The Operon platform
The platform that understands the physical world.
Two layers. A context layer that connects your drawings, time series, documents and email into one model of your company, and the AI employees that work on it, inside your controls and your systems.
AI employees
Context layer
One model of your company. Every fact traced to its source.
Your data
AI is only as good as the context it works from. In industry, building that context is the hard part. It's what the platform does.
01 · Context
Every source, connected.
The context layer reads every kind of data your company runs on and links it. The pump on the P&ID, its vibration trend, its datasheet, its work orders and the emails about it become one record, and every fact keeps a link to where it came from.
- Drawings read symbol by symbol, with 97%+ detection accuracy on P&ID symbols, tags and lines
- Assets, lines, rooms and jobs linked across every source
- Every fact traced to the sheet, reading or message it came from
- Low-confidence items go to a person before anything relies on them
Context
P-101 · Crude feed pump
Unit 300 · Running
- Drawings3 sheets, latest Rev D
- Vibration7.1 mm/s, rising
- DatasheetAPI 610, single seal
- Work orders12, last in March 2025
- Email2 vendor threads
Every fact links to its source
02 · AI employees
Built for your company's work.
On the context layer we build AI employees for specific jobs at your company: the bid inbox, take-offs, submittal reviews, maintenance planning, MOC packages and compliance documents. Each one follows your rules and templates and works in the systems your team already uses.
- Built around how your company works, not a generic template
- On live work from the day it goes live
- Drafts in your formats, sent through your systems
- Each new AI employee reuses the same context, so it starts faster
Northline Mechanical
AI employees · Today
- Done
Bid inbox
38 emails filed to their jobs, 4 replies drafted
- Approve
Take-off · Maple Office, Level 2
Updated for Addendum 2
- Approve
Submittal review · AHU-1
2 deviations from the spec, each cited
- Done
Service agreements
12 renewals due, quotes drafted
03 · Knowledge
Ask anything. Get the source.
Plain-language questions across drawings, readings, specs, work orders and email. Answers cite the sheet, reading or message behind them, follow the same permissions as your projects, and say plainly when the data doesn't support an answer.
- Every answer carries its sources
- One question can span units, jobs and years of records
- Access mirrors your projects, companies and roles
- No answer is better than a confident wrong one
What protects R-201 from overpressure, and is anything changing?
Two relief valves: PSV-1302 on the reactor, set at 150 psig, and PSV-1305 on the feed line. MOC 2291 proposes raising PSV-1302 to 165 psig. It's in review.
04 · Approvals and control
AI employees do the work. People sign it.
Every AI employee works inside the same controls. It reads only what it's allowed to, logs every step against the data it used, and stops at an approval gate before anything leaves the company, goes into a price or changes a controlled record.
- Every step is logged against the data it read
- Nothing leaves the company without a named approver
- Access mirrors your projects, companies and roles
- Runs on demand, on a schedule, or when something new arrives
Work order WO-4417 · P-101
Waiting for approval- Read the P-101 vibration trend, last 7 days
- Found standby pump P-101B on P&ID 300-PID-0114
- Compared the seal's age with its datasheet rating
- Drafted the work order in your template
05 · Integrations
Inside the systems you already run.
Operon reads from email, file shares, document control, the historian, the maintenance system and ERP, and writes back through the same systems. For anything else there are exports, REST, GraphQL and an SDK. Our engineers build the connections each deployment needs.
- Nothing to replace: Operon works where your data already lives
- Read-only to start; write access only where you allow it
- Exports to spreadsheets, registers and downstream systems
- Partners and integrators can build on the same layer
Connected systems
- Connected
Email
Bid and project inboxes
- Connected
File shares
Drawings and specs
- Connected
Document control
Revisions and transmittals
- Connected
Historian
Tags, readings and alarms
- Connected
Maintenance system
Work orders and history
- Connected
ERP
Jobs, costs and purchase orders
How it works
From data to action, in five steps.
Every step can be inspected, corrected and audited. Nothing is a black box.
01
Connect
Plug into email, file shares, the historian and the systems you already run.
Read-only to start
02
Read
Drawings, readings, documents and messages, each item with a confidence score.
Scans · CAD · PDFs · tags
03
Link
One model of your company: assets, lines, rooms and jobs across every source.
Every fact traced to its source
04
Act
AI employees do the work: take-offs, reviews, work orders and drafts.
In your formats
05
Approve
A person signs before anything leaves, and every step is logged.
Approval gates · audit log
The jobs on the platform
Every job Operon does today, and the template every new deployment starts from.
See all jobsMaterial take-off
HVAC take-off
Estimating support
Submittal review
Revision control
Management of change
LDAR coverage
PSM documentation
PHA / HAZOP prep
Turnover & handover
Brownfield digitization
Asset registers
The security and scale you expect
Operon runs in an isolated tenant or inside your own network. Your data and everything derived from it stay under your control, and we never train models on it.
View the security overview
Cloud or your own metal
An isolated tenant per customer, or the full stack inside your network for data that can't leave it.
No training on your data
Own your IP
Your data, your context and your outputs are yours. We don't train on them and don't retain them after you leave.
Live in three weeks
Our engineers connect your data in the first week, and your first AI employees go live in three.
Questions about the platform
A connected model of your company, built from its drawings, time series, documents, field records and email. Assets, lines, rooms and jobs are linked across sources, and every fact keeps a link to where it came from. It's what lets AI employees work reliably on industrial data.
A chatbot answers questions. An AI employee does a defined job end to end, like reviewing a submittal or drafting a work order, in your systems and your formats, and hands its work to a person to approve.
No. Operon reads your data where it lives today and builds the context layer on top. Your document control system stays the system of record for files; Operon becomes the system of record for what is in them.
On P&IDs, detection accuracy on symbols, tags and lines runs at 97% and above. More important, every item carries a confidence score, low-confidence items go to a person, and nothing reaches your systems until it has been verified.
Email, file shares, cloud drives, document control, the historian, the maintenance system and ERP. It writes back through those systems, exports, REST, GraphQL and an SDK. Our engineers build the connections each deployment needs.
Yes. The full platform can run inside your network, for data that can't leave it.
Start with the work that costs you the most.
Tell us where your team loses the most hours. Three weeks later, an AI employee built for your company is doing that work.