What happens when legal AI becomes buildable?
On 28 October, Anthropic introduced a gallery of ready-to-use artifacts for legal work in Claude. The examples include citation checks, contract comparisons, matter boards, regulatory horizon scans and tools for reviewing contract terms. Teams can adapt them with their own data and connectors, share them across the business, and follow links from outputs back to sources. Anthropic says a lawyer applies judgment to Claude's first pass.
The interesting part is not the gallery itself.
It is how much shorter the distance is becoming between a general AI model and a working legal tool.
A team can now start with an example, connect its own material and adapt the workflow without beginning with a blank page. Anthropic also says Claude for Legal has more than 60 legal connectors. That makes the platform around the model part of the story: the place where work happens, the systems it can reach, and the people with whom a result can be shared.
The pressure on expensive legal technology
This creates a difficult question for legal technology sold at a premium.
If a product's main value is a general-purpose model wrapped in a focused interface, a set of prompts and a standard workflow, a broader AI platform may make that package easier to reproduce. Buyers will ask what they are paying for — and whether the specialist product still earns its price when a similar first pass can be built closer to the tools they already use.
That does not mean legal software is about to disappear, or that every legal workflow can be recreated inside a general AI platform. It means that the easy-to-copy parts of a product are less likely to support a high price on their own.
The stronger cases for a specialist product will be harder to reproduce: trusted legal content, strong coverage of a jurisdiction, dependable links to matter and document systems, tested performance on real work, audit trails, access controls, privacy safeguards, and clear responsibility when something goes wrong. A product can also create value by fitting deeply into a firm's actual process and helping people carry work through to a reliable result.
The question for a vendor becomes sharper:
What does the customer receive beyond access to a capable model and a useful interface?
That answer may include expertise, data, integrations, controls, support and measurable outcomes. The vendor needs to show which of those are distinctive and why they matter enough to justify the price.
The platform may own the work surface
There is also a distribution question. If lawyers and business teams can access legal tools from a general AI workspace, specialist products may have to become part of that workspace, or offer enough value for users to leave it deliberately.
Anthropic's announcement points in both directions. It presents artifacts built with legal technology companies and law firms, and it emphasizes connectors to systems legal teams already use. That suggests an ecosystem where specialist products can remain important while being reached through broader AI environments.
For legal tech companies, the strategic choice may not simply be whether to compete with the model provider. It may be where they sit in the customer's workflow: as the system of record, the source of trusted legal intelligence, the specialist execution layer, or some combination.
What a law firm should do
A law firm should start by mapping the work it wants to improve.
Which steps repeat? Which can be prepared from documents and established rules? Where does a lawyer need to interpret context, choose a position, advise a client or approve an action? What must be checked before a result leaves the firm?
That map gives a firm a practical way to evaluate tools. A pilot can be measured against the current process: time, quality, missed issues, correction effort, client experience and total cost. The firm can then decide whether a broad platform, a specialist tool or a combination works best.
The firm also needs to decide how it will explain the economics to clients. If routine work takes less time, clients may expect faster answers, lower costs or a clearer connection between fees and value. The opportunity is not simply to do the same work faster. It is to use saved capacity for better judgment, more responsive advice and work that clients value.
What changes for an individual lawyer
As AI takes on more first-pass work, the lawyer's contribution becomes more visible in the review: understanding what matters, spotting exceptions, checking sources, recognizing weak assumptions and deciding what advice fits the client's situation.
A link to a source makes an answer easier to inspect. It does not prove that the source supports the conclusion, that the analysis is complete, or that the right question was asked.
That makes verification a core skill. Lawyers will also need to describe a process clearly enough for a system to assist with it, know when the system is outside its lane, and explain the result in a way a client can act on.
The profession's value does not vanish when a first pass becomes cheaper. But the value has to be visible in the quality of the decisions, the advice and the responsibility that follow.
What changed
I have tended to think of legal AI as a collection of products: research, contract review, drafting, matter management. This announcement makes me focus more on the work surface around those products.
The boundary between a general AI workspace and a purpose-built legal tool is becoming easier to cross. That raises the standard for every product that asks customers to pay a premium: it must make clear whether its lasting value lies in legal expertise, data, workflow integration, reliability, governance or results.
What I think matters
The next phase of legal AI will be decided less by who can demonstrate an impressive first draft and more by who can make a whole workflow dependable.
For legal technology vendors, that means proving the value that cannot be copied by a quick configuration. For firms, it means redesigning work around defined human decisions and measurable quality. For lawyers, it means becoming excellent at the parts of the work where context, judgment and accountability matter most.
What remains uncertain
Anthropic's announcement describes a product direction. It does not independently establish how accurate the artifacts are, what time or cost savings they produce, or how well they perform on Norwegian legal work. It also does not show that every function or connector is available to every team.
Those questions should be answered through careful testing on real work, with clear controls and a way to compare the result with the existing process.
Sources
- Introducing artifacts for legal work, Anthropic, 28 October 2026. The product capabilities above are described in Anthropic's announcement; the commercial and professional implications are my analysis.