Background
AfvalAlternatief is a Flemish specialist in the collection and management of business waste, headquartered in Dessel. Founded in 2012 by Jo Smolders and Aart Van Oekelen, the company set out to offer SMEs a more personal and flexible alternative to the larger players in the waste market, starting with a single truck and a single driver. Today, AfvalAlternatief supports around 15,000 SMEs across Flanders, with 110 employees and a fleet of more than 40 trucks, and is part of the Bruco Group.
Challenge
Because AfvalAlternatief doesn’t have its own truck coverage everywhere in Flanders, it works with subcontractors in some regions. These subcontractors collect waste on AfvalAlternatief’s behalf, but invoice separately, using their own ERP systems. Each subcontractor works differently: some send a single monthly invoice covering all pickups, running to 40 or 50 pages, others send a separate invoice per pickup, sometimes hundreds per month. On top of that, the layout and naming conventions on these invoices, such as company name, street name, VAT number or customer code, change regularly. Every incoming invoice had to be matched manually against the original order in Business Central, AfvalAlternatief’s ERP system, for example to check whether a customer with a contract for one container, who had two containers picked up, was invoiced correctly for two. This manual check was labour intensive, error prone, and delays could affect both cash flow and VAT administration.
Solution
To streamline this process, CROPLAND developed the Smart Matching Agent, running on the Microsoft Azure platform. The solution performs 3 way matching: it compares data from sales invoices, purchase invoices and contract terms, and combines that information to determine the correct match with an order. The approach is deliberately semi automatic: the AI proposes matches, after which employees can verify the suggested processing and handle exceptions themselves. A key part of building the solution was making the underlying business logic explicit: the pricing agreements and codes used by each individual subcontractor had to be documented and understood in detail, since this context proved essential for accurate matching. Results are presented clearly in an Excel dashboard. Colour codes make discrepancies between documents immediately visible, giving employees insight into differences and into the profitability of a job, without having to compare hundreds of invoice lines one by one. CROPLAND also provided a clear user guide, so employees could use the tool independently. The solution currently runs for one subcontractor, with a rollout to the others underway. It was originally built on a pre trained Microsoft Document Intelligence model; CROPLAND is now migrating it to an LLM based approach, making it easier to onboard new subcontractors with their own invoice formats.
Benefits
Results
Since September 2024, across 24 measurement points, the Smart Matching Agent has processed 82,074 order lines, covering 18,751 unique invoices, 874 client locations and more than €1.4 million in matched purchase amounts. Since the solution has been running at a stable pace, the average share of correctly matched order lines stands at 53%, with a clear growth path towards 80% or more as source data is further cleaned up. For every 100 orders the AI Agent can successfully match and process semi automatically, AfvalAlternatief saves an estimated one full working day of manual review. At a typical volume of 500 to 1,000 orders per month, that represents a potential of five to ten working days saved per month. In July 2025, 512 orders were booked semi automatically, a conservatively estimated, realised time saving of well over five working days that month. CROPLAND’s own root cause analysis of the non matches shows significant further potential: address mismatches, quantity differences and incompletely matched documents together explain 92% of mismatches, and are solvable by cleaning up source data. The initial implementation represented an investment of around €25,000, alongside an ongoing monthly fee for hosting and support. Based on the estimated time savings, that investment has already paid for itself within the nearly two years the solution has been running.
“Thanks to this solution, we’ve tripled the number of automatically processed orders. It not only saves us time, it brings peace of mind.” — Bjorn Anthonis, CFO AfvalAlternatief
