How a Plantel
pilot works.
A pilot is a paid, scoped trial of Plantel on one line or one use case in your plant, for at least 20 days. We start from the exports and systems you already have, agree one number to measure against before we start, and at the end you decide whether to go on. Nothing is replaced and nothing is ripped out.
What a pilot is
Plantel is an AI layer designed to sit over the software a plant already runs: ERP, production, SCADA and historian, maintenance and quality systems. The aim is to read across them, so the people who run each department can ask a question directly instead of asking someone to put a report together. What it connects to today is a shorter list, set out under What we connect first.
A pilot tests that on a small, defined piece of your operation first. It runs for at least 20 days, and it is paid as a running cost rather than a capital purchase. What it costs depends on how many systems we connect and how much setup work is needed. We set that with you after a first call.
Before we start, we agree one number with you: the hours per week your team spends moving data between systems by hand. You give us the figure. It is the same question the self-check asks under manual data movement. At the end we ask the same question again, so the pilot is judged against your number rather than one we chose. How that figure is counted, and what it is worth, is set out in Counting the hours.
What we need from you
- Access to the systems in scope Read access to them, or the exports they already produce. We ask only for what the agreed use case needs.
- Recent exports or reports A few from those systems, so we can see how your data is really laid out before we connect anything.
- One named person Someone who knows the systems and can say what a field means when it is not obvious. Plantel asks when it is unsure rather than guessing, and this is who answers.
- About an hour of their time a week Checking figures and answering questions about fields. Most of those questions come early, while the connections are set up.
- Your starting number The weekly hours figure above, written down before anything changes.
What we connect first
A pilot starts from what we can connect today, not from everything Plantel is designed to sit over. There are two routes.
- Files your systems already export. CSV and Excel files, including the reports that production and machine systems produce. Your team uploads them, checks a preview of how each column was matched, and confirms before the rows are added.
- Direct connections. We have these for online spreadsheets and databases, accounting and ERP software, maintenance, fleet tracking, sales and reporting software, registers kept as spreadsheets, and some accounting software that runs on your own premises, through a small program on one of your computers (more on that below).
The direct connections were built first for one kind of plant. Each one is set up for the records that kind of plant keeps, so for most plants a direct connection needs setup work before it can be used. Before a pilot starts, we tell you which of your systems we can connect and how, what setup work that needs, and how far each connection has been tested against the real product.
We do not yet connect directly to machines or control systems. Where machine data matters, a first pilot works from the reports those systems already export.
What you get, and when
- Setup. We connect the agreed sources, then check with your named person that the figures Plantel reads match what your team knows to be true.
- Questions across your systems. Once the first sources are in, your team can ask questions in plain language about what happened and what is happening, and get answers from the connected data. Each answer is built to state the period its figures cover.
- Throughout. Where a value cannot be read with confidence, or two systems disagree, Plantel asks your team instead of choosing for itself, and remembers the answer when asked to.
- At the end. A written readout, which you keep: what was connected, what could and could not be answered, and your weekly hours figure asked again, beside the starting one.
The numbers are worked out by fixed calculations, the same ones behind Plantel's screens, and the AI puts them into words. When something is not measured, the AI is built to say so rather than estimate it.
No forecasting on day one. Plantel answers questions about what happened and what is happening. Questions about what to plan next need a clean, connected history, roughly six months of it. The pilot is where that history starts.
Paper has no history to read. If your logs are kept on paper, there is nothing yet for Plantel to work from. A pilot can start the record. It cannot recover what was never written into a system.
No direct machine connections yet. A first pilot works from the reports those systems export.
Not every direct connection fits your plant yet. They were built first for one kind of plant, so one may need setup work before it can read your records.
It does not replace anything. Your ERP, production, maintenance and quality software keep doing their jobs.
How your data is handled
- Where it runs is agreed with you before the pilot starts, and so is which AI model provider writes the answers.
- The connections that bring your data in are built to read it. They are not built to change anything in your systems. Logins you give us are stored encrypted, and where your software allows it, we ask for a read-only one.
- How a connection reaches your software. For software on the web, Plantel's servers read it at its web address. If that software runs on your own server, the server has to be reachable from the internet for this, and we agree that with you first. For software that stays inside your network, a small program on one of your computers sends the data out to us instead. Nothing dials in to that program, and all it gets back is a delivery status.
- Sending messages is separate, and only if you ask for it. One kind of connection sends reports, alerts and reminders from your own business messaging account to the people you choose. Only people with manager or owner access can send through it, and the AI cannot send a message on its own.
- The AI can use only the tools we have opened to it. Most of them only read. The few that can add something, such as a new record or a new measure, only propose it: a person on your team approves it before anything is saved.
- The AI cannot invite people, change who has access, or change your settings.
- Each person sees only the departments they have been given, and the AI answers each person with that person's access.
- Your data is kept separate from every other client's.
How we handle information about the people we talk to is set out in How we handle your information.
How a pilot ends
At the end we go through the readout with you, and you decide: carry on, change the scope, or stop.
If you stop, your systems are as they were, because nothing in them was replaced or changed. You keep the readout. If a small program was installed on one of your computers, it is removed. Within 30 days of the pilot ending, we delete the data Plantel holds for you and confirm that in writing.
Who we are
ProArc is an AI company, and Plantel is what we build for manufacturers. You work with the founders directly.
We started inside a family-run pharmaceutical business, automating its own operations. Every system there worked. The gap sat between them, and closing it by hand took skilled people away from better work. Plantel is built for that gap.
We are taking on a small number of first pilots. That is why this page has no case studies yet: everything on it describes how a pilot is set up and what the software does today, and none of it is a customer result.
Start with the self-check
It asks how your systems, machines and registers actually hold information, and gives you a score and the first things worth fixing. That report is the most useful thing to bring to a first call.