Demo proposal · Foodservice distribution
The nightly order call is where you lose accounts. A rep runs out of time, a callback slips, an order gets keyed wrong. Mise is an AI voice agent that calls every account on schedule, takes the order in a natural conversation, suggests the reorder, and writes a validated order into your system before the truck rolls.
No system access. No integration. Just 3 accounts + their order guides + 30 days of history, enough to build a working demo on your data.
The problem
Small and mid-size distributors capture the next day's orders through a nightly phone call. A rep dials every account on the route, walks their usual items, and keys it in before a hard warehouse cutoff. It's human, time-boxed, and repeated every single night. It fails in five predictable ways:
Rep forgets or runs out of time before cutoff. Order lost.
Order lost, and it feels like the distributor's fault.
Wrong pack size, wrong SKU, unapproved substitution. Credits and returns.
A rushed call skips the nudge. Margin left on the table, every night.
The rep leaves and the accounts wobble. Key-person risk you can't see.
You feel it as lost accounts and a business that runs on whether someone made their calls.
How it works
A real reorder call: quantities, a history-aware nudge, and a live out-of-stock substitution that never happens without a yes.
Identifies itself and confirms it's reaching the right person, no guessing.
Knows this account's SKUs, pack sizes and price, and captures quantity changes conversationally.
Out of stock, offers an approved substitute at a stated price, and never swaps without an explicit yes.
Uses past orders and par levels to nudge the item they're likely low on, every eligible call.
Reads the full order back and states the cutoff so nothing locks in by surprise.
A structured, validated order in your required format, ahead of cutoff.
What makes or breaks it
Highest value, highest risk. Nothing is ever substituted without explicit approval on the call, at a stated price.
A validated order in your exact format, before cutoff. Anything ambiguous routes to a human review queue.
Price disputes, credits, off-script asks, it captures the request and hands off cleanly. No bluffing.
How we get there
Phase 0 · now
Agent calls a test number, walks a real account's order guide, suggests a reorder, handles one substitution, and outputs a structured order file. Proves the experience.
No integration neededPhase 1
One design-partner distributor. Real accounts, real cutoff, a human reviews every order. We watch the missed-order rate drop toward zero.
Phase 2
Full ERP write-back. Human review shrinks to exceptions only. Account coverage expands across routes.
Phase 3
Demand forecasting, suggested ordering, AR follow-up and load-sequencing, all riding the order flow you now own.
What changes
What we need from you
Exported spreadsheets, CSV or Excel, are perfect. No system access, nothing for your customers to learn. The realer the data, the more convincing the demo.
Your demo readiness
Check off what you can send. The three starred files alone are enough to get started.
Short on time?
Just send 3 accounts + their order guides + 30 days of history. That alone is enough to demo the agent taking an order and suggesting a reorder.
A handful of real accounts on the same route. Anonymize the names if you like.
The items each account normally buys. This single file is what lets the agent sound like it knows them.
Last ~30 days, or the last 5–10 orders per account. This powers the reorder suggestions.
Product info for the items in those guides, including in/out-of-stock if you have it.
A few out-of-stock pairs, e.g. romaine → green leaf, same price. Lets us show the most impressive part of the call.
A short call recording, a script, or just the phrases reps use, so the agent sounds like your business, not generic.
And whether it can export guides & history and import orders. Not needed for the demo, it tells us what writing orders back automatically looks like.
Format: CSV or Excel, most ordering systems export these directly. Anonymize customer names if you prefer; the structure is what matters.