Argomenti trattati
First published April 12, 2026; last updated April 22, 2026. This report retells how a rapidly expanding roster of customers adopted agentic AI and production-ready systems. Organizations large and small are moving beyond simple assistants toward integrated, autonomous workflows that coordinate multiple models and services. To analyze that momentum, our team uploaded the complete dataset of deployments and case studies into Gemini Enterprise, ran deep research with the latest Gemini Pro models, and distilled the results into a compact set of actionable observations.
Working with the model outputs, our analysts selected five insights that repeatedly surfaced in conversations with customers and partners. Each insight reflects how teams are reshaping work: from unlocking fifty-year-old mainframes with natural language to using generative video models to produce thousands of personalized ads at near-zero incremental cost. The compiled list is organized across 11 industry groups and six agent types — Customer, Employee, Creative, Code, Data, and Security — and includes hundreds of new entries since the first edition.
How we distilled large-scale customer data into practical lessons
We fed the dataset into Gemini Enterprise and ran iterative queries using Gemini Pro for deep synthesis. The process combined automated trend extraction with human review to ensure context and accuracy. That hybrid workflow let us surface recurring patterns across thousands of deployments — patterns that are not merely theoretical but visible in production systems spanning automotive, retail, finance, healthcare, logistics, and media. The resulting five trends capture where the most impactful change is happening and where product and governance work is needed most.
Five defining trends shaping the agentic enterprise
From single helpers to coordinated agent teams
The most striking shift is away from solitary assistants toward agentic teams that orchestrate multi-step processes autonomously. In practice this looks like a supply chain agent coordinating with compliance, finance, and forecasting agents to execute complex decisions without human micro-management. Companies such as Accenture and Manhattan Associates are building governance and orchestration layers to manage these task forces, while vendors provide protocols and monitoring tools to keep behaviors auditable and safe.
Natural language as the bridge to legacy systems
Another powerful trend is using natural language as an interface to unlock decades-old systems. Organizations are placing an NL interface in front of SAP, mainframes, and COBOL-based systems so non-technical staff can query and act on siloed data. This approach preserves investment in legacy infrastructure while enabling rapid productivity gains — a pragmatic alternative to full migrations that many enterprises prefer.
Technical waves that matter: media, multimodality, and cybersecurity
Generative media is shifting from manual studios to computational factories. With models like Veo 3 and Imagen 4, marketing teams create hundreds or thousands of personalized video variations from a single creative brief, enabling real-time, test-driven creative strategies. Meanwhile, multimodality lets systems ingest live video, blueprints, and sensor feeds — powering applications such as robotic shelf audits, athlete biomechanics from smartphone footage, and factory safety monitoring. In security, defenders are deploying agentic auto-remediation systems that can write detection rules, isolate compromised workloads, and even create deceptive honeytokens to slow attackers.
Representative industry implementations
Across sectors we observe consistent patterns. Automotive examples include MercedesBenz using Gemini via Vertex AI for conversational in-vehicle assistants and Volkswagen’s myVW app leveraging multimodal queries. Logistics players like UPS and Stord build digital twins and predictive fulfillment agents on BigQuery and Vertex AI. In finance, banks from Banco Macro to Revolut are deploying conversational advisors and compliance agents; in healthcare, organizations such as CVS Health and Quest Diagnostics integrate agentic experiences into patient and clinician workflows. Media and retail teams use Veo and Imagen to scale creative production and personalization at previously impossible speeds.
What this means for leaders and builders
These developments create immediate opportunities and responsibilities. Builders must design for observability, access controls, and lifecycle governance while product leaders must rethink roles and processes so humans and agents collaborate effectively. Security teams should prioritize agent-aware defenses, and data teams must ensure grounding and provenance via retrieval-augmented generation and verified data sources. For those ready to experiment, the same platforms that power these case studies — Gemini Enterprise, Vertex AI, BigQuery, and the AI hypercomputer stack — offer a path from prototype to production.
If you want to explore the full dataset and try a similar synthesis, upload your data to Gemini Enterprise or use NotebookLM to surface ideas for your teams. The list continues to grow: this edition adds more than 300 new entries, marked with an asterisk in the dataset, and we expect future editions to showcase even broader adoption and new agent types as the technology matures.