Why Ryedore
A monitoring platform tells you something changed. Ryedore tells you why — and lets you check.
Side by side with monitoring / IoT platforms and generic AI: what you get, how you know it’s right, what happens over time, where your data goes, who decides — with an example and a how‑to‑check link on every row.
Seven differences — each with an example and a way to check it
The foil is the category — monitoring / IoT platforms and generic AI — not a named vendor. Examples are worked scenarios. illustrative
| Dimension | Monitoring / IoT & generic AI | Ryedore | Example | How you can check |
|---|---|---|---|---|
| What you get | An alert: a threshold crossed, a score, a colour. | An answer: cause chain, remaining life, what to do, confidence. | Vacuum pump VP‑2 on a paper machine: “vibration high” vs “bearing wear from a misaligned coupling — ~9 days — re‑align first, replace bearing at the next stop, 88 %.” | Read a full trace → |
| How you know it’s right | Trust the vendor; a case study you can’t reproduce. | Public datasets, signed results, re‑runnable — including where we lose. | The remaining‑life and anomaly results are published on NASA turbofan and hydraulic‑rig data with a signature and a re‑run script. | Re‑run a benchmark → |
| What happens over time | The model is static — or it is retrained and may quietly get worse. | Every new version must beat the current one on every task, or it is blocked. | A candidate that would have dropped category accuracy from 0.90 to 0.75 was blocked; the log is public. | See the gate log → |
| Where your data goes | To the vendor’s cloud — the model needs it there. | Nowhere. Model, data and learning stay on your hardware; air‑gap capable. | A refinery runs the whole platform inside its OT network; the vendor never sees a sensor value. | Security & certification status → |
| When something must be done | A ticket, or an autonomous action you didn’t see coming. | An agent that proposes, passes six checks, and waits for a human. | Ryekronix requests a work order to re‑align VP‑2; a person approves it in Approvals — or it never happens. | How actions are gated → |
| Where it is going | A static product; updates on the vendor’s timetable, unexplained. | A platform that improves by design — it proposes its own experiments and a gate decides what ships — with a public roadmap whose proof points are signed. | Improvement proposers fire centrally; a candidate that would regress on any task is blocked; the ten‑program roadmap is published with a proof point each. | What the round builds → |
| Before you spend | Change it and see. | Six AI specialists test the change on a physics twin first. | “Would a ceramic bearing on VP‑2 last longer?” — baseline vs intervention, statistics, provenance — before the purchase order. | How an experiment runs → |
What each kind of tool can actually show you
Categories, not vendors. ● built in and checkable · ◐ partial or vendor‑asserted · ○ not there. Every ● on our column links to the page where you check it.
| Capability | Monitoring / IoT platform | Generic AI / chat | Single‑purpose AI tool | Ryedore | Check it |
|---|---|---|---|---|---|
| Cause chain with evidence on every prediction | ○ | ◐ | ○ | ● | open → |
| Remaining life with a range, not a score | ○ | ○ | ◐ | ● | open → |
| Signed, re‑runnable public benchmarks — losses included | ○ | ○ | ○ | ● | open → |
| A model that cannot get worse (gated, logged) | ○ | ○ | ○ | ● | open → |
| Runs entirely on your hardware; air‑gap capable | ◐ | ○ | ◐ | ● | open → |
| Learns from your labels without forgetting | ○ | ○ | ◐ | ● | open → |
| Cross‑industry learning — one model, many sectors | ○ | ◐ | ○ | ● | open → |
| Physics‑ and standards‑checked answers | ○ | ○ | ◐ | ● | open → |
| Agent that acts — only after checks and a human | ○ | ○ | ○ | ● | open → |
| Controlled experiments on a physics twin before you spend | ○ | ○ | ○ | ● | open → |
| Calibrated confidence — it says how sure it is | ○ | ○ | ◐ | ● | open → |
| Own model — no third‑party model anywhere | n/a | ○ | ◐ | ● | open → |
| An audit you can run on your own data | ○ | ○ | ○ | ● | open → |
Alert vs answer.
The alert tells you that something changed. The answer tells you why, how long, and what to do — and shows how it got there.
Everything else on this page is that sentence, checked.
Four kinds of tool, four blind spots
Categories, not names. Each is good at what it was built for — none was built to answer why with proof, on your hardware.
Great at work orders; the “AI” is a threshold with a nicer name. No cause, no proof.
One failure mode, one sensor type, one industry. Every new asset is a new project.
Powerful building blocks — you build, host and prove the model yourself, in their cloud.
Knows its own machine; blind to the line around it, and the data stays with the OEM.
What changes for each person in the room
Fewer surprises: a remaining‑life number with a range instead of a red light — and the reason.
Cause chains you can argue with, actions you approve, a memory that never loses a confirmed root cause.
Nothing leaves the site; certification status stated honestly; a signed audit chain.
Signed benchmarks and a gate log instead of vendor slides; an audit you can run yourself before any purchase.
Four answers
- Do we have to replace our monitoring platform?
- No. Ryedore is the intelligence layer on top of it. Your thresholds, dashboards and work‑order system stay; Ryedore adds why, how long and what to do — with a trace behind each.
- Why not just use a general AI assistant?
- A chat assistant matches patterns in text and has no access to your plant. Ryedore runs real tools on your live data, checks answers against physics and standards, states a calibrated confidence, and never acts without six checks and a person.
- How is this different from the AI bolted onto our CMMS?
- That “AI” is usually a threshold with a nicer name: it tells you something changed. Ryedore gives the cause chain and the remaining life with a range, and every model version is gated so it cannot quietly get worse.
- What can we verify before we ever talk to sales?
- The signed benchmark results on public data (losses included), the never‑degrade gate log, the certification register with honest statuses, the published price list — and a self‑serve audit you can run on your own data.
Check us before you talk to us.
Open the signed report, read a trace, see the gate log — then bring one machine to a demo.