The State of AI: What Is Actually Happening Across Healthcare, Finance, and Beyond
Artificial intelligence is no longer a technology category that requires qualification with words like “emerging” or “potential.” It is operational infrastructure across healthcare, finance, manufacturing, media, and professional services — and the pace of capability development is accelerating. The question for businesses, investors, and individuals is not whether AI will be consequential but how to navigate a landscape where the applications are proliferating faster than the frameworks for understanding them. This piece maps the most significant application areas, the underlying technological shifts driving them, and the structural risks that deserve more attention than they typically receive.
- → Healthcare AI has moved from research to deployment — diagnostic models now match or exceed specialist performance on specific tasks, and the bottleneck is regulatory approval and integration, not technical capability
- → Multi-modal AI — systems that process text, images, audio, and video simultaneously — is the architectural shift that makes AI genuinely useful across complex real-world tasks rather than narrow benchmarks
- → In financial services, fraud detection and algorithmic trading are mature AI applications; the frontier is real-time risk modelling and AI-driven regulatory compliance
- → Generative AI has restructured the economics of content production — the marginal cost of text, image, audio, and video generation is approaching zero, which is transforming media, marketing, and software development
- → The critical constraints are not technical but institutional: data quality, regulatory frameworks, workforce adaptation, and the alignment of AI capabilities with human oversight structures
Healthcare: From Proof of Concept to Clinical Deployment
Medical AI has crossed the threshold from academic benchmarks into clinical practice. Radiology is the most advanced front: AI diagnostic systems for chest X-rays, mammograms, and CT scans now perform at or above the level of specialist radiologists on specific tasks. Google’s DeepMind demonstrated AI detection of over 50 eye diseases from OCT scans with a diagnostic accuracy comparable to world-leading specialists. Paige.AI received FDA approval for AI-assisted prostate cancer detection. These are not projections — they are deployed systems generating clinical outcomes today.
The more transformative near-term application is personalised treatment planning. AI systems can analyse genomic data, patient history, and real-world evidence databases to identify optimal treatment protocols faster and more comprehensively than any individual clinician. For oncology, where the right drug-mutation match determines survival, this represents a genuine capability step change. The bottleneck is not the AI — it is the integration of these systems into clinical workflows and the regulatory pathways that govern their use.
The most under-appreciated AI application in healthcare is not diagnostics but drug discovery. AI platforms like AlphaFold have solved the protein structure prediction problem that stumped biochemistry for 50 years. The implications for pharmaceutical development — identifying drug targets, predicting binding efficacy, designing novel molecules — are only beginning to materialise in clinical pipelines.
Multi-Modal AI: The Architecture That Changes Everything
The shift from single-modality AI (systems that process one input type) to multi-modal AI (systems that process text, images, audio, video, and structured data together) is the most significant architectural development in applied AI. Earlier systems were useful but narrow: a language model that could not see, an image classifier that could not reason. Current frontier models — GPT-4o, Gemini, Claude — process multiple input types simultaneously and reason across them.
The practical implications are substantial. A multi-modal system can watch a manufacturing process via video feed and simultaneously read equipment telemetry and maintenance logs to predict failures before they occur. A medical AI can simultaneously read a patient’s chart, examine a scan, and listen to a physician’s spoken notes. An educational system can watch a student’s facial expressions and response patterns to adapt instruction in real time. These are qualitatively different capabilities from anything available five years ago.
The marginal cost of generating a 1,000-word article, a photorealistic image, a 60-second video, or a two-minute audio segment is now measured in cents. This is a structural economic shift that affects every industry producing information-based goods. For media, marketing, and advertising, the implications are existential for certain cost structures and enabling for others. The constraint is no longer production — it is curation, quality assessment, and the maintenance of trust in a world where synthetic content is indistinguishable from authentic at scale. For the investment implications of this shift, see our analysis of investing in the AI economy.
Financial Services: Mature Applications and Emerging Frontiers
AI in financial services is not a future prospect — it is a decade-old operational reality. Fraud detection at major card networks has been ML-powered since the early 2010s, processing millions of transactions per second and flagging anomalies that no human team could identify at that scale. Algorithmic trading systems process news, sentiment, and market microstructure data to execute strategies in microseconds. Credit scoring models increasingly incorporate non-traditional data sources to improve accuracy for underserved borrowers.
The current frontier is more consequential: real-time systemic risk modelling, AI-assisted regulatory compliance (a genuinely costly function that AI can substantially automate), and the integration of large language models into investment research and wealth management. The latter raises significant questions about fiduciary duty and the appropriate disclosure standards for AI-assisted advice — questions that regulators are beginning to address but have not yet resolved.
The Risks That Receive Insufficient Attention
The public discourse about AI risk tends to oscillate between dismissal and catastrophism. The more grounded risks are structural and near-term. Algorithmic bias — AI systems that encode and amplify historical discrimination in hiring, lending, and criminal justice — is documented and ongoing. Data privacy concerns are acute: the training data sets for large models include scraped personal information at scales that existing privacy frameworks were not designed to address. The economic displacement of specific job categories is not speculative; it is happening in legal research, software testing, graphic design, and customer service.
Perhaps most importantly, the concentration of AI capability in a small number of very large organisations creates structural dependencies and competitive moats that have significant implications for market structure, geopolitical competition, and the distribution of AI’s economic benefits. The EU AI Act, the US executive orders, and China’s regulatory framework represent the beginning of a governance architecture — but the pace of capability development consistently outstrips the pace of regulation.
AI is transforming industries not because of a single breakthrough but because a series of compounding improvements in model capability, training efficiency, and multi-modal integration have crossed practical thresholds in domain after domain. The frame of “AI replacing jobs” is too simple — the more accurate description is that AI is restructuring the production function across knowledge work, shifting the bottleneck from execution to judgement, curation, and oversight. The organisations and individuals who will benefit most are those who understand specifically what AI can do well, what it does poorly, and how to integrate it into workflows in ways that multiply rather than displace human capability. The time for “monitoring the situation” is over — the decisions being made now about AI adoption and governance will shape competitive and social outcomes for a decade.
Responses