AI for the PI System.
Six products, running live.
Conversational AI, excursion capture, high-resolution recorded data and predictive analytics, on one platform that connects them to your PI data. Every screen on this page is a live demo.
Conversational AI for PI, inside Microsoft Copilot.
An AI agent for Microsoft 365 Copilot. Engineers and managers ask about any PI tag in plain language from Teams, Outlook or the Copilot app, and get values, statistics, trends and Excel files back in seconds.
- No extra AI subscription. It runs on the Microsoft 365 Copilot licenses you already have.
- Natural-language analytics on live and historical data: time-weighted averages, min, max, totals and counts over any time range.
- Inline charts and Excel exports, the shift report you would otherwise build by hand.
- Packaged for your tenant. Your Microsoft 365 admin uploads one app package and assigns users.
- Needs
- SypinIQ-Plus, Microsoft 365 Copilot licenses
- Network
- SypinIQ-Plus endpoint published to Microsoft 365 over TLS
| Shift | Avg °C | Max °C | Above H |
|---|---|---|---|
| Night | 86.9 | 92.4 | 0 min |
| Day | 89.7 | 96.1 | 18 min |
| Evening | 88.3 | 94.8 | 0 min |
Animated demo · illustrative data
An AI copilot on every PI Vision display.
A context-aware AI assistant embedded in PI Vision. It reads the tags on the open display, so operators get trend explanations and root-cause hints without leaving the screen.
- Display-aware context, so “why is this rising?” is answered with the tags in front of you.
- Explains trends in plain language, pointing to related tags that moved at the same time.
- Your choice of large language model: Azure OpenAI, OpenAI or Anthropic Claude, including your own private Azure deployment.
- Entra ID sign-in and a full audit trail of every question asked.
- Needs
- SypinIQ-Plus, PI Vision 2023, an AI provider key
- Network
- Browsers reach SypinIQ-Plus on TCP 8443
Animated demo · illustrative data
Capture every excursion before it becomes an incident.
Safe operating limits are the last line of defence before alarms, trips and integrity damage. The SOL symbol draws LL, L, H and HH on the live trend and captures every excursion automatically: when it started, how far it went and how long it lasted.
Why it matters. Most excursions are short and recover on their own, so nobody counts them. Repeated short excursions are often the early warning of corrosion, catalyst damage or off-spec product. SOL makes them visible and countable.
- Automatic excursion capture: every crossing timestamped, with peak value and duration, per limit.
- Time-in-window KPI: the share of running time spent inside the operating envelope, per shift or campaign.
- Equipment running flag. Link a pump or reactor status tag so excursions only count while the equipment is in service. No false counts during shutdowns.
- Evidence for process safety and IOW reviews. Shift handovers, incident investigations and integrity operating window reviews all use the same numbers.
- Needs
- PI Vision 2023, PI Web API
- Network
- Runs inside PI Vision; no AI provider needed
| Start | Limit | Peak | Duration |
|---|---|---|---|
| 05:41:30 | H | 95.8 °C | 04:30 |
| 05:12:00 | L | 77.6 °C | 02:00 |
| 04:38:30 | H | 96.4 °C | 06:00 |
Animated demo · illustrative data
Every archived event. Timestamp for timestamp.
A high-resolution table symbol that reads raw archive events through PI Web API, not the interpolated plot values a trend shows. Compare tags timestamp for timestamp, with the exact numbers engineers need for investigations.
- Timestamp comparison: several tags side by side on their recorded events, with tag and unit headers.
- Statistics block: event count, average, minimum, maximum and total, time-weighted or event-weighted.
- Limit and multistate highlighting using the same bands as PI Vision, plus four number formats.
- CSV export with two-row headers, ready for Excel or Python.
- Needs
- PI Vision 2023, PI Web API
- Network
- Runs inside PI Vision; no AI provider needed
| Timestamp | 21TI1103.PV | 21PI2101.PV | P-101 | Δt |
|---|---|---|---|---|
| °C | bar | state | ||
| 06:11:58 | 91.26 | 8.44 | Running | 2:36 |
| 06:09:22 | 93.90 | 8.47 | Tripped | 2:42 |
| 06:06:40 | 96.08 | 8.60 | Running | 1:33 |
| 06:05:07 | 95.31 | 8.52 | Running | 2:16 |
| 06:02:51 | 90.17 | 8.39 | Running | 2:37 |
| 06:00:14 | 88.42 | 8.41 | Running | 1:58 |
Animated demo · illustrative data
Predictive analytics on your PI trends.
Period Compare overlays past periods on today's trace and projects where the process is heading. The forecast blends physics-based and statistical regression models trained on your own PI history, with a confidence band that tightens as live data arrives.
- Hybrid forecasting: first-principles process behaviour combined with statistical regression on historical periods.
- Confidence band and horizon that update with every new recorded value.
- Timestamp-aligned comparison with last year and the year before, plus a live deviation strip in %.
- Built for seasonal KPIs: throughput, energy per tonne, yield, cooling-water temperature.
- Needs
- PI Vision 2023, PI Web API
- Network
- Model runs on your own servers
Animated demo · illustrative data
The AI engine that connects agents to PI, safely.
The platform every SypinIQ product runs on, and an MCP (Model Context Protocol) server on its own. Give Claude, Copilot Studio or your own AI agents governed, read-only access to PI data.
- MCP server for PI: tag search, current values, recorded history and summaries as tools any MCP client can call.
- Admin console for PI Web API connections, data servers, TLS, licensing and settings.
- Security built in: Entra ID tokens, optional group restriction, DPAPI-encrypted credentials and a full audit log.
- One Windows installer that sets up the service and discovers your PI Data Servers.
- Needs
- Windows Server 2019 or 2022, PI Web API read-only account
- Network
- Listens on TCP 8443 (configurable), reaches PI Web API on 443
Animated demo · illustrative data
See what each package includes and build your own.
See it on your own PI data.
A 30-minute call. We connect to a test PI Web API, ask real questions of your tags and drop the symbols onto one of your displays.