Documents, emails, databases.
One memory your agents run on.
Connect your sources once and the memory builds itself. Answers with their source, warnings before something slips, and agents that do the follow-up for you.
EU-hosted · Your data never trains AI
Nobody can read it all. So things slip.
Your ERP knows what was typed in. Your document system knows where the files are. Neither connects a supplier’s delay email with the order, the milestone and the customer invoice behind it. fluctor does.
RE: delivery week 22, PO 4471
Dave Corbett <dispatch@corbett-fasteners.co.uk>
Tue 19 May, 09:12
To: purchasing@yourcompany.com
Morning,
the 40 units for PO 4471 won’t make week 22 after all. Earliest we can ship is Wednesday the 27th. Sorry for the short notice.
Dave
Dave Corbett · Dispatch · Corbett Fasteners Ltd
THIS EMAIL
PO 4471 · 8 days late
FEEDS
Order #88 · 40 units
BEHIND IT
Milestone 3 · Halden Steel Ltd · 30 May
AT THE END
invoice_2088 · €38,000 · slips to July
Connect everything once, and every new document files itself.
Every document is read as it arrives and filed under the customer, order or project it belongs to, even when none of them mention each other. You end up with one file per customer, with every source still attached.
invoice_2041.pdf
INVOICE · PDF
Order #88
ORDER · PDF
framework_2024.docx
CONTRACT · SHAREPOINT
Delivery #88-1
DELIVERY · SCAN
Payment · open
OPEN ITEMS · SQL
RE: order #88, partial delivery
EMAIL · OUTLOOK
QC_batch_88.xlsx
QUALITY · EXCEL
Site acceptance #88
PROJECT · DRIVE
invoice_2041.pdf
INVOICE · PDF
Order #88
ORDER · PDF
framework_2024.docx
CONTRACT · SHAREPOINT
Delivery #88-1
DELIVERY · SCAN
Payment · open
OPEN ITEMS · SQL
RE: order #88, partial delivery
EMAIL · OUTLOOK
QC_batch_88.xlsx
QUALITY · EXCEL
Site acceptance #88
PROJECT · DRIVE
Halden Steel Ltd
CUSTOMER · ONE FILE
Delivery #88-1 was signed for 14 days ago. Invoice #2041 belongs to order #91. Nothing has billed order #88: est. €12,400 outstanding.
Ask how it looked on any past date
Walk it through once. That is the whole setup.
Describe the task in your own words. fluctor does it for real while you watch, and stops for your approval before anything leaves the house or changes a record. What you approve becomes the agent. Every step that reaches outside is checked against your instructions before it happens.
It does the task for real.
Reminder for order #88 drafted for Halden Steel Ltd · €12,400 · waiting for your approval
You approve it.
The steps it took become the agent: check open invoices → delivery note on file? → draft the reminder → send
Every run after follows those steps.
Anything that reaches outside is checked against your instructions first · no improvising
It can start from a schedule, a signal, an incoming email, a webhook, or a single click. How an agent is built
When something is missing from a file, it says so.
Because everything about a customer sits in one file, a gap in it stands out: a delivery with no invoice behind it, a date in a contract coming up, a customer whose orders have dropped off against their own history.
REVENUE
A delivery went out and no invoice followed.
Order #88 · signed 14 days ago · no invoice on file · est. €12,400 open
CUSTOMER
A customer has gone quiet against their own history.
Marlowe Freight · this week: 2 orders · a normal week: 20
DEADLINE
A date buried in clause 12 is coming up.
framework_2024.docx · price review due in 19 days
25+ signal types · one case per problem, not forty alerts · dismiss one and it learns
Every answer names its page.
Ask the way you would ask a colleague. The answer comes with the document and the exact line it was taken from. If nothing in your data matches, it says so rather than guessing.
Has order #88 been invoiced yet?
No. Delivery #88-1 was signed for on 14 May. There is no invoice on file. Est. €12,400 outstanding.
“…received in full, 40 units.”
delivery_88-1.jpg · PO_88.pdf, page 1
Say what you want to keep an eye on, and the board builds itself.
One sentence is enough. Every tile is checked against your live data before it is saved. Forecasts are built from this morning’s invoices and tested against what actually happened before you see them.
“The five biggest overdue invoices, by customer.”
one sentence → three numbers, one chart, one table
Overdue
checked against live data · 12:04
Overdue, total
€41,600
Older than 30 days
3
Oldest
37 d
Overdue, last 90 days
EUR · net
Made for businesses that run on paperwork.
Orders, deliveries, contracts and the emails in between, spread across several systems and several parties. If that is how your business runs, this is built for you.
Manufacturing & Suppliers
One chain per order, from confirmation to payment.
Construction & Plant Engineering
Contract, change orders, acceptance, final invoice.
Trade, Wholesale & Logistics
Shipments, delivery papers, unbilled freight.
Project Business
Three years, four project leads, one memory.
Service & Maintenance
Contracts, service reports, SLA deadlines.
What it reads stays yours.
Software that reads your contracts and your mail has to be held to a higher standard than the systems it reads from. These are the terms, and they are in writing.
Hosted in the EU, processed in the EU
All AI processing runs on EU endpoints. No call ever goes to a US model.
Your data never trains a model
That is written into our privacy policy.
Every organization fully isolated
Kept apart in every record, from your first upload.
A record nobody can edit
The audit log is kept for seven years and cannot be deleted, not even by an administrator.
A second EU provider on standby
It takes over automatically when the first one is at capacity.
See it on your own paperwork.
Half an hour, live, on documents you bring along. If it finds nothing worth your time, that is all you have lost.
No integration project to start · Your data stays in the EU