AI Is Coming for Drudgery—But That’s Not the Whole Story
On February 3, Wall Street watched a handful of legal-tech stocks lose their footing after Anthropic rolled out agentic tools that can rip through contracts, NDAs, and compliance minutiae. Within hours, a nervous murmur about software became a loud forecast about layoffs. Two weeks later, The Washington Post’s Editorial Board stepped in with a counterpoint: the machines are targeting tedium, not people. It was a call for perspective, and a meaningful one—yet the subtext of the market’s panic tells us just as much about where white‑collar work is actually headed.
A familiar argument, on unfamiliar terrain
The Board’s thesis reaches back to earlier bouts of technological worry. In those cases, automation lifted the low-end tasks and the ceiling of what humans could do rose accordingly. This time, the stage is the office rather than the factory. Law is the right place to test the claim: the United States employs more than 1.3 million attorneys—about four per 1,000 people—an outlier among peers. That isn’t because Americans are uniquely litigious by nature; it’s because our rules and paperwork have accreted for a century. If AI trims the procedural thicket, the argument goes, legal professionals can spend less time crawling through documents and more time exercising judgment, the part of the job no one hires a robot to do.
There’s credibility here. Estimates that roughly forty‑odd percent of legal tasks are automatable don’t imply forty‑odd percent of lawyers vanish. They imply that the hours once absorbed by search, triage, and template work get compressed. When the meter stops running on repetition, volume rises. More disputes get tidy resolutions; more small firms get compliance right the first time; more individuals finally find affordable help. That’s the optimistic read—and markets, fixated on quarterly margins, rarely price it in on the first pass.
Tasks collapse, not titles—but titles will feel it
The legal profession is organized around a pyramid: juniors learn by doing the tedious parts, seniors monetize judgment. If software devours the bottom rungs, the ladder doesn’t disappear; it becomes steeper and stranger. Firms will need to invent training that doesn’t rely on grinding through discovery to teach young lawyers how to think. Apprenticeship will shift from “hours in the trenches” to supervised decision‑making with AI drafts as raw material. Billable hours, already under pressure, will give way to fixed fees and outcome pricing. The job description of “associate” won’t evaporate, but it will be rewritten around audit, verification, and client strategy instead of brute‑force review.
The sell‑off wasn’t about headcount—it was about moats
The February shock in information‑services stocks was less an omen of job loss than a referendum on defensibility. If agentic systems can read, reason, and act across repositories, then incumbent advantages tied to search, treatises, and structured datasets look less impregnable. Value migrates toward proprietary matter histories, workflow orchestration, guarantees about accuracy and liability, and deep integration with client systems. In other words, the profit pool shifts from selling the map to guiding the expedition. Markets overreacted, but they didn’t hallucinate the direction of travel.
This is why “AI comes for drudgery” undersells the disruption. It comes for the profit centers that monetize drudgery—document review benches, premium access to reference data, hourly billing premised on slow process. Those institutions can adapt, but the business models that grew fat on friction won’t make the transition unscathed.
Access expands—and so does activity
Cheaper routine legal work does not simply mean lawyers do less; it means more people can afford to do anything at all. The long tail of latent demand—startups incorporating cleanly, tenants challenging unlawful fees, local nonprofits navigating grants—enters the market. That expansion can be welfare‑improving while still creating headaches: when filings get cheaper, more filings appear; when compliance gets simpler, regulators tighten expectations. Productivity gains don’t end the story; they change the tempo.
Judgment is scarce, verification is costly
The limiting reagent in professional AI isn’t output; it’s trust. Drafts are easy. Consequences are hard. In high‑stakes contexts, humans will check the machine’s work, and that supervision eats into the apparent efficiency dividend. The frontier shifts from “can a system summarize this?” to “how do we prove this is correct, and who is on the hook if it isn’t?” Expect the premium to gather around audit trails, versioned reasoning, and insurance products that underwrite the risk. The bottleneck moves from generating text to certifying decisions, a transition that favors firms willing to build rigorous oversight into their workflows.
What changes next
In‑house teams will assemble stacks of agents tuned to their own clause libraries and regulatory footprints, then press outside counsel for speed and certainty rather than volume. Law firms will quietly retire line‑items that once billed for manual review and compete instead on playbooks and outcomes. Legal publishers will pivot from selling access to content to selling embedded answers with indemnities. Law schools will teach tool‑use and judgment side by side, because neither works without the other. And yes, some niches bloated by repetitive work will shrink.
The Post is right to deflate the panic. This isn’t a pink‑slip apocalypse. But “coming for drudgery” understates the stakes. AI is reassigning power inside the professions: away from those who controlled process, toward those who command judgment, distribution, and liability. If you want to know who prospers, follow who owns the verification step and who sits closest to the client’s real problem. The rest is just paperwork—and paperwork, at long last, has met its match.
