ERP complexity debt: how workarounds pile up and why an AI-native ERP removes them
What is ERP complexity debt?
Most finance teams run on a set of tools that each do one job: one or two bank accounts, an expense management tool for cards and receipts, an accounting tool that is often managed externally by the tax advisor, an invoicing tool embedded in the CRM, a payroll and HR tool, and spreadsheets that bring all of this together for reporting. Sometimes there is an ERP in the middle, built around the standard process of an industry. None of these tools are built around how your company actually works. So wherever the business does not fit the tool, or two tools do not talk to each other, someone on the finance team has find a workaround and fill the gap by hand, usually with a spreadsheet.
Each of these workarounds makes sense on the day it is introduced, the problem is that it stays. Every workaround adds a few minutes to every transaction or every month-end close that follows, it usually lives in one person's head, and it grows with the business. That recurring manual work is what we call complexity debt: you saved time when you set the workaround up, and you pay interest on it every month for as long as it exists.
Where does the complexity show up for the finance team?
In most growing companies, the same pattern comes up: spreadsheets and manual steps connecting systems that do not talk to each other. Here is where it shows up most often.
Booking the same transaction twice. Several companies we spoke to book every transaction twice, once in their ERP and once in DATEV for their tax advisor, because the two systems are not connected.
Matching payments to invoices by hand. Your bank account records every payment going in and out. Your invoicing tool holds the invoices you send to customers, supplier invoices arrive by email or in your accounting tool, and your expense management tool records card transactions and their receipts. None of these tools are connected, so someone goes through the bank statement and matches every payment to the invoice or receipt it belongs to, one by one. The tools do not make this easier: one company told us their expense management tool only exports the invoice date, not the payment date, so every card payment has to be found and checked by hand. Another told us that moving each receipt into their accounting tool takes about four minutes.
Business models your tool stack cannot represent. A standard setup books an invoice as revenue in the month it is issued. If your customers pay for a year upfront, that revenue belongs to twelve months, so finance keeps a schedule in a spreadsheet and books a correction by hand every month. The same happens with deposits, projects with milestone payments and grants: the system follows a standard process, the business does not, and the difference is handled in Excel. In one case, customers sign three-year contracts but are still billed every three months, because the middleware between the company's systems cannot do anything else.
Reporting in a structure the books do not have. Bookkeeping in Germany sorts costs by type, following the SKR03 or SKR04 chart of accounts: salaries, rent, software, travel... Investors usually want the same costs sorted by what they are for: cost of sales, sales and marketing, research and development, and admin. A developer's salary and a salesperson's salary sit on the same salaries account, so to split them you need to know which team each person works in. That is what cost centres are for, and when they are not kept in unison with the books, someone splits salaries, software and other shared costs by hand before every board report. One company we spoke to had built this split with AI-generated scripts in Google Sheets, as a workaround, which worked until a team or an account changed, and it all had to be rebuilt.
Why does adding another tool not fix it?
The usual reaction is to buy a tool for the gap, such as a connector between two systems or a planning tool for the forecast. Each one removes a manual step, but it also adds a new system with its own data, its own mapping to maintain and a new place where numbers can disagree. And the spreadsheet stays. The reports the board and investors ask for every month or quarter, the P&L, cash position, forecast and KPIs, are still put together by hand in Excel and slides, from an export of every tool.
Moving to a large traditional ERP does not fix it either. It can be adapted to almost anything, but the adaptation is custom development: code written for your company on top of the standard product, which you pay for, document, test again with every update and maintain yourself. For a growing company, that means a long and expensive project, and a system that is hard to change as your business scales and changes.
What does a workaround cost?
One workaround on its own costs very little, a step that adds ten minutes to each month-end close costs two hours a year, and twenty such steps add up to a working week. Steps that repeat for every transaction cost far more: four minutes per receipt at 300 receipts a month is 20 hours, every month.
And because the work grows with the number of transactions, the only way to keep up as the company grows is to hire.
To take stock, list every recurring manual step in your finance processes and ask four questions about each one: how often does it happen, how long does it take, who is the person who knows how to do it, and what happens if it is skipped. The steps that repeat every month, take real time and depend on one person are the expensive ones, so remove those first. And limit new debt before it is added: when someone asks for a new field, report or tool, ask first whether the existing setup can already provide it.
How to remove it with an AI-native ERP
Removing complexity debt means two things: one place where all bookings live, and a system that fits your company instead of the other way round.
Agent F is an AI-native ERP that brings the data from your bank accounts, expense management tool, invoicing tool and payroll tool into one ledger. Instead of fitting your company into a standard template for its industry, it is configured to your business model through what we call the ERP Factory: modules are built around how you actually work, whether that means annual contracts, deposits, projects or grants, and they change as your company changes. That configuration is part of the product, not custom code you have to own and maintain.
Inside that ledger, the manual steps become an approval flow, every transaction comes in with the booking suggested: the matching invoice, the account in your chart of accounts and your own cost centre logic, each with a confidence score. Your finance team approves or corrects it, and corrections feed into the next suggestions, so the rules that used to sit in a spreadsheet are kept in the system and the same mistakes do not accumulate over time. The reports your board and investors ask for are built from the same ledger and exported directly as a PDF or presentation, as we explain in our board reporting deep dive.
Which of your workarounds costs your team the most time each month? Chances are we are solving it :)
Learn how Agent F works, and book a demo at agent-f.ai/demo, to see it run on your numbers.