White paper
INTEL FOUNDRY × FORMANT
Proving the value of context in AI‑driven industrial operations
Intel Foundry Automation and Formant tested agentic AI on 47 real alarms. With connected operational data and domain knowledge in the loop, context‑aware AI cut time to resolution by 60% on average — and answered more accurately.
Published by Intel Corporation · Read on intel.com
THE RESULTS
Same alarms. Same operators. Far less time to resolution.
60%
TIME TO RESOLUTION
Average reduction across 47 alarms (~54% weighted by alarm volume).
7.5×
DOCUMENT RETRIEVAL
From ~15 minutes to ~2 minutes per alarm.
66%
STEP AUTOMATION
Across 47 alarms and 380 TSS procedures.
~393h
TECHNICIAN TIME
Saved per year, ~46 minutes per alarm.
THE CONTEXT
Connected data plus domain knowledge. Answers that hold up.
The study ran agentic AI in advisor mode: the system recommends, the operator commands. With connected operational data and procedures in view, responses were faster and more accurate, and institutional knowledge stayed in the system.
STUDY SCOPE
DATA
540 days of historical data
SYSTEM
SCW ion‑exchange system
MODE
Advisor mode
ALARMS
47 alarms · 380 TSS procedures
01
Context drives accuracy
Industrial AI is only as good as its context. Connecting live operational data with domain knowledge let the AI respond faster and more accurately than prompts alone.
02
60% less time to resolution
Across 47 real Intel Foundry alarms, context‑aware agentic AI reduced average time to resolution by 60%, shortening the path from signal to action.
03
Institutional knowledge, preserved
Expert know‑how was captured and made available to every operator, so resolution no longer depends on who happens to be on shift.
04
Built for remote operations
Designed for centralized Remote Operations Centers that monitor and respond to events across many sites from a single location.
The challenge
As Intel Foundry Facilities moves toward centralized Remote Operations Centers (ROCs), engineers increasingly need to monitor and respond to alerts and alarm events across multiple industrial facilities from a single location. Every minute spent reconstructing what a signal means, which system it belongs to, and what has worked before adds to time‑to‑resolution — and much of that knowledge lives with individual experts.
The proof of concept
To explore how agentic AI could reduce time‑to‑resolution, preserve institutional knowledge, and enable the next generation of remotely managed facilities, Intel Foundry Automation collaborated with Formant on a proof of concept. The AI was tested against 47 real alarms, with Formant’s operating layer supplying the operational data, equipment context, and domain knowledge it needed to reason about each event rather than answering from generic training data.
“In AI‑driven industrial operations, connected operational data and domain knowledge — that is, extensive context — enable faster, more accurate AI responses.”
Intel white paper — Proving the Value of Context in AI‑Driven Industrial Operations
What it proved
The PoC demonstrated that in AI‑driven industrial operations, extensive context is what separates a plausible answer from a useful one. With connected data and domain knowledge in the loop, the AI resolved alarms 60% faster on average and more accurately, pointing the way toward ROCs where engineers can resolve issues across many sites without losing the expertise that makes each site run.
The full white paper, including the PoC architecture, methodology, and findings, is published by Intel and available on intel.com.
From operational chaos to systemic command.
Read the full study, or book a diagnostic of your own operation.
Results from a proof of concept using 540 days of historical data from a single ion‑exchange system, in advisor mode. Your costs and results may vary. Performance varies by use, configuration, and other factors. Intel and the Intel logo are trademarks of Intel Corporation or its subsidiaries.