Applied Deep Dive: What Companies Are Building with the New Technology

# What companies are actually building

# Concrete deployments and what they show

Narayana Health in Bengaluru redesigned a new hospital without an onsite diagnostic lab. Instead, the facility relies on drones to move diagnostic samples about 2.5 miles in about seven minutes, a trip that can take three to five hours by road when batching is included. The hospital's design change — omitting a lab entirely — shows how the technology can influence physical infrastructure when integrated with logistics.

Frito-Lay is working with driverless trucks operated on predictable routes. The approach started with fixed trips under 10 miles and expanded to dynamic routes covering up to 400 miles and dozens of stops. The pattern: automate a predictable, valuable route rather than trying to make every truck intelligent.

Google Research and the UK air-traffic provider NATS are testing weather forecasts to redraw flight paths so aircraft avoid creating contrails over the North Atlantic. The trial includes operational phases and satellite observation feedback to confirm whether predicted contrails formed — a closed-loop test where a prediction leads to an action and the sky is checked for results.

Caterpillar's field assistant is a voice interface for technicians, but the real advantage is the 1.6 million connected machines and 16 petabytes of structured data behind it. The assistant must map a voice request to the exact machine, applicable manual, permission rules, required logging, and handoff points for human control. The operating context, not the assistant alone, is the competitive asset.

# What companies report publicly — and what they don't

The directory's striking gap is outcome reporting. Of 136 entries, 98 had no reported measurable result. When companies do disclose details, filings and technical reports are more informative than demos. SEC filings frequently reveal governance steps, hardware and cooling plans, offline compute strategies, and the decision controls that production systems require.

• Intapp documents expert agents that follow firm-specific playbooks with permission controls and decision-tracing logs, making governance part of the product.

• ChronoScale plans large-scale compute and liquid cooling installations, highlighting infrastructure needs in production deployments.

# Common patterns for successful deployments

• Narrow scope: single task, route, or machine where success is measurable.
• Proprietary data and institutional memory that models can access and act on within business rules.
• Workflow integration: results must fit into how people already work, including logging, permissions, and escalation.
• Feedback loops to verify outcomes, whether satellite observations, flight logs, or lab turnaround time.

# Where the hard work is

# Bottom line

Companies are applying the technology to real-world tasks beyond conversational use, but evidence of impact is still sparse in public reporting. The most convincing cases pair a focused use case with clear measurement and system-level engineering that ties predictions to verifiable outcomes.