On Tuesday, $300 billion evaporated from software stocks. Adobe dropped 7.3%. Salesforce fell 6.9%. The catalyst was Anthropic releasing new coding tools, but the fear is older: something new can do the job of what these other things did, and so the value of the older thing is lessened. AI can now do what these companies charge $500 per seat for, and so the free market does its thing.
The same week, lawyers on LinkedIn are in a panic about ChatGPT replacing them. They have it backwards. The lawyers will adapt to use the tools, but generic legal software vendors are the ones in trouble.
The accountability problem
A contract review AI can scan documents faster than any paralegal. It can flag non-standard clauses, highlight risk language, and summarize a 200-page agreement in seconds. What it cannot do is sign its name at the bottom.
When the AI misses a liability cap buried in an amendment, someone has to explain that to the client. Someone has to carry malpractice insurance. Someone has to sit across from opposing counsel and decide whether to push back or fold. That someone has a law license and a reputation to protect.
AI has no skin in the game, but the lawyers do.
Two reasonable interpretations
Most real legal questions are not “what does the law say?” They are “which of two defensible readings helps my client more?”
A machine can give you Option A and Option B. It can rank them by how often each interpretation won in similar cases. But the machine does not know the judge assigned to your case tends to favor plain-language readings. It does not know opposing counsel has a reputation for bluffing. It does not know your client would rather settle for less than risk a public trial.
That is judgment, which requires context that lives outside the training data.
Where legal tech thrives
The legal software companies that will survive are the ones AI cannot easily replace. They fall into three categories:
1. Court-specific workflow tools. Filing requirements vary by jurisdiction. The Southern District of New York has different e-filing rules than the Northern District of California. A tool that tracks local deadlines, formats documents to local specifications, and integrates with PACER has value that a general-purpose AI does not replicate without significant customization.
2. Compliance monitoring for regulated industries. A hospital system needs software that tracks changes to HIPAA, state privacy laws, and CMS billing requirements simultaneously. The software needs to map those changes to internal policies and flag gaps. This is not a task you hand to ChatGPT and hope for the best.
3. Practice management for small firms. Solo practitioners and small firms need billing, calendaring, conflict checks, and client portals in one package. Clio, MyCase, and PracticePanther built businesses here because the alternative is duct-taping together five different tools. AI does not solve the integration problem.
The vendors in trouble are the ones selling document automation that a prompted AI can now match. If your product is “fill in the blanks on a standard NDA,” a la Turbotax, you have a problem. That case was always begging to be automated way.
The paralegal question
Paralegals and junior associates doing doc review will see their roles change. Some of those jobs will disappear. This is true.
But the number of documents that need review keeps growing. E-discovery volumes double roughly every two years. A firm that previously needed ten associates to review a document set might now need three associates supervising AI tools. The work per person changes. The total work does not shrink.
The junior lawyers who learn to use AI well will handle more matters, bill more efficiently, and make partner faster. The ones who refuse to adapt will struggle. This is the same pattern we saw with spreadsheets, search engines, and email.
Adaptation, not extinction
Law firms are already adjusting. Allen & Overy integrated Harvey (a GPT-4 based tool) into their practice in 2023. They did not fire their associates. They gave associates a tool that handles first-pass research so the associates can focus on analysis and strategy.
Latham & Watkins built an internal AI system for contract review. The system flags issues. Lawyers decide which issues matter. The division of labor is clear: speed from the machine, judgment from the human.
These firms are not worried about being replaced by AI, just about being outcompeted by other firms that use AI better.
Who should worry
The real casualties are not lawyers. They are the mid-tier software companies that built products a large language model can now approximate.
Adobe charges $55 per month for Acrobat Pro. A competent AI can now summarize PDFs, extract data, and redact sensitive information. Salesforce charges hundreds per user for a CRM that, increasingly, an AI agent can replicate with a spreadsheet and some API calls.
These companies have moats built on switching costs and enterprise contracts. Those moats are real, but they are not infinite. When the contract comes up for renewal, the CFO will ask why they are paying six figures for software an AI can handle.
Legal software is different because legal work requires accountability. Someone has to be responsible when things go wrong. Until an AI can be sued for malpractice, lawyers have job security.
The short version
Generic software that does generic tasks is in trouble. Specialized tools for specialized contexts are safer. And any job where humans bear responsibility for outcomes is safer still.
Lawyers will use AI the way they use Westlaw today: as a tool that makes research faster, not a replacement for the lawyer. The profession will have fewer entry-level research jobs and more leverage for experienced attorneys. That is a change, not an extinction. The $300 billion that vanished from software stocks is not coming back, and the legal profession will be fine.
In the end, this article isn’t about lawyers. Every field is going to face this panic, and every field will see routine processes automated away. Nuance requires human touch, and these same humans will also be piloting the AI powered software than will only make them better at their jobs.
As with other revolutions, the things that we thought would end just ended up being opened up for further expansion. AI is an opportunity to expand our frontiers.
Playing with Sharp Objects. 