Questions Gulf Coast plant managers actually ask.

ProcessIQ talks to PI via the PI Web API, so anything from PI Server 2016 forward works out of the box — including PI Server 2018 R2 and the current AVEVA PI System. Older versions (PI Server 2012 and earlier) usually work too as long as the Web API is exposed. We've connected to OSIsoft Asset Framework installations, collective-style PI Historians, and the newer cloud-hosted deployments without modification.

A typical pilot is live in 5–10 business days. Day 1–2 is read-only PI Web API credentials and tag selection. Day 3–5 is data ingestion and baseline pattern detection on one unit or area. Day 6–10 is the diagnostic engine tuned to your specific operations and a working session with your reliability team. We don't rip anything out during a pilot — your PI, DCS, and alarm systems stay exactly as they are.

No. Your data is never used to train any model. ProcessIQ uses OpenAI's GPT-4o through Polsia's enterprise proxy only at query time to interpret patterns we detect — the raw tag values stay inside your tenant. The AI sees a structured summary of detected anomalies (rate of change, correlation, limit violation, timestamp), not your full historian. Connections are TLS 1.2+ and we can deploy fully air-gapped if your IT team requires it.

Pricing is per-plant and tiered by tag count and connection type. A single-unit pilot typically starts in the low four figures per month. Multi-site enterprise rollouts are quoted against total monitored streams and support tier. We don't charge per-user, we don't charge per-diagnostic, and the pilot cost is credited toward your first year if you go live. Reach out via /contacts for a number tailored to your plant.

Yes — that's the entire premise. ProcessIQ is read-mostly against whatever historian you already run: PI System via the Web API, OPC UA via standard servers, OSIsoft Asset Framework, IP21, Honeywell PHD, Aspen IP21, and most SQL-fronted historians. If you can query it, we can pull from it. CSV and historian-export uploads are also supported for plants that don't allow direct outbound connections.

Yes. On-prem and fully air-gapped deployments are supported for plants in chemical, refining, and aerospace that can't reach the public internet from the control network. The diagnostic engine runs in your environment; the only optional outbound call is to the AI interpretation layer, which can be disabled in favor of an on-prem LLM. We've delivered this configuration to plants under NERC CIP and ITAR constraints.

No, and that's intentional. ProcessIQ sits alongside your DCS, your alarm management system, and PI Vision — we don't replace any of them. Think of us as the layer that tells you why a tag drifted three hours before the alarm fired, not another screen to watch. Your operators still use the screens they trust; we just give them (and you) better answers faster.

The diagnostic engine catches the patterns it was designed for — rate-of-change anomalies, correlated drifts, limit violations, stuck sensors, sustained deviations — with high precision because the detection step is deterministic. The AI interpretation step adds a ranked shortlist of likely causes and corrective actions. In pilots, the top-ranked hypothesis has matched the engineer's post-incident write-up on the first pass the majority of the time. We publish accuracy honestly: false positives happen, and the UI is built so engineers can mark a diagnostic as off-target so it learns what your plant looks like.

Almost always no. The default implementation is software-only: we connect to your existing PI / OPC / historian via read-only credentials, ingest the tags you nominate, and start running. No new transmitters, no new PLC code, no new wiring. The only on-prem component in a typical install is a small connector service (a single container or Windows service) that brokers the read-only connection — and even that can be replaced by a CSV upload pipeline if your network policy forbids installing anything in the control environment.

A chatbot answers whatever you ask it. A BI dashboard shows you whatever you built a screen for. ProcessIQ does neither — it watches your tags continuously, detects the patterns engineers would look for if they had the time, and surfaces only the anomalies worth a human's attention, ranked by likely cause and recommended next step. The AI is a layer on top of deterministic pattern detection, not a replacement for it. That's the difference between a tool you ask questions of and a tool that brings you the right question.