When we talk to a company about optimising its processes, the conversation usually starts with what should be added: a new module, a new integration, a new report. Once we connect to their system and look at the data, the picture is almost always different. A missing feature is rarely the problem. The problem is money that gets stuck between departments, because nobody sees the whole story.
Quotes live in sales, deliveries in the warehouse, invoices in accounting, complaints in an inbox. Each department has its own version of the truth, and none of them shows that a customer received the goods but never got the invoice.
Below are the seven patterns we find most often. The numbers come from real companies whose systems we have reviewed, and they are anonymised.
1. Overdue invoices without a reminder
Finding: one manufacturer had EUR 1.2 million in overdue receivables. The payment reminder module was installed — and switched off.
This is not unusual. A common scenario: reminders are set up during implementation, then switched off after one goes to a customer who had already paid, or because the letter sounded too harsh. Nobody switches them back on, and chasing payments becomes something accounting does when it has time.
What we build: overdue invoices get reminders on a schedule, with escalation steps and templates that you define. Difficult cases go to a person, and every reminder sent is logged on the invoice.
2. Quotes nobody follows up on
Finding: more than 1,000 sent quotes worth EUR 3.4 million in total, older than 30 days, with no follow-up of any kind.
A salesperson sends a quote and moves on to the next one. If the customer does not reply, the quote sits in the system as sent — not rejected, not accepted, and nobody looks at it again.
What we build: every quote gets a scheduled follow-up, for example after 3, 10 and 30 days, until the customer responds. The message refers to the specific quote, and the reply is routed back to the salesperson.
3. Delivered, but never invoiced
Finding: 598 delivered orders without an invoice at one company. At another, a single order worth EUR 24,000.
Between the warehouse and accounting there is a handover that in many companies depends on a person: someone has to notice that the goods have left and issue the invoice. When that person is ill, on holiday or overloaded, deliveries pile up without invoices — and nobody notices, because the warehouse sees the delivery as done and accounting sees nothing missing.
What we build: delivered orders create draft invoices the same day. A daily overview lists everything that got stuck between delivery and invoicing, and unusual cases are flagged for review.
4. Opportunities that never leave the first stage
Finding: 93% of 400 open opportunities never left the first stage. Nine out of ten had no planned next step.
A CRM without next steps is a list of names. An opportunity gets entered because there was a call or an enquiry, and then nobody decides who does what next, and when. Such an opportunity does not die — it simply never moves forward.
What we build: an assistant qualifies every new opportunity, plans its next activity and drafts a first reply for review. When a deal stalls, it reminds the owner and attaches the context of the deal.
5. Delays hidden among stale records
Finding: 2,066 overdue warehouse transfers, with real delivery delays hidden among them.
When there are thousands of overdue transfers, the list loses its meaning. Many of them are stale — cancelled orders, duplicates, transfers nobody closed. But among them are deliveries that really are late, and those can no longer be found.
What we build: an assistant closes stale records, merges related issues into a single record and highlights the deliveries that really are late. The result arrives as a short morning overview.
6. Customer tickets that wait for months
Finding: between 370 and 3,000 open tickets per queue. The oldest had been waiting for seven months.
The answer to many tickets already exists somewhere in the system — in the customer's history, in a similar earlier ticket, in the product documentation. The problem is not that the team does not know the answer, but that it has to look it up again every time.
What we build: tickets are sorted and routed as they arrive. An assistant drafts a first reply from the history in your ERP, and your team reviews and sends it. Special cases go straight to the person responsible.
7. Abandoned carts in the online shop
Finding: 128 abandoned carts worth EUR 29,000 in 60 days. The cart recovery tool was installed and unused.
A shopper who has already put a product in the cart is as close to buying as it gets. They often do not finish for entirely everyday reasons, and without a reminder they do not come back.
What we build: an abandoned cart triggers an email written around what is in the cart. The timing adapts to the customer, and the flow stops on purchase or unsubscribe.
What these findings have in common
In two of the seven cases, the tool that solves the problem was already installed — payment reminders and cart recovery. Money was not leaking because the company needed a new system, but because nobody had finished setting up the existing one, or it had been switched off along the way.
In the rest, the cause is the same: every step depends on someone noticing something. A quote needs someone who remembers to call. A delivery needs someone who notices the invoice is missing. An opportunity needs someone to set the next step. When there are too few people or too much work, these steps are quietly skipped — and they show up in no report, because reports show what happened, not what did not.
That is why optimisation in these cases does not mean replacing the system. It means handing the steps that depend on someone remembering over to automation — with human approval wherever you want it.
How we arrive at these numbers
We connect to your system read-only — to the ERP, email, online shop and bank. Nothing is migrated and nothing is replaced. An AI assistant reads the connected data and reports where money is leaking. Each finding can then become a working automation that uses the same connected data, so each one is quicker to build than the last.
If you would like to see what your numbers look like, we offer a free 30-minute review of your system. We connect read-only and go through the findings together with you. You decide whether they are worth fixing. Book a review.
Where companies lose money in their ERP: seven findings from real systems