As companies rush to adopt GenAI, one question keeps coming up: How do we make all this intelligence useful on the plant floor?
SAS Vice President of IoT Jason Mann and Jobi George, Global VP of Business Development at Weaviate, discussed how vector databases and retrieval-augmented generation (RAG) are enabling manufacturers to transform raw data into contextual intelligence.
The conversation highlights a simple but powerful shift: Industrial IoT is no longer just about detecting anomalies – it’s about explaining them.
1. GenAI adds the missing context to anomaly detection
For years, SAS has helped industrial companies flag anomalies and outliers. But as Mann pointed out, alerts often lack the “why,” making root-cause analysis slow and complicated.
George explained that this is where vector databases and RAG change the game.
Most industrial data still resides in isolated systems – some structured (historian data) and some unstructured (PDF manuals, maintenance logs, technician notes, and faxes). Without context, an alert is just noise.
“With a vector database–based RAG approach, you’re able to add a lot more context into that information,” George said. “Now you can build applications that are more relevant and support intelligent decision-making.”
By combining SAS IoT analytics with Weaviate’s vector search, teams can automatically link sensor anomalies to recorded maintenance history, operating procedures, parts documentation and more.
The result? Faster answers, smarter decisions and fewer blind spots.
2. No-code tools make GenAI accessible on the plant floor
Weaviate and SAS have been collaborating for years, with one clear goal: simplify GenAI adoption for real industrial users.
“We make it a lot easier to create RAG applications… with a no-code approach,” George said.
That accessibility matters. Many manufacturers want GenAI but lack the specialized skills (and time) to build large-language-model applications from scratch. No-code workflows enable teams to experiment, deploy, and refine solutions without requiring a machine learning team behind them.
As Mann noted, this is what unlocks scale: “No-code is really important for being able to make this accessible to the masses and generate extensive benefit.”
In other words, GenAI becomes a tool on the factory floor, not just an IT project.
3. Unstructured data is the next frontier for industrial intelligence
Manufacturers are deeply familiar with historical systems – structured time-series data that has powered operations for decades. But historical data only captures part of the story.
“This industry has a ton of information trapped in unstructured spaces,” George said. “PDFs, maintenance records, technician notes – all of it matters, but none of it has context on where it’s sitting.”
New GenAI advances – from vector databases to emerging models like ColBERT, ColPali and others are making it possible to finally put that information to work. These tools can ingest documents across disparate systems, index them semantically and transform them into real-time intelligence.
This unlocks use cases that weren’t feasible before:
- Automatically linking an anomaly to past repairs.
- Surfacing the correct troubleshooting steps instantly.
- Building natural-language access to decades of operational knowledge.
- Reducing reliance on tribal knowledge as experts retire.
As George put it: “The technology is here now, and we’re able to solve use cases that were not possible to solve before.”
Meeting industrial problems that exist right now
Despite the excitement surrounding GenAI, many manufacturers remain hesitant because they are unsure where to begin or how to demonstrate ROI. Mann and George agree that hesitation is the biggest challenge.
However, the tools are maturing rapidly, and the problems they solve – workforce shortages, downtime costs, and documentation overload – are urgent today, not in years to come.
SAS and Weaviate are focused on closing that gap.
