PwC partners with AI startup Harvey to launch a chatbot for its 4,000 lawyers, built using OpenAI's GPT-4, as law firms rush to find ways to integrate AI
Technology from start-up Harvey aims to make contract analysis and due diligence quicker — PwC has announced a chatbot experiment …
Context & Ripple Effects
This is PwC's first move in what became a deepening OpenAI relationship: within months it expanded from an internal tool into client-facing consulting on tax, legal and HR questions built on OpenAI models, then became the largest customer and first reseller of ChatGPT Enterprise across roughly 100K US and UK staff. For Harvey, the PwC deal was the credibility anchor that carried into an $80M Series B at a $715M valuation by year-end.
The deal lands amid a broader experiment wave: firms were already testing AI on work done by entry-level lawyers, with billable hours as the exposed metric. Harvey's later launch of its own proprietary legal model, Harvey Tenet, shows where this pilot arc ended up — rented OpenAI infrastructure first, owned models later.
First-order effects
- PwC's 4,000 lawyers get a working assistant for contract analysis and due diligence, making PwC one of the first Big Four firms to put generative AI into daily legal workflow rather than a sandbox.
- Harvey gains a marquee reference customer inside a global professional-services firm, validating its OpenAI-based approach to law-firm buyers.
Second-order effects
- Rival Big Four firms face pressure to match an internal AI capability that directly targets the hours-based economics of legal due diligence, pushing them toward their own vendor deals or builds.
- OpenAI secures a template for distributing GPT-4 through professional-services intermediaries — a channel PwC would later formalize by reselling ChatGPT Enterprise itself.
Third-order effects
- If entry-level legal work keeps migrating to tools like Harvey, law firm and advisory business models built on billable junior hours face structural repricing, with firms competing on oversight and judgment rather than document volume.
- The sequence points toward vertical AI vendors graduating from wrapping foundation models to owning their own trained models, shifting value capture from OpenAI toward domain-specific players like Harvey.
The trend: Professional-services firms are converting cautious internal AI pilots into enterprise-wide deployments and client-facing offerings, with legal work the leading edge.