18 Steps

Between the moment something goes wrong and the moment somebody fixes it, the information changes hands eighteen times.

An instrument records a drift. A shift engineer notices it. It goes into a log, into a handover, into a morning meeting. Someone raises it with maintenance. Maintenance raises a notification. Planning schedules it. Procurement checks the spare. It comes back for approval. Somewhere in there, the lab result that explains the whole thing is sitting in a different inbox.

6–8
hops · the best-run plants
18
hops · the average
30–35
hops · the worst

Every hop costs time, and every hop loses context. That is where the money goes — not in the fault, but in the distance between noticing it and doing something about it.

We are named after the eighteen hops. We built the company to remove them.

How we think about this

The problem was never a lack of data

A modern plant records everything — thirty thousand tags in the historian, every assay in the lab system, every notification in SAP, every shift written down. The problem is that no single system, and no single person, reads all of it together.

So we did not build another monitoring product

There is no shortage of software that watches one machine and tells you a bearing is failing. We built the layer above it: a system that reads across the historian, the lab, maintenance, materials and the shift logs at once, and reasons about what they mean together — the way a plant head would, if he had the time and all of it in front of him.

Nothing gets installed

No sensors, no control-system connection, no capex cycle. We read what your plant has already been recording for twenty years. That is why deployment takes weeks, not quarters.

And the human stays in charge

The system diagnoses, quantifies and drafts. Your engineers decide. On a shop floor, an AI that acts on its own is not an innovation — it is a liability.

Who you are dealing with

Practitioners first.

Sandeep Kumar — Founder

Digital Expert at McKinsey & Company  |  Digital Practice Lead at Accenture  |  10+ years spent optimizing and digitizing some of the world’s largest plants, manufacturing setups, and mines — including Asia’s first fully integrated mine-to-smelter control tower, spanning five mines and three smelters on a single operating picture.

Author of CFO Niti, a best-selling book on financial leadership.

The people who do the work

Industrial AI fails when it is built by people who have never stood on a shop floor. So the work here is done by practitioners first.

Process & metallurgy

Engineers who have run kilns, roasters, cell houses and boiler islands, and who know why the textbook answer is often the wrong one at three in the morning.

Reliability & rotating equipment

Careers spent on vibration, lubrication and failure modes, and the judgement to tell a real signal from a mill start-up.

Instrumentation & OT

The historian, the DCS, the tag dictionary, the twenty-year-old naming convention nobody documented. Somebody has to know where the data actually lives.

AI & data engineering

Senior engineers who deploy on site, not from a distance.

Every answer the system gives at a new plant is reviewed by a practitioner with thirty years in that specific process before an engineer sees it. That is slower than shipping software. It is also the only way an industrial buyer should accept an answer.

What we will not do

The lines we do not cross.

Connect to your control system
Change a setpoint
Touch an interlock
Advise on isolation, permits, confined space or hot work — those go to your SOP and your responsible authority, every time
Use your plant’s data to train anything
Let your data reach another company, in any form

These are not preferences. They are how the system is built.

See what your plant looks like from the outside.

Ninety days of your own data, read together for the first time. Under NDA. Nothing touches production.

Request a plant scan