Browse
Log in / Register
Share
April 20, 2026 — Enterprise AI has shifted from pilot projects to operational workhorses, and the data now shows exactly where. A new IoT Analytics analysis of 1,533 real-world generative and agentic AI projects, announced between Q4 2022 and Q1 2026, finds operational process optimization is the most common enterprise AI application, appearing in 52.2% of recent deployments. The analysis, based on the Generative AI & Agentic AI Projects List 2026, also documents a fast but still early rise in agentic AI: the share of projects running multi-step tasks autonomously jumped from 1.1% in 2025 to 17.8% in Q1 2026.
Generative AI has crossed the chasm and is moving into late-majority adoption. The sequencing is visible in the project data. Technology companies were first, testing capabilities and embedding them into products. Financial services followed, using the technology to modernize customer service and back-office work. Retail trade and professional services are now the early majority and posting results. Manufacturing and healthcare have moved more slowly, held back by long investment cycles and data-security concerns, though activity in both sectors has picked up recently. Education and administrative services are still finding their footing: many have pilots, but few have a clear strategy.
Agentic AI is a different story. Technology companies such as Spotify, Shopify, and Ericsson lead the way, and finance and professional services firms like OffDeal and EY are already building complex agentic workflows. Most other industries remain in the pilot stage. In 2025 and Q1 2026, single-agent systems accounted for 11.3% of projects and multi-agent systems for 2.0%, up from 0.3% each in 2022–2024. Non-agentic chatbots and assistants still dominate, representing 60.1% of projects, down from 65.0%.
The dataset, which grew from 530 to 1,533 projects, points to three notable movements.
Operations is the center of gravity. Process optimization jumped from 30% of projects in 2022–2024 to 52.2% in 2025 and Q1 2026. It was already the top-ranked use case; now it leads by a wide margin.
Agentic AI has become a real category. Single- and multi-agent systems combined rose from 0.5% of projects in 2022–2024 to 13.4% in 2025 and Q1 2026.
Autonomy is climbing. Only 1.1% of projects executed multi-step tasks autonomously in 2025. By Q1 2026, 17.8% did. Over the same stretch, output-only projects fell from 59.5% to 32.2%. The Q1 figures are based on just 174 projects, so they indicate direction rather than an established trend.
IoT Analytics ranked the business activities most frequently targeted by 734 projects announced in 2025 and Q1 2026. Here is the full ranking.
AI is improving the speed, cost, quality, and reliability of core workflows — manufacturing lines, loan processing, order fulfillment. In agentic setups, the software also performs parts of the work itself.
NielsenIQ (NIQ) shows the payoff. The US market researcher automated item coding, converting product packaging images into structured data for its database of 220 million products, using Microsoft Foundry and Azure OpenAI. Coding once took roughly four minutes per item. In one deployment, NIQ coded 32,000 products in 10 hours, work that previously took 300 hours. NIQ reports 90% less item-coding time and expanded the system to Germany, France, Ireland, and Canada within two months of launch. "It was a labor-intensive process that required deep market knowledge and meticulous attention to detail," said Gabriel Harris, principal data scientist at NIQ. "We saw an opportunity to eliminate the manual steps by using language models to automate the entire pipeline."
AI produces copy, images, video, and campaign assets; agentic versions split work into parallel tasks and assemble the final output. The benefit is elapsed time from idea to published asset, plus more variants for different audiences and languages at lower cost.
Kraft Heinz built a generative AI platform called TasteMaker with Google Cloud and partner Apply Digital. It combines the company's BigQuery data with Google's Gemini, Imagen, and Veo models on Vertex AI, and employees use it without writing prompts. By May 2025, Kraft Heinz was using it for product concepts and marketing materials matched to customer preferences. What previously required about eight weeks to reach market now ships in as little as eight hours.
Assistants and agents answer routine questions across chat, voice, and email, escalating complex cases to staff. Companies cut cost per contact, shorten response times, and get 24/7 coverage in multiple languages. In May 2025, US online used-car retailer Carvana reported results from two AI customer-support tools built with Microsoft, including an AI agent named Sebastian that guides customers through the process.
AI assists with code generation, code review, testing, documentation, and debugging. Developer productivity and faster delivery are the direct outcomes.
AI converts raw operational data into summaries, dashboards, and narratives. Reporting projects reduce the effort behind business reviews and get insights to decision-makers faster.
Transcription, meeting notes, knowledge retrieval, and internal assistants cut the time teams spend searching for information. These are among the easiest AI business cases to justify.
A step beyond inquiry handling, these systems diagnose problems and recommend fixes for complex cases. Routine questions are solved automatically; harder issues reach staff with full context.
AI speeds analysis for product development and research teams, from customer insights to technical literature review. Shorter discovery cycles mean more ideas can be tested.
AI monitors regulatory changes, screens contracts and filings for risk, and keeps compliance teams current. As AI-specific rules multiply, this category is likely to grow.
AI inspects products and production processes, catching defects earlier and more consistently than manual checks. It feeds directly into the error-rate and downtime metrics operations teams own.
The list sends several clear signals.
Enterprise buyers are steering AI toward measurable operational problems. Process optimization, reporting, and quality control all map to cost per transaction, cycle time, and error rate. The ROI conversation has therefore shifted from "what can it do?" to "what does it save?" Vendors that lead with model benchmarks rather than workflow-level results will struggle to win deals.
Customer-facing use cases remain substantial, but they are no longer the headline. Content creation, inquiry handling, and issue resolution together account for a large chunk of activity; operations and IT use cases are where momentum is consolidating. The presence of regulatory compliance at number nine also signals that governance is moving from overhead to a first-class AI application — a trend likely to accelerate as AI-specific laws take effect.
For enterprises, the practical lesson is that the most common applications are not exotic. They are data-heavy workflows with clear metrics and room for human oversight. That makes them easier to pilot, evaluate, and scale in stages.
The strongest movement in the data is toward autonomy. The jump in multi-step execution, the shrinking share of output-only projects, and the rise of agentic architectures all point the same way: AI systems are increasingly expected to do work, not just produce content.
The change is still early, and the quarterly swings should be read with caution. Agentic AI represents only 13.4% of recent projects, and the Q1 2026 figures are built on a modest base. Reliability, observability, and human oversight remain open questions, especially for multi-agent systems. Companies that buy "agents" in 2026 should ask whether those systems can be audited, contained, and measured against the same operational KPIs that made process optimization the top enterprise AI application in the first place.
The ranking covers 734 generative and agentic AI projects announced in 2025 and Q1 2026, drawn from the Generative AI & Agentic AI Projects List 2026. The full database tracks 1,533 projects from Q4 2022 through Q1 2026. IoT Analytics cautions that the Q1 2026 figures, based on 174 projects, indicate direction rather than a stable share.
What is the most common enterprise AI application? Operational process optimization, which appears in 52.2% of projects
