MIT Technology Review’s EmTech Future 2026 put a familiar AI claim into a more operational context: the technology’s largest effects may emerge not from standalone models, but from its integration with other technical and industrial systems.
In a session highlighted from the event, Yossi Matias, Google’s vice president and head of Google Research, discussed AI’s growing role in biology, infrastructure, manufacturing and science. The three-day program also covered quantum computing, energy systems, robotics, climate technology and AI’s effects on work.
The conference material is promotional rather than a set of product announcements or deployment results. Still, its central framing is useful for operators: AI strategy is increasingly inseparable from strategy for data, physical operations, energy, scientific workflows and governance.
The shift is from tools to systems
Many early enterprise AI efforts focused on discrete tasks: drafting content, searching internal knowledge, writing code or supporting customer-service teams. Those uses can deliver value, but they generally fit inside existing software workflows.
The harder opportunity is AI embedded in systems that touch the physical world or high-consequence decisions. In manufacturing, that can mean connecting models to production data and operational processes. In biology and science, it can mean augmenting research workflows. In infrastructure and energy, it means dealing with long asset lifecycles, reliability requirements and multiple interconnected systems.
That is a substantially different implementation problem from licensing a general-purpose AI assistant. The quality of data, integration with existing tools, human review, measurement and failure handling become central to whether a project delivers value.
Convergence expands both the opportunity and the dependency list
EmTech Future’s agenda paired AI with quantum, energy, robotics and climate systems. The practical takeaway is not that every company needs a quantum or robotics roadmap. It is that technical capabilities increasingly depend on one another.
AI workloads require computing capacity and energy. Robotics depends on software, sensing and reliable physical operations. Scientific AI depends on access to specialized datasets and evaluation methods. Infrastructure applications introduce security, safety and regulatory constraints that do not disappear because a model is capable.
For executives, this makes cross-functional ownership more important. AI programs that sit only with an innovation team or a single business unit can struggle once they require data-platform changes, operational technology access, procurement changes, security approval or workforce redesign.
Build around a workflow, not a technology label
For founders and builders, the convergence theme favors products that solve a narrowly defined workflow across the software-to-operations boundary. A generic AI feature may be easy to replicate; durable value is more likely to come from domain data, deep integration, evaluation infrastructure and a clear role in an existing process.
For enterprise teams, the starting point should be a workflow where performance can be measured. Identify the decision or task to improve, the systems of record involved, the accountable human operator and the conditions under which automation must defer. That structure is especially important in science, manufacturing, infrastructure and other settings where an incorrect output has material costs.
What to watch next
The key signals will be evidence of deployment rather than broad convergence narratives: repeatable gains in research or industrial workflows, reliable integrations with legacy systems, and clearer operating models for human oversight.
Leaders should also watch the constraints. Energy availability, compute costs, proprietary data access, cybersecurity and regulation may determine which AI-plus-industry applications move from demonstrations to durable businesses.
EmTech Future’s broader message is straightforward: AI is becoming a component of more systems, not a separate category of work. The organizations that benefit most will be those able to manage the integration work that follows.
