Vertical AI: Legal, Finance, Media and Services

Vertical AI is AI built for one industry's work: a legal assistant that drafts and checks contracts, a research tool that reads every filing a bank analyst needs, a customer-service agent that resolves a retailer's refund requests, a video model trained for film studios, a scribe that writes a clinician's note. The pitch is that general models are now good enough at reading, writing and reasoning that the bottleneck has moved to workflow, data, trust and liability, which are specific to each industry, and that the prize is not the software budget of a profession but its labour budget. Law firms, banks, contact centres, agencies and studios spend far more on people than on software, so a product that does part of the work can, in principle, charge for outcomes rather than seats. 2024–26 produced the first large vertical AI companies on that thesis: Harvey (legal, valued at ~$15.5 billion in September 2026, reported), Sierra (customer service, ~$15.8 billion, May 2026), Decagon, Hebbia and AlphaSense (finance research), Runway and ElevenLabs (media), and, in healthcare, Abridge (covered in the Healthcare knowledge base). It also produced the first large cautionary tales: lawyers sanctioned for citing cases the AI invented, Klarna's partial retreat from AI-only customer service, and record-label and studio lawsuits against generative media companies. A vertical AI business is therefore an answer to four questions: whose labour does it replace or augment, who is liable when it is wrong, what proprietary data or workflow does it own that a general model and the industry's incumbent software vendor do not, and how is the value it creates split between vendor and customer? This section covers legal, finance, customer service and media in depth, links to the Healthcare and Finance knowledge bases rather than repeating them, and compares incumbents with AI-natives, pricing models and the evidence so far.