You have the data.
What you are missing is someone to read them every day.
Hiring an analyst is expensive and they leave. A new dashboard ends up like the others: nobody opens it.
Patok Cognitive is an analyst who lives in your inbox. You ask, and it replies with the report already built —with charts, with the detail spreadsheet, and with the recommendation— in the same thread.
Any ERP knows how to email a CSV. That is enough to start — no API, no connector, no IT project.
Inbox · 07:00
Today's operation · Aug 4 · 12 overdue, 4 due today
Replenishment 304-2B · order 94 t this week
Weekly summary · 4.1 d cycle · 87% on time
The subject line already carries the alert. You do not have to open anything to know whether today needs action.
You send it an email.
It replies with the report already built.
It is not a chat, and not a dashboard you have to go open. You send the question —or your spreadsheet— and the answer arrives in the same thread.
- To
- analyst@yourcompany.patok.in
- CC
- purchasing · warehouse
- Subject
- Replenishment 304-2B
Good morning. How much do we need to order of 304-2B this week? I am attaching the last three months of consumption.
You write the way you would write to an analyst. No commands, no formats, no logging into any system.
The engine calculates
consumption 3.5 t/day
transit 20 days → 69 t
safety stock σ√L → 25 t
ROP = 25 + 69 ≈ 94 t
A deterministic engine does the math and draws the dashboard. The model does not compute the number — it only reports it.
- De
- analyst@yourcompany.patok.in
- To
- you · purchasing · warehouse
- Subject
- Re: Replenishment 304-2B
Replenishment
304-2B · cutoff Mon 07:00
Order 94 tons this week. No 60, no 120.
You consume 3.5 t/day and the material takes 20 days to arrive: that is 69 t in transit, plus 25 t of safety stock. ROP = 25 + 69 ≈ 94 t. Safety stock scales with σ√L —not in a straight line— which is why doubling the lead time does not double the buffer.
Place the order today. Waiting until Monday leaves you 4 days uncovered.
I am attaching the detail by gauge and the consumption history.
01 · The question
You write to your analyst's address the way you would write to a person. You can attach your spreadsheet.
02 · The engine
The math and the dashboard are done by proven code, not by the model. That is the difference between an analyst and a generator of pretty text.
03 · The answer
It arrives in the same thread, with everyone on CC included, and with the dashboard and the detail spreadsheet attached.
And if the answer claims an attachment that does not exist, or a figure the system cannot compute, it gets corrected before it goes out.
Whoever receives it
managed by replying.
The reason reports die is that changing them means asking someone. Not here: you reply to the report in plain language and it adjusts itself.
And nobody can touch anyone else's: identity comes from the sender of the email, never from what the model says. Anything that deletes or cancels also asks for confirmation.
And also:
Someone with nothing assigned gets nothing sent to them. The recurring empty email is exactly what makes people stop opening them.
You reply to the email:
“Send it to me at 7 in the morning”
Changes the schedule of your delivery. Only yours.
“Northern plant only”
Builds your own filtered version, without touching anyone else's.
“Drop the quality section”
Adjusts the content and sends you a sample to confirm.
“Stop sending it to me”
Unsubscribes you. You, not the report.
“I want a new one on stockouts by store”
Designs it with you over email, shows you a sample, and schedules it.
Your ERP gives you today's snapshot.
We give it memory.
Your system's export has no history: it is a flat extract, with no yesterday. Patok compares each file with the previous one, and that is where the questions that actually matter come from.
What came in
What shows up for the first time in today's file and was not there yesterday.
What disappeared
What stopped coming — it is flagged, not deleted, because sometimes it is an export error.
At what rate
How much is consumed per day, calculated against previous extracts.
How long it lasts
Days of coverage at the current rate. The question a flat extract can never answer.
It works the same with what arrives by email, with what is in your cloud, and with the data your floor already generates — because the engine is the same.
Where it gets the data.
From what you already have. Nothing needs to be installed in your ERP and no port has to be opened for anyone.
By email
Your ERP sends its export
A dedicated address where your system drops the file, scheduled or by hand. It is the shortest path and the one we almost always use to start.
From your cloud
The folder you already use
We mount the folder where your team already drops files and it gets checked automatically every few minutes.
From your floor
What Patok Gemba already captures
Work orders, lots, downtime, inspections, movements, and the log — with nothing to export, because they already live in the same place.
What you have to tell it once
Which column is the committed date, which is the amount, and which is the customer. That is the only real configuration, and it is done once. From then on, every new file comes in on its own.
One-click questions
And one rule that is not negotiable: the model may propose a plan, never claim that something happened on the floor.
Connect your plant
to Claude or ChatGPT.
If your team already works with an AI assistant, Patok connects as one more source: your operational data becomes available there, with nothing to export and no copy-paste.
And it is not just the data that travels: so does the analysis criteria — how to read a queue, when downtime matters, what it means when variability rises.
And if the data does not exist?
Patok Gemba generates it
Downtime with reasons, queue by machine, lot traceability. Captured by your people with a sticker and one tap — and from there it goes straight to the analyst.
And if you want to go further?
An engineer we set up for you
When the analysis needs industrial engineering judgment —experiments, variability, capacity— we build it with you and it becomes a routine.
Send us a file.
We send back the first analysis.
A CSV, a spreadsheet, whatever your system exports. We tell you what can be answered with it — and what cannot, which is usually the most useful thing to know.
Start with a file · Nothing to install · No ERP changes