We messaged 900 dormant retailers. 5.8% reordered in a week.

900 retailers had stopped ordering from an FMCG distributor we work with. Not angry, not lost to a competitor as far as anyone could tell, just quiet. The kind of list a sales team writes off because working it by hand does not pay for itself: no rep has a day free to call 900 accounts that might say no.

So an AI agent worked the list instead. It did not treat every retailer the same way. The default move was the lightest one: a WhatsApp message, in the retailer's own language, asking if they wanted to reorder their usual lines. Retailers who did not respond after a few days were escalated to an AI voice call, the same conversation a field rep would have had, minus the travel time. Every retailer who said yes was logged and moved straight to the order desk. Nobody split the list, nobody assigned territories, nobody wrote a script by hand for each response.

The results split cleanly by channel. Of the retailers WhatsApp actually reached, 6.2% reordered. Of the ones who never answered a message and were then called, 7.5% reordered. Measured against the full list of 900 — including everyone who was never reached on any channel — 5.8% reordered within seven days of first contact. For a list that had already been mentally written off, that is not a rounding error, it is real revenue recovered from accounts nobody had time to chase.

What is more interesting than the one-off number is what happened after. The distributor did not run this as a single campaign and move on. The same approach was turned into a standing workflow: any retailer who goes 30 days without ordering gets picked up automatically and worked the same way, on a loop, with no one deciding week to week which accounts deserve attention. That workflow has now logged 1,994 engagements and keeps running in the background of the business.

Two things are worth pulling out of this. First, the win-back rate is not uniform, voice outperformed chat on the retailers where chat had already failed, which suggests escalation order matters more than picking one channel and sticking to it. Second, the value here was less about winning back 900 accounts once and more about not having to decide to try that again. Nobody had to notice the list had gone stale in the first place; the 30-day trigger noticed for them.

None of this means every dormant list will convert at 5.8% of the whole list. This distributor sells to retailers who already knew the brand and the product, so a light touch was enough to remind them it was still an option. What it does show is that a list too large to work by hand is not the same as a list not worth working. It just needs someone, or something, willing to send the first message.