AI and the legal profession: an evolving landscape

What Anthropic’s new legal solution signals for in house counsel 8 min read

In early February, Anthropic introduced a new legal plugin for its Cowork platform; one of the first signals that a major AI technology player sees the legal industry as a market worth explicitly targeting.

Anthropic is the company behind Claude, a consumer large language model known for its 'thoughtful' reasoning capabilities. While the new features are not solely aimed at legal users, its impact was felt most sharply in the legal technology sector. Since the plugin's release on 2 February, share prices of several leading legal AI technology providers, including Thomson Reuters and LexisNexis, have declined notably.

Beyond short‑term market volatility, the release poses a meaningful question for in-house leaders: how should general counsel interpret Anthropic’s entry into legal AI, and what, if anything, should change in their adoption of AI and longer‑term strategy for integrating AI into their legal operations?

What is this new development from Anthropic?

The plugin combines Claude’s language capabilities, similar to ChatGPT, with pre‑configured legal 'modules'. In short, it's a starter kit providing ready‑made workflows that can be installed directly into Claude Cowork.

As at 12 February, Anthropic indicates that the plugin supports the following capabilities:

• contract review: playbook‑driven clause review, deviation spotting, redline generation
• NDA triage: automated screening, classification, and routing recommendations
• compliance workflows: GDPR/CCPA checks, DPA review, data‑subject request handling
• template responses: category management, standard responses, escalation criteria
• legal risk assessment: risk classification and escalation models
• meeting briefings: structured preparation, context gathering, action tracking.

However, it's important to note these are early-stage features. Given the technical expertise required to deploy and operate them effectively, it may be challenging for corporate legal teams to derive tangible value from the Claude plugin in the short term, especially those without dedicated legal operations professionals or engineering support.

In our testing, once we navigated the initial limitation of Cowork requiring macOS (rather than Windows)¹, we found the functionality to be well-executed for an early‑stage offering. While there is some functional overlap with specialist legal AI platforms such as Legora, Harvey and CoCounsel, in our assessment, these specialised tools offer deeper, purpose‑built legal capabilities, a better user experience for lawyers and more robust enterprise‑grade data governance.

In a legal environment, user experience is critical and a key factor in whether a new solution will be adopted by lawyers. Our experience has shown that adoption is greatly enhanced when tools operate within lawyers’ existing workflows and meet them where they work (eg within Outlook and Word). There is still some way to go until Anthropic is able to deliver on these key areas.

Beyond market impact: understanding the current legal AI landscape

While the initial market response has been sharp, whether, and to what extent, this development will impact the legal industry in the longer term is more nuanced. Anthropic acknowledges that the plugin is not a deeply specialised legal AI function, but rather a general‑purpose AI adapted for legal tasks. This means that, while it facilitates broad reach into Claude Cowork's user base, it lacks the depth, nuance and domain‑specific safeguards of purpose‑built legal AI tools.

Over recent years, we've observed legal AI capabilities evolve from narrow, rules‑based or machine-learning tools designed to classify the relevance of documents or surface information into far more context‑aware systems that can support substantive legal work. Early systems delivered scale and speed but struggled with nuance, legal reasoning and variability across matters. Recent advances have materially improved how AI systems understand legal context, structure arguments, interact with source material and engage with users.

Making sense of legal AI tools

In practice, AI applications used within legal functions fall into several broad and overlapping categories.

1. General productivity assistance: general purpose AI assistants not designed specifically for law, but widely used for day‑to‑day productivity (eg Microsoft Copilot, ChatGPT or Claude).
2. Legal‑tailored productivity tools: large language model‑based tools tuned for legal tasks and outputs such as research, drafting and analysis (eg Legora, Harvey, CoCounsel and Lexis Protégé).
3. Domain or outcome‑specific legal tools: highly specialised solutions with narrow but deep capabilities that address a specific practice or need within the legal profession such as eDiscovery, contract review or document automation (tools such as Relativity, Nuix or Canopy).
4. Practice management and legal operations systems: platforms that support the business of law, including matter management, workflow tracking and time capture (eg Intapp, Smokeball AI and Laurel).
5. Agentic legal AI: systems that can take actions or make decisions in pursuit of defined goals, often across multiple steps, with limited human intervention (these systems are starting to take shape, in some way, across all the examples above).

While individual tools may span more than one category, the risk profile, reliance level and governance requirements differ significantly across them. Anthropic’s new Claude plugin sits primarily across categories (1) and (2). It takes a general‑purpose model and layers legal‑specific workflows on top. It does not yet compete directly with the depth or specialisation of category (2) and (3) tools, and it remains to be seen how it will perform longer term against the current leading legal AI tools as it evolves.

To date, specialist legal AI tools have had a clear advantage because they are built solely and specifically for legal practice and how lawyers work. This focus makes the tools easier for lawyers to learn, easier to use in their day to day, and better integrated into their work day. While Anthropic is likely to improve its product over time, unless it makes the strategic decision to focus sharply on legal functions, it will be difficult to replicate the depth and sophistication that a dedicated, fast‑moving specialised legal AI provider can deliver. The leading legal AI platforms also benefit from the flexibility to select the most suitable underlying model for each task, rather than being constrained to a single proprietary model.

Ecosystem impacts: lessons from Microsoft and eDiscovery

The entry of a major technology provider into a legal domain does not automatically translate into market dominance. A useful parallel is Microsoft’s introduction of eDiscovery capabilities within the Purview (Office 365) suite.

When Purview launched, some predicted it would displace well-established eDiscovery tools such as Relativity and Nuix. Instead, most organisations adopted a hybrid approach: using Microsoft’s tools for early‑stage or lower‑complexity matters, while continuing to rely on specialist platforms for larger, complex and higher‑stakes work.

Microsoft’s entry into this area made eDiscovery functionality more accessible to a broader range of users, but it did not eliminate the need for specialist tools. In fact, it prompted those providers to innovate and differentiate more aggressively.

A similar pattern is likely to emerge across the legal AI landscape.

What this means for corporate legal functions

A large, well‑funded AI technology company entering the legal space will almost certainly accelerate innovation and increase competitive pressure—however, it is unlikely to displace existing legal AI providers in the short to medium term. For legal departments, this is a broadly positive development as it increases the range of available tools and is likely to result in pricing and deployment models for established legal AI tools becoming more flexible.

For general counsel and legal operations leaders currently exploring tools such as Harvey, Legora, CoCounsel or Lexis Protégé, Anthropic’s Claude release should not prompt a dramatic change in approach.

Instead, it presents an opportunity to pause and reflect on several key principles when considering an investment in legal technology.

• Stay informed and adopt a healthy scepticism. Track developments closely but prioritise demonstrated outcomes over feature lists. Ask for evidence of measurable impact in real legal workflows. Consider general AI providers against specialist legal AI providers and evaluate the differences regularly. Revisit these assessments every 6-12 months as the market evolves and engage with your panel firms to learn from their experience and insights.
• Prioritise experimentation and 'learning by doing'. Use pilots and proofs of concept to test AI against clearly defined business problems and legal workflows. This helps quantify both value and limitations in your specific context. It also provides an opportunity for lawyers to build their capabilities by engaging directly with the technology. Experimenting with AI is the best way to understand its potential, as well as its limitations. Create space and permission for your legal team to experiment.
• Invest in solid foundations. The fundamentals remain unchanged: lawyer curiosity and digital fluency, well‑structured data and clear processes greatly enhance the likelihood of a successful legal AI tool deployment. Spend some time considering your valuable data sources, where they are located and how they are structured. Map the role of legal, key services delivered to the business and the key processes that are carried out on a regular basis. An AI tool, regardless of its functionality, will never compensate for weak foundations.