Every Australian managing partner asking about legal AI eventually hears two names: Harvey and Claude. Harvey is the purpose-built legal AI platform used by large firms such as Allen & Overy and PwC for document review and litigation work. Claude is Anthropic's general-purpose model, available directly through Claude Cowork or the API, and also sits underneath parts of Harvey's own stack for certain workflows. The real question for a Sydney, Melbourne or Brisbane firm isn't which name sounds more legal. It's whether the firm wants a fixed legal workflow product or a flexible platform it can shape around its own precedents, matter types and risk appetite.
What Harvey Actually Does
Harvey packages document review, contract analysis, litigation research and drafting assistance into a single interface built for large corporate and litigation practices. It ships with legal-specific prompt templates, citation checking against case law databases, and integrations designed for BigLaw matter management systems. For a firm of two hundred lawyers running standardised due diligence across dozens of concurrent deals, that packaging genuinely saves setup time. The tradeoff is cost and rigidity. Harvey's enterprise pricing typically starts well above $100,000 a year for a mid-sized deployment, and the workflows are built around US and UK practice patterns that don't always map cleanly onto Australian conveyancing, family law or the Personal Property Securities Register.
Where Claude Fits Differently
Claude isn't a legal product. It's the model underneath a growing number of legal products, including parts of Harvey's own stack, plus every custom tool an Australian firm might build for itself. That distinction matters commercially. Instead of paying for a fixed legal workflow suite, a firm can use Claude directly to build exactly the review, drafting and research assistants it needs, trained on its own precedent library and firm style guide. A boutique commercial firm in Sydney with twelve partners doesn't need Harvey's litigation-scale document review engine. It needs a fast, accurate assistant that drafts first-pass commercial leases and flags unusual indemnity clauses against the firm's own precedents, which is a narrower and cheaper build.
Pricing model: Harvey sells enterprise seats with minimum contract values; Claude can be deployed per-use or as a fixed-fee build, which suits firms under thirty partners.
Customisation: Harvey's workflows are templated for large-firm litigation and M&A; a Claude-based tool can be built around a specific firm's precedent bank and matter mix.
Data handling: Harvey processes matter data through its own infrastructure; a Claude deployment can be scoped so client data stays inside the firm's own systems.
Speed to value: Harvey switches on faster if the workflow is already standard; a Claude build takes longer up front but fits the firm's actual practice rather than the other way around.
The Real Cost Comparison for an Australian Firm
Numbers vary by vendor negotiation, but the pattern holds. Harvey's published enterprise pricing puts a serious deployment at roughly $150,000 to $400,000 a year depending on seat count and modules, which only pencils out once a firm has enough deal flow to keep dozens of lawyers using it daily. A Claude-based build for a similarly sized Australian practice, commissioned as a fixed-fee project through a local consultancy, typically lands between $15,000 and $60,000 to build, plus ongoing API and hosting costs often under $2,000 a month for a firm of twenty to forty lawyers. This isn't a like-for-like feature comparison. It's a comparison of who is paying for what: a packaged product licence versus a purpose-built tool sized to the firm.
Data Residency, Privilege and the Privacy Act
For any Australian firm, the harder question sits behind the pricing table. Client matter data is subject to legal professional privilege and the Privacy Act 1988, and a partner needs a clear answer on where that data goes, who can access it, and whether it trains a third-party model. Harvey's enterprise terms address this at a platform level, but the firm inherits whatever architecture Harvey has chosen. A Claude-based build gives a firm more direct control over that architecture: which data touches the model, whether prompts and outputs are logged, and how long anything is retained. For firms handling family law, criminal defence or anything with heightened confidentiality obligations, that control is often the deciding factor over feature parity.
Which Firms Should Choose Which
The honest answer depends on scale and workflow standardisation, not brand preference.
Large firms with more than a hundred lawyers running high-volume, standardised due diligence or litigation document review: Harvey's packaged workflows are likely worth the licence cost.
Mid-sized and boutique Australian firms under fifty lawyers with practice-specific workflows: a Claude-based build sized to the firm's actual matters, often for a fraction of Harvey's enterprise pricing, tends to fit better.
Firms still deciding: a two-week paid pilot on one workflow, such as contract review, first-pass drafting or client intake triage, shows the real time saved before committing to either path.
There's no universal answer here, only a fit question. Before comparing price sheets, map the actual workflow you want automated. Book a session and we'll walk through what a Claude build would look like for your practice, sized honestly against what Harvey would cost for the same outcome.
