The Other Economy of AI

Building the New Economy

The distinction between a general model and a system adapted to the way a specific industry operates is where durable business value is likely to emerge. Andreessen Horowitz has argued that foundation models may become more like infrastructure than traditional software businesses, with greater value accruing to companies that build AI products around specific workflows, proprietary data and industry expertise. The investment flowing into legal AI, clinical documentation and industrial applications reflects this shift: investors are looking beyond the underlying models toward companies that can embed AI into valuable business processes.

The advantage of vertical AI comes from what it adds beyond the underlying model. Harvey, the legal AI company, is not competing by building another general-purpose language model. It applies AI to specific legal workflows, incorporating the documents, citation practices and professional standards that shape work inside law firms. The same principle applies in healthcare. A clinical documentation product that drafts a physician's notes from a recorded consultation creates value because it incorporates the coding rules, billing requirements and compliance obligations of the healthcare system in which it operates. The advantage lies in this industry-specific layer: the knowledge, processes and constraints that a general model does not provide and that European companies must navigate carefully as the EU AI Act introduces new requirements for high-risk AI deployments.

For much of the history of enterprise software, companies relied on broad horizontal platforms combined with specialised tools for particular functions. Those specialist products succeeded because they understood one job better than a general-purpose suite could. That dynamic is changing because the underlying capabilities of AI models are becoming increasingly accessible. The advantage is shifting from the model itself to how well the technology fits a specific workflow — something vertical AI systems are designed to provide.

For European firms, the implication is practical and demanding. The value will not come from acquiring the same models available to everyone else, but from the harder work of adapting those systems to real business processes. That means selecting tools that match the needs of a specific organisation and then redesigning how work is carried out around them. Companies that treat AI as a software purchase risk remaining among the majority of firms reporting limited financial impact. Those that treat it as an operational transformation are more likely to capture the gains. The next phase of AI adoption will be shaped by the companies willing to do the detailed implementation work that turns general technology into specific business value.