This transition reflects the core of the Ken Burns observation: while history doesn’t repeat itself, human nature does. Our innate need for a trusted, responsible arbiter of risk is an inelastic demand that technology, by its very nature, cannot fulfill. AI will fundamentally change the yield, but the responsibility for the harvest remains human.

Disclaimer: The views reflected in this piece are mine alone and are not made on behalf of any employer or client.

While legal AI undeniably is a force multiplier (the potential of which I encounter daily and I consider myself an advocate for early adoption of any AI-enabled gained efficiency possible), human-delivered advice isn’t dead yet (I’ll put a pin in what happens when the agents start needing lawyers). The value of human-derived counsel is concentrated into the following three moats and their value is increasing daily:

Judgment (Data vs. Wisdom)

AI models can summarize, benchmark and suggest revisions to a 200 page indenture or credit agreement. However, these models lack the ability to apply judgment comparable to a human that has spent a decade negotiating them and seeing how the sector has evolved; AI does a great job responding to specific text-based questions, but can fail to understand how certain provisions interact with each other. Nor are they currently able to navigate board-level politics or weigh potential reputational risk. AI will struggle to understand the humanistic motivations of a counterparty or the inclinations of a particular activist investor. I have no doubt that, someday, a model with a large enough context window, coupled with exhaustive iterative feedback and enrichment, can get there, but it will take an extreme effort (and willingness) from those with the relevant insights to digitize that institutional knowledge. We’ll begin to see automation adoption start lower in the value chain, but we are a long way before we let the models independently run a $2 billion bond offering or a $5 billion sell-side M&A transaction; the potential downside risk of a faulty tax analysis or a broken acquisition is far too great to outweigh any efficiencies gained from automation. Strategic risk assessment remains the domain of humans with context for human implications and consequences.

Creativity (Pattern Recognition vs. Pattern Breaking)

AI excels at pattern recognition, but I have not personally seen evidence of pattern breaking. This isn’t surprising – we don’t see this from junior associates either and not because they aren’t talented, bright and hardworking. I’ve been fortunate to work with many category-defining enterprises that disrupted legacy industries because the incumbents had grown innovatively sclerotic; these clients were bold enough to forge a new path where the current law never contemplated the technological advancement, scientific breakthrough or business application (while I wasn’t practicing during the dot-com era, I did witness the cloud migration, rise of SaaS, the emergence of blockchain and cryptocurrency and now the AI revolution). These entrepreneurs were willing to ignore legacy legal frameworks or exist just outside of them following carefully iterated discussions and brainstorming sessions. Perhaps the AI models will evolve on a similar trajectory that our junior attorneys do; after enough repetitions, they can demonstrate an ability to advise beyond their training materials. Again, however, I think this starts out in smaller transactional environments. What’s most likely to happen is that prior outputs become inputs (to borrow a phrase from Packy McCormick); as AI lowers the cost of what humans can initially build, the volume of complex matters arising as a result will actually increase. This is exactly where humans will need creativity and the ability to reason how existing legal frameworks (including general policy considerations) will impact the trailblazers that foster innovation.

Niche Specialization (Liability Backstop)

I fully expect rudimentary, mundane legal work to be commodified shortly (to the extent that it isn’t already). Frankly, this is not work that needs to be done at a large law firm and, to the extent it is currently handled by them, is routinely done by junior attorneys or paralegals. This is an area where large firms will need to adopt AI tools to drive efficiency if they want to retain this work (which is an open question; perhaps these AI-enabled capabilities aren’t part of the value proposition going forward). However, for the highest stake matters where the consequences for errors, omissions or inaccurate analysis dwarf any cost-efficiencies, law firms will still supply the human element as a liability backstop, for which they will charge a commensurate risk premium. I don’t see a world where anyone has the time or inclination to create a highly bespoke model to handle once-in-a-lifetime matters, or the willingness to provide any financial guarantee of accuracy. Perhaps this is really a derivative moat of the first two but there are enough differentiated and complex corporate transactions and litigation where the return won’t justify the investment.

AI-driven technology achievements will cause each of these moats to narrow; but as it does, I believe they get deeper and the value will continue to expand as long as human innovation persists on the exponential curve we’ve witnessed since the industrial revolution. Each decade, humans move up the value chain into more specialized functions. There will clearly be disruption in every field, including legal services, but this has always occurred. As an industry, we need to figure out how to embrace AI capabilities to realize efficiencies from a competitive dynamic, though I believe this adoption will ultimately be in service of each of the distinctly human advantages outlined above.

AI is amazing for automating efficiency. Excellent humans will have transformational tools to raise the ceiling on that excellence. But it would be a mistake to confuse the mathematician with the calculator or the scientist with the lab. As an industry, the survivors will figure out what their strengths are and will reinforce them through AI.

Disclaimer: The views reflected in this piece are mine alone and are not made on behalf of any employer or client.