Pharmacy: clearing the third-party receivables backlog before it turns into a write-off

In a pharmacy, almost all activity runs through third-party billing. The counter handles a continuous flow of invoices whose financial counterpart never arrives the same day, and rarely in a single payment: one share goes to the public health insurance scheme, the other to the complementary insurer, with different lead times, references and channels. The consequence is structural: at any given moment, a pharmacy carries a backlog of receivables awaiting payment, living its own life behind the daily flow.

That backlog is rarely managed. Most teams handle what comes in, fix the day's rejections, and leave behind a residue of files that could not be settled in a few minutes. This residue does not go away: it accumulates, it ages, and it eventually forms an aged receivables balance nobody looks at until the accountant flags a gap between invoiced revenue and actual cash collected.

Before working the backlog, it helps to understand that it does not contain one population but three, with different causes, different treatments and different urgency. The first covers invoices that never entered the payment chain at all. The second, invoices that entered and were rejected. The third, invoices that were genuinely paid but never reconciled in the accounts. Conflating them is the main reason balance clean-up campaigns fail.

The first population is the most expensive, because it is invisible. After electronic transmission, the payer returns a technical acknowledgement indicating whether the batch was accepted. A negative acknowledgement means the invoice is not rejected in the business sense: it simply does not exist on the payer side. No payment return will ever contradict it, and no rejection will ever flag it. Unless someone explicitly monitors these negative acknowledgements, the receivable disappears from view on the day it is issued and resurfaces only at the annual stocktake, usually past the deadline.

The second population, rejections, is better known but poorly bounded. Easy rejections are fixed within the day: entitlements to refresh, an exemption to requalify, an obvious data-entry error. Rejections that require a supporting document, a call to the prescriber or input from the patient get set aside with the intention of coming back to them. Those are exactly the ones that age, and their difficulty grows over time: the patient has not returned, the prescription has been filed away, and whoever held the context has moved on.

The third population is different in nature: the receivable is settled, the money is in the bank, but the entry was never cleared. This case is heavily concentrated on the complementary share, paid by third-party platforms that bundle dozens or hundreds of files into a single transfer, with no patient identifier on the bank line. Until the transfer is broken down, the corresponding files stay open. The risk is not only accounting: chasing organisations that have already paid wastes time and damages the relationship with the platform.

The first operational move is therefore to date the backlog and sort it by cause and by paying organisation, never by patient. Sorting patient by patient produces an unmanageable list of unique cases. Sorting by cause produces homogeneous batches: all negative acknowledgements from a given period, all rejections carrying the same code, all files pending with the same platform. On an ageing backlog, these batches are usually highly concentrated: a handful of causes explain most of the volume, and a single correction replicates across dozens of lines.

The second move is to prioritise by deadline, not by amount. Every billing channel has a limit beyond which an invoice can no longer be resubmitted, set by regulation or by the agreement applicable to the organisation concerned. Those limits are not uniform, which is precisely why they should be checked channel by channel rather than assumed. Past the limit, a file is no longer a receivable to recover: it is a loss to record. Working the large amounts still recoverable in six months while letting small amounts expire next week is the most common prioritisation mistake.

The third move is to decide write-offs explicitly. Not every receivable is worth recovering: on lines worth a few euros, reprocessing time exceeds the amount at stake. Setting an owned threshold, writing it down, and taking those lines out of scope is better than letting them clutter the balance indefinitely and demoralise the team. What matters is that it be a decision, taken once and documented, rather than abandonment through exhaustion.

This is the kind of work where an AI agent genuinely changes the economics of the task, because it is high-volume, repetitive and fully rule-based. The agent navigates the practice management software, the payer web services and the insurer portals using the same access as the team, rebuilds the complete backlog, sorts it by cause and by deadline, and prepares the matching correction or resubmission for each batch. On the complementary share, it reconstructs grouped transfers by matching them against expected receivables, which clears in one pass files that had been open for months without ever having been unpaid.

The principle is the same as across the rest of the chain: the agent prepares, the human validates. On an ageing backlog this split is particularly well suited, because the reconstruction work accounts for most of the time and involves no judgement, while the decision to resubmit, chase or write off stays short and must stay human.

The real benefit is not the one-off catch-up, however visible it is in cash terms. It is that the aged balance becomes a monthly indicator, with a target for residual backlog rather than an annual observation. A pharmacy that knows, every month, how many receivables are over sixty days old, for what cause and with which organisation, stops rebuilding the backlog it has just cleared. For a buying group or a multi-site structure, this consolidated view also exposes gaps between locations, which almost always come from a local billing practice rather than from the patient mix.