Meridian FoodsA wholesaler doubles its capacity.

Every morning, Meridian's back office drowns in documents. Orders arrive by email, fax, text photo, voicemail. Invoices get keyed by hand. Delivery confirmations and supplier paperwork get reconciled line by line. Three people spend most of their day moving information from one place to another.

So that's where we start. Not because it's exciting. Because it's the most obvious, reclaimable burden in the building. An AI-native intake system reads every inbound order and invoice, normalizes it, matches it to the catalog, flags exceptions, and posts clean data straight into the system. The three people aren't gone. Their day is.

This is the move most automation stories skip. The temptation is to bank the headcount and move on. Don't. Those three people know the customers, the catalog, and exactly where orders go wrong. They're the most valuable input the engineering team has. So they get reassigned: sitting with engineers to build the next thing.

What do they build? A custom ordering surface for each restaurant. Not a generic portal. A control surface tuned to each kitchen: standing orders that match the menu cycle, par-level reminders, one-tap reorders, substitutions when something's short, pricing they can actually see. Ordering from Meridian becomes easier than ordering from anyone else.

That convenience compounds. Existing restaurants order more often and churn less, new ones switch because the experience is obviously better. Demand rises, and now there's a new constraint to chase.

Rising demand pushes the constraint downstream. Purchasing is now the pinch point: buyers guess quantities and get caught short. So the team builds a purchasing surface that turns cleaned order data into demand signals, suggests buy quantities, and watches supplier lead times. Then it's delivery's turn: a routing surface that sequences trucks against the day's actual orders.

None of these are moonshots. Each is a focused control surface for one role, built fast because building is now cheap, and funded by the capacity the last one freed. This is the rolling investment in motion.

Now it gets interesting. Intake, ordering, purchasing, delivery, and collections were each built to lift one constraint. But they all write to the same place. Once that data lives together, the combined system does things no single automation was designed for, abilities that simply weren't available before.

Ordering data plus inventory plus supplier lead times becomes auto-replenishment. The full order history becomes demand forecasting that tells buyers what next week looks like. Orders plus routing becomes dynamic delivery routing that re-sequences trucks as the day changes. None of these were on the original roadmap. They fell out of the combination.

The back office that drowned in paper now runs the exception queue and the customer relationships. The sales team spends its reclaimed hours selling. Restaurants order more and leave less. The company carries materially more volume without a proportional increase in headcount. The constraint on growth, lifted. And every stage paid for the next, so the whole thing ran on a small rolling investment, not a single large bet.