This week at AVEVA World 2026 in Milan, TrendMiner is introducing two features that change how engineers and operators get answers out of their plant data: the TrendMiner Agent and Routines. Both are live inside the platform, and both reflect the same conviction. AI should work alongside the people running the plant, not invisibly in the background.
The TrendMiner Agent: now available in your view, and grounded in it
The Agent is an autonomous, in-app assistant. It reasons across your historian and asset data to deliver conclusions with explicit assumptions, a full report of root cause and evidence, and recommended actions. You can configure monitors or dashboards without leaving the conversation.
What sets it apart is the grounding. The Agent isn't generic. It auto-grounds on the view you're already working in, so when you ask about heat rate on a specific asset, it doesn't try to scan your whole historian. It reasons through the most logical places first, then expands from there if needed.
It also goes beyond what TrendMiner's MCP can do on its own. You get the things you'd expect from a modern AI agent: the ability to connect any LLM, work with TrendMiner data and relationships, and keep persistent items for reuse. You also get capabilities that only make sense within the UI of the application. The Agent can manipulate analysis and visualizations directly, and it pairs natively with Routines (more on that next).
Three capabilities do the underlying work. Data relationship inference, so the Agent understands how tags, assets, and events connect. Autonomous multistep reasoning, so it can plan and execute investigations end to end. No-code workflow configuration, so the outputs are usable by people on the floor.
Routines: the analysis, readable, editable, reusable
A Routine is a way to document and structure an analysis as an ordered sequence of steps that any colleague can interpret. It pairs naturally with the Agent — you can assign a Routine to capture and reproduce an Agent investigation as a shareable analysis or report — and it works equally well with AI-assisted analysis and visualization steps you build yourself. Edit a step, swap a tag, rerun next month to update the analysis. Share the same analysis across lines or sites for scaled solutions.
This sounds simple on paper. In practice, it's the difference between an investigation that lives in one engineer's head and an investigation that becomes institutional knowledge. The features doing the heavy lifting here: drag-and-drop step ordering for quick edits, shareability across lines and sites, and one-click conversion to a monitor, so a one-off troubleshooting question can become a continuous early warning the next time it matters.
Why this matters
The broader AI conversation in industrial analytics has tended toward agent and workflow tooling that requires heavy manual configuration: significant time investment and specialized skill to reach the intended outcomes. TrendMiner has taken a different view. The Agent auto-configures on the view you're already working in, shows its reasoning step by step, and produces Routines anyone on the team can read and edit. Plant knowledge sits with the people running the plant, and the right job for AI is to give them more time, more context, and more institutional memory.
The Agent is the investigative power. Fast, patient investigation that shows its work. Routines are the memory. Reusable, shareable, editable analyses that scale beyond a single shift. Together they let process and production engineers on the shop floor move faster on the questions they already know how to ask, and start scaling answers across the rest of the business.
One conversation. One readable artifact. Reusable on any plant, any time.
See them live this week
Both features are on the floor at AVEVA World 2026, Allianz MiCo, Milan, May 19–21. Find us at booth S4 in the Innovation Zone. The team is set up for live demos and one-to-one conversations.
Event details: trendminer.com/events/aveva-world-2026
Book a demo: https://www.trendminer.com/request/demo
