Ask your AI about your factory.
Claude, ChatGPT, Gemini — the AI your team already uses can now see your jobs, your machines, your orders. And it answers with today's numbers, not guesses.
What is Bulk AI & MCP?
Bulk AI & MCP is one governed connection that lets the assistant your team already uses — Claude, ChatGPT, Gemini, or any MCP agent — see the live records inside Bulk: jobs, machines, orders, quality, and OEE. It answers with today's numbers, not guesses. Reading is free; the moment it wants to write, a person approves first.
What it did this week.
Not a feature list — the unglamorous work. Ask it in plain language and it reads the live thread, then hands back the answer. Here's a week of it.
"Give me the morning brief for Line 2."
OEE, the downtime reason, and the two orders at risk — before the kettle boiled.
"Summarise the open NCRs for the customer call."
A clean summary, with every claim linked back to the record it came from.
"Draft invoices for everything finished this week."
6 drafts prepared…
Held · waiting for your approvalEmployed like anyone else.
Reading is one thing. The moment your AI wants to write — issue that invoice, create that task — it's employed exactly like a person: it carries a badge, it answers to a supervisor, and everything it does goes in a file.
A badge
It works as its user — the same permissions, the same sites, not one lever more.
A supervisor
Reading is free. Writing waits — for a person to say yes or no.
A personnel file
Every action — allowed, held or refused — on the same audit trail as a person's.
Bring the AI you already use.
No new chatbot to learn, no integration project. One connection, and the assistant your team already lives in can see the factory.
Claude 
ChatGPT Codex 
Gemini Claude Code 
Copilot + any MCP agent For IT: your team connects it in an afternoon — sign-in for assistants, scoped keys for coding agents. Anything that speaks MCP works the same way.
Bulk comes with its own AI already on staff.
Alongside yours, Bulk employs its own AI for a few specific chores — on exactly the same terms. It proposes; a person accepts; everything lands on file. It never writes on its own.
It reads the paperwork
Customer POs and delivery notes become structured, reviewed orders — set up per customer.
It drafts the plans
Task plans and KPI / report content, proposed — you accept before anything is written.
It proposes the safety work
Control measures and corrective tasks drawn from evidence — human-approved, every time.
It runs on Google's enterprise Gemini, on Vertex AI — the grown-up, contractual kind. Which means it works where your data lives.
Questions, answered.
What is Bulk AI & MCP?
It is one governed connection that lets the AI your team already uses read the live records inside Bulk — jobs, machines, orders, quality, OEE — and answer in plain language with today’s numbers, not guesses.
Which AI assistants can I use?
Claude, ChatGPT, Codex, Gemini, Claude Code, Copilot, or any MCP agent. There is no new chatbot to learn — one connection, and the assistant your team already lives in can see the factory.
Can the AI change my data on its own?
No. Reading is free, but the moment it wants to write — issue an invoice, create a task — it is held for a person to approve or deny. It never writes on its own.
What permissions does the AI have?
It works as its user, with the same permissions and the same sites as that person — not one lever more. Every action, allowed, held, or refused, lands on the same audit trail as a person's.
How long does IT need to set it up?
Your team connects it in an afternoon: sign-in for assistants, scoped keys for coding agents. Anything that speaks MCP works the same way.
Does Bulk include its own AI?
Yes. Bulk employs its own AI for specific chores — reading paperwork into orders, drafting plans and reports, proposing safety work — on the same terms: it proposes, a person accepts. It runs on enterprise Gemini on Vertex AI.
Bring your AI onto the floor.
We'll connect the assistant you already use to a demo factory. You ask it your questions, approve its first job, and read its file afterwards.