Nobody at HKR checks the bank for new payments. Every hour from 7:00 to 20:00, the AI checks Wise and Banca Transilvania and tells the team who paid. A person still marks each payment paid.

Every hour from 7:00 to 20:00, the AI checks Wise for new incoming transfers. Banca Transilvania sends a text when money lands, and the AI reads those texts in the same pass. Each payment is checked against the open invoices.

The result goes to the team’s Slack channel: who paid, how much, and which invoices it clears. On Wise, a transfer between two HKR companies is named as an internal transfer and matched to no invoice. A person still marks each payment paid in Zoho Books.

Nobody at HKR works out by hand which invoices a payment clears. The AI clears the oldest invoice first and asks a person about any amount that does not fit. A person marks each payment paid.

A payer’s name rarely matches the name on the invoice, and a partial payment rarely matches an invoice total. The AI keeps a copy of the open invoices, refreshed four times a day, and finds the client who sent each payment. When the payment could clear several invoices, the oldest goes first.

When the amount does not fit an invoice total, the AI shows how it would split the payment and asks a person to confirm. When it cannot tell who paid, it says so and leaves the payment for a person to place. A person marks each payment paid in Zoho Books.

Nobody at HKR collects balances bank by bank. The AI refreshes Wise and Revolut each working hour, reads Banca Transilvania’s statement when it arrives each day, and shows the age of every balance. A person names the owner of any account it does not know.

HKR banks with Wise, Revolut and Banca Transilvania, across three companies and four currencies. Wise and Revolut balances refresh every working hour. Banca Transilvania sends one statement a day, and the AI reads it when it arrives. A few accounts are typed in by hand. Every balance shows how old it is, so nobody reads yesterday’s number as today’s.

Each account belongs to one HKR company. When the AI finds an account it does not know, it holds that account until a person names its owner. The question “how much do we have” gets one answer.

Nobody at HKR copies numbers out of the books for a report. Four times a day, the AI copies invoices, expenses, customers and payments from Zoho Books, so every report reads the same numbers. People still keep the books and decide what goes in them.

HKR keeps its accounts in Zoho Books. Four times a day, the AI copies the new and changed invoices, expenses, customers and payments into one database. Reports read from that copy, so two people who ask the same question get the same number.

The copy only reads the books. Invoices, corrections and every other entry stay in Zoho Books, with the people who keep them.

Nobody at HKR builds the overdue list by hand. On the 5th and the 15th, the AI shows the whole team every overdue invoice and how late it is. People do the chasing.

Twice a month, on the 5th and the 15th, the AI posts every invoice past its due date in the team’s Slack channel, with how many days late each one is. Everyone sees the same list on the same day.

The AI only lists. People decide who to chase, when and how, and a person records each payment when it arrives.

Nobody at HKR reads every bill for surprises. On the 18th of each month, the AI checks each expense against what that vendor usually charges and lists what stands out, with the reason. A person decides what each one means.

A large bill tells you nothing until you know what that vendor usually charges. On the 18th of every month, the AI checks each expense in the books against the vendor’s own history: a single bill far above its usual size, a vendor that was not there in the last six months, a software tool that stopped charging.

The flags come back in Slack as one list, alerts first, then warnings, each with its reason. A person decides what each one means. The AI reads the books and never writes to them.

Nobody at HKR re-reads past months for bills that came in late. On the 18th, the AI fetches three months of expenses again, so late bills count, and ranks what moved, with the reason. A person decides what each change means.

A bill booked late changes a month that already looked finished. On the 18th of every month, the AI fetches the last three months of expenses again and scores them on nine checks: category swings, large single bills, capital spending that drifts, expenses without a category label and more.

The result lands in Slack: the total, the change month on month and year on year, and the lines that moved, each with its reason. A person decides what each change means. The AI reads the books and never writes to them.

Nobody on HKR’s finance team digs through folders and the books for an answer. They ask the AI in Slack, and it answers from the company’s records: contracts, payments, spend and payroll rules. The person who asked decides what to do with the answer.

The finance team asks in Slack, in plain words: what a contract says, whether a client has paid, what we spend on a vendor, what a payroll rule allows. The AI looks it up in the books, the signed contracts and the company’s other records, and replies in the same thread.

Contracts come up most often, then payments, follow-up tasks, spend and payroll rules. The AI finds and reports. The person who asked decides what the answer means.

Nobody at HKR ties salary and contractor payments to people by hand. Every week, the AI ties each payment to a person and lists the ones that do not tie. A person resolves each one on the list.

Every Monday, the AI goes through the salary and contractor payments in the books and ties each one to the person it paid, using the HR records.

A payment it cannot tie goes on a short list, and a person resolves each one. When every payment ties, no list is sent, so a list arrives only when there is work on it.

Nobody at HKR searches folders for a signed contract. Every signed document is in one index, so “what did we sign with this client” is a question the AI answers. A person decides what the contract means.

Every day, the AI walks the folders where HKR files its signed documents and adds each new one to a single index, with who signed it and what kind of document it is.

So “what did we sign with this client” is one question in Slack, with no folder search. The AI finds the documents and says where they are. A person reads them and decides what they mean.

Nobody at HKR rebuilds the forecast by hand. On request, the AI rebuilds it from the books and lays cost-cut scenarios over it on one chart. A person picks the plan.

The forecast is rebuilt only when someone asks for it, from the books as they stand that day.

Cost-cut scenarios sit over the forecast on one chart, so each option is compared on the same numbers. The AI lays out the options. A person picks the plan.