How to Train Restaurant Staff on a New Menu (Without Burning a Whole Pre-Shift)

Hey Team!

I've rolled out well over a hundred menu updates in thirty years. Seasonal changes, full overhauls, new chef coming in and blowing up the whole thing, ownership change that required a rebrand in two weeks. I've done it every way you can imagine. And I'll tell you what I've learned: the single biggest mistake operators make when a new menu drops is treating it like an event instead of a process. They gather the whole staff, run a pre-shift that's supposed to cover everything, everyone nods, service starts, and within forty-five minutes you've got a server telling table seven the gnocchi is gluten-free when it absolutely is not.

That's not a training failure. That's a sequencing failure.

Here's how I think about it now, after doing this badly for the first decade or so and then slowly figuring out what actually works.

The first thing I do when a new menu is finalized is sit down with it, just me and a printed copy, and I read it like a server who's been on their feet since three in the afternoon and is now trying to describe dish fourteen to a guest who won't stop talking. I'm looking for the things that will trip them up under pressure. Not the things I want them to say. The things they will actually say wrong when they're in the weeds.

Every menu has about ten of them. You know the ones. A French technique word that half your staff will butcher. An ingredient that's regional and nobody's heard of. A protein preparation method your chef spent six years learning and your newest hire thinks is just "kind of rare." A wine pairing with a producer name nobody in Chicago grew up saying out loud. When we added the walnut muhammara to the Black Barrel menu last fall, I clocked it immediately. Muhammara. It's not hard once you've said it, but the first time a server sees it in print at table three during a Saturday dinner rush, they're going to stumble. I flagged it. We drilled it. Not a single guest got a garbled description.

So step one is: make your list. Write down the ten items that will cause problems. Not "servers should know the ingredients." The specific words, preparations, and concepts that are landmines under service pressure.

Step two is doing the drilling before anyone steps foot on the floor. This is the part most operators skip because they think pre-shift is the right time. It isn't. Pre-shift is too late. By the time you're doing pre-shift, you've got servers tying aprons, side work happening, someone calling in and you're reshuffling sections. That's not a learning environment. That's a managed-chaos environment. People retain almost nothing from a ten-minute verbal download when they're also mentally tracking who has the bar section tonight.

What I want is for my staff to encounter the new menu on their phones at home, or in the parking lot before they walk in, or in the fifteen minutes between their regular job and their shift here. Not because I'm asking a favor. Because that's when human brains actually absorb things. When there's no noise. When it's just them and the content and a little bit of idle curiosity. The research on this is pretty clear: spaced, low-stakes repetition before high-stakes performance is how training actually works. This is why I built ShiftTrained the way I did. Upload the menu, the AI generates the quiz, servers take it on their phones before service. The average staff member at Fat Tommy's will take a new menu quiz two or three times before they even clock in, because it's on their phone and it's fast and there's a little competitive element to the scores.

But here's the thing: even if you don't have a platform doing this for you, you can still do the sequencing right. Text your staff the PDF three days before the rollout. Give them a pronunciation guide you typed up yourself. Make a voice memo walking through the ten landmine items and drop it in your group chat. It's not fancy. It works anyway, because the timing is right. Early exposure, low stakes, repeated.

Step three is the pre-shift itself, and once you've done steps one and two correctly, this part gets a lot shorter. You're not explaining from scratch. You're confirming what they already know and filling the gaps. You do a quick tasting if the budget allows. You answer actual questions that came up because people actually engaged with the material beforehand. You spend four minutes on the two dishes that you know need extra context, the ones where the story behind the ingredient matters, the ones where the allergen situation is non-obvious and you really need them to understand why.

Four minutes on the stuff that actually matters beats twenty minutes of hoping everyone absorbed a full rundown at once.

Step four is the verification, and this is the step most operators treat as optional. It's not optional. You need to know, before service, which of your servers can describe the new halibut prep correctly and which ones can't. Not because you want to embarrass anyone. Because the guest at table nine deserves accurate information, and your server deserves not to be put in a position where they're guessing.

A quick quiz does this. Five to ten questions. Focused on the landmine items you identified in step one. Not a comprehensive exam. Just a fast check that tells you who's solid and who needs thirty more seconds of coaching before they hit the floor. When I can see quiz scores before service at Fat Tommy's or Black Barrel, I know exactly where to spend my last five minutes. I walk straight to the two people who got the allergen question wrong and I stand with them for sixty seconds. That's it. That's the whole intervention. But without the data, I'm just guessing.

Now let me tell you what burns a pre-shift. It's when you skip steps one through three and try to do all of it at once. You stand at the front of the room with a new menu and you go dish by dish, ingredient by ingredient, and you can see it happening in real time: the eyes glazing, the nodding that isn't retention, the server in the back who's mentally already at her section. You've given everyone a fire hose of information right before they need to use it, in a noisy environment, with no follow-up mechanism. That's not training. That's reciting.

I've done it that way. A hundred times in the first ten years. I understand why operators default to it. It feels like training. It looks like training. The staff was there, you talked at them, check the box. But information delivery and information retention are not the same thing, and service will always reveal the gap.

The other thing I'd add, because I see this get overlooked constantly: don't treat a menu update as a one-time event. The new menu is live for how long? Three months? Six months? New staff will get hired after the rollout meeting. Seasonal items will rotate. The allergen situation on the salmon might change when the supplier changes their prep method. You need a training system that can update and be redelivered, not a one-time speech that lives in nobody's memory after the first weekend.

A document on a shelf doesn't train anyone. A PDF in an email thread that nobody re-reads doesn't train anyone. Training that lives on the device your staff already has in their pocket and can be retaken, requizzed, and refreshed whenever the menu changes, that's the infrastructure that actually protects your guests and your staff.

The sequence isn't complicated. Identify your ten landmines. Get the material to staff early, on their phones, before service. Use pre-shift for confirmation and context, not first exposure. Verify with a quick quiz so you know where to spend your last five minutes. Repeat as needed when the menu changes again.

That's it. I've been doing it that way for years now and I haven't had a server describe a dish incorrectly to an allergic guest in longer than I can remember. That's the goal. Not a perfect pre-shift. A protected guest and a server who actually knows what they're talking about when it counts.

Have a great day! — Terry
Your AI Restaurant Guy