MIT Technology Review Insights

Global healthcare providers are deploying agentic AI agents to address a projected 11 million worker shortage by 2030, with over two-thirds of providers already adopting the technology. Unlike previous digital health tools, agentic AI can handle complex scenarios autonomously, make decisions, and iterate without manual intervention. Hospital for Special Surgery has deployed AI agents to process insurance claims at scale and triage patients 24/7, reducing administrative burden on clinicians.

Organizations are adopting AI agents faster than their infrastructure can support them. While 85% of enterprises want to deploy agentic AI within three years, 76% lack the operational readiness across people, processes, and workflows. The core problem is that companies are layering AI agents onto existing human-centric operating models rather than fundamentally redesigning how work flows, preventing them from capturing the full value these systems can deliver.

Enterprises are shifting away from the early AI adoption model of outsourcing data and models to third-party providers, driven by concerns over IP loss and competitive advantage. A movement toward AI and data sovereignty, defined as breaking dependence on centralized providers and establishing control over models and data estates, is gaining momentum across global companies. Survey data from EDB shows 70% of executives believe they need sovereign data and AI platforms to succeed, while the conversation is also becoming a policy priority at the national level, with leaders like NVIDIA's Jensen Huang advocating for countries to build their own AI infrastructure.

Financial services firms deploying agentic AI face a critical bottleneck: data quality, security, and accessibility. More than half of financial services teams plan to implement agentic AI, but the technology amplifies existing weaknesses in data infrastructure. Success requires centralized, well-governed data stores that can handle both structured and unstructured information at scale while meeting strict regulatory audit requirements. A Forrester study found 57% of financial organizations still lack the internal capabilities to fully leverage agentic AI.

Organizations that capture outsized value from AI investments tend to adopt a customer-back engineering approach, starting with customer needs and working backward to technology solutions rather than the reverse. Capital One's Ashish Agrawal describes how embedding engineers directly with customers through empathy sessions, support rotations, and ride-alongs creates a motivational feedback loop and unlocks faster problem-solving. The approach is particularly powerful in the AI era, where engineers with direct data access can rapidly prototype agentic solutions like conversation summarization and follow-up automation that would be harder without high-quality customer data and close collaboration.

Finance departments are adopting AI tools faster than leadership can establish governance frameworks, creating a bottom-up transformation that outpaces top-down strategy. AI is embedding itself across workflows from fraud detection to contract review, particularly where unstructured data once created bottlenecks. The shift is forcing executives to reconcile productivity gains with oversight and risk management, while the real constraint emerging is not technology but talent: the gap between domain expertise and AI fluency, combined with the need for auditability and proper tool understanding.

Health care AI adoption is accelerating, with over 1,300 FDA-approved AI-enabled medical devices and rapid growth in administrative applications, but success requires deep understanding of clinical, technical, and business contexts. Developers and providers increasingly recognize that off-the-shelf solutions often fail because they misunderstand health care's complexity, leading 61% of health care organizations to pursue partnerships with vendors for customized solutions rather than building in-house or buying generic products. The sector faces a maturity challenge, with 77% of technology leaders citing immature AI tools as a significant adoption barrier, while regulators continue to shape an evolving policy landscape.

Enterprise AI adoption is being constrained not by model capability but by fragmented, ungoverned data infrastructure. Most organizations have information scattered across legacy systems and siloed applications, making it impossible for AI systems to generate trustworthy outputs at scale. Leading companies are now consolidating data into unified, open architectures with precise governance and real-time context, tying AI deployment directly to measurable business outcomes rather than treating it as isolated innovation.

As AI moves from experimentation into core business workflows, organizations are discovering that model performance and computing power are not the primary bottleneck. Instead, the critical challenge is ensuring AI systems have access to business context alongside raw data. Without understanding policies, processes, and real-world decision tradeoffs, AI can generate fast answers that optimize for the wrong outcomes. A well-designed data fabric that preserves semantic meaning across applications and systems is emerging as essential infrastructure to scale AI safely and align automated decisions with actual business priorities.

Public sector organizations face distinct operational constraints that make standard large language models impractical for government deployment. Small language models (SLMs) offer a more viable path forward, allowing agencies to maintain data control, ensure operational continuity, and avoid GPU infrastructure bottlenecks while delivering comparable performance to larger models. A Capgemini study found 79 percent of public sector executives worry about AI data security, and 65 percent struggle with real-time data use at scale, highlighting why purpose-built, locally-housed SLMs are better suited to government environments than cloud-dependent LLMs.

A survey of 300 engineering executives finds that agentic AI in software development is moving from narrow task assistance toward autonomous management of entire software projects and lifecycles. Currently 51% of teams use agentic AI in limited ways, with 45% planning adoption within 12 months. While most expect only incremental improvements over two years, nearly all anticipate 37% faster time-to-market, and 41% aim to have agents managing full product and software development lifecycles for most or all products within 18 months.

Privacy-led UX, which treats data transparency and user consent as core elements of customer relationships rather than compliance checkboxes, is emerging as a business growth lever alongside regulatory necessity. MIT Technology Review Insights examines how organizations can build consumer trust through well-designed consent experiences, particularly as AI systems add complexity to data governance. The report identifies privacy infrastructure as foundational to responsible AI deployment at scale, with cross-functional ownership and clear consent frameworks as prerequisites for success.