Tech has to support the Guest Experience, not supplant it. Fourteen waves of receipts.

Since 1982, my position on technology in our business has been the same.

Tech has to support the Guest Experience, not supplant it.

That year was the first time I heard the industry beat the drum that more tech was going to create the guest experience. Cash registers, dot-matrix printers, paper dupes to the kitchen, a reservation book on the host stand. The trade press was already running “computers will transform restaurants” copy. The pitch was that the next wave of technology would solve the human work.

It didn’t. It hasn’t. And it isn’t about to.

Forty-four years later, I’m watching the same play run with new equipment. So before I get to AI — which is what most of us are actually arguing about right now — let me walk the receipts.

The waves

Mid-1980s — networked POS systems. Squirrel, Micros, Aloha. Sold as the end of paper, the end of mis-rings, the end of kitchen miscommunication. Verdict: supported the GX. Removed friction. Earned its seat.

Late 1980s / early 1990s — Kitchen Display Systems. Screens replacing paper tickets at the line. Pitched as the answer to ticket times and expo accuracy. Verdict: supported when it amplified an already-built kitchen. Failed when it became the kitchen’s brain. The screen doesn’t run the line. The expo does.

1990s — back-office software. Inventory, scheduling, accounting integration. Sold as “control your costs from your office.” Verdict: supported when used as a tool. Supplanted when used as a substitute for being in the building.

Mid-to-late 1990s — caller ID and early online ordering. Pizza chains first. Verdict: supported the transactional side. Didn’t touch the GX where the GX actually lives.

Early 2000s — handheld POS and pay-at-the-table. Wireless ordering devices for servers. Verdict: supported when servers stayed servers. Supplanted when the device became the relationship.

Mid-2000s — OpenTable scales. Reservation books retired. Verdict: supported the guest’s access to the reservation. Whether it supported the GX itself depended entirely on what happened after the guest walked through the door.

Mid-to-late 2000s — loyalty platforms and CRM. Fishbowl, Paytronix. Verdict: supported when the operator used the data to deepen the relationship. Supplanted when the operator outsourced the relationship to a points program.

2008–2012 — cloud POS arrives. Toast, Square Register, the iPad wave. Verdict: supported — lowered the entry bar for independent operators. The first wave I’d give a clean win to since networked POS.

2010–2015 — third-party delivery launches. GrubHub-Seamless, DoorDash, Uber Eats. Verdict: supplanted. Reframed the restaurant as a fulfillment center for someone else’s app and moved the guest relationship onto someone else’s platform. The operator stopped owning the moment the guest decided whether to come back.

2015–2018 — online ordering as table stakes. Direct platforms plus the four-tablet station behind every counter. Verdict: mixed. Direct ordering supported. Tablet aggregation hell supplanted, because the operator started running the host stand for four apps instead of for guests.

2017–2021 — ghost kitchens and virtual brands. CloudKitchens, Reef, Kitchen United. The spreadsheet thesis: rent is killable, dining rooms are killable, brands are spinnable. Verdict: supplanted entirely. The whole model removed the GX from the equation by design. It collapsed because the math didn’t disappear — it moved.

2018–2022 — QR-code ordering, kiosks, tip-screen culture. Verdict: split by segment. QSR kiosks supported. Full-service QR-only ordering supplanted. The server isn’t a noun. Removing the server doesn’t remove the cost — it moves the cost onto the guest, who now does the work and feels it.

2020–2023 — COVID-driven tech surge. Whole industry retooled in 18 months. Verdict: necessity, not strategy. Most of what stuck was already on the supported side; most of what’s been quietly walked back was on the supplanted side.

2021–2024 — labor automation. Robot fry cooks, voice-AI drive-thru, automated phone answering. Verdict: supplanted, mostly. Sold as the answer to the labor crisis. Most of the early deployments have been quietly pulled. The reason is simple. The line wasn’t broken because no one was running the fryer. The line was broken because no one was running the leadership.

2023–present — generative AI. The current drumbeat. Verdict: pending — and that’s the actual subject of this post.

The rule

Fourteen waves. One stance. The line that has held across all of them is what I call the [Amplification Principle]: Legitimate tool amplifies a capability already built. Substitution fails. Build first. Amplify second. Subtract nothing.

That’s the test. Run any technology through it. If the tool amplifies a capability the operator has already built into the building — the people, the standards, the systems, the way the room is run — it supports the GX. If the tool is being asked to replace a capability that was never built or has been allowed to atrophy, it supplants the GX. Substitution fails.

Underneath that sits the [Subtraction / Addition Master Test]: does this tool add something the guest can feel, or does it subtract something the guest was getting? If the guest ends up doing more of the work, paying more for less contact, or talking to a screen instead of a person who used to be there — that’s subtraction sold as addition. The math doesn’t disappear. It moves. Usually onto the guest.

And the floor underneath both of those is the [Connection Floor] — the floor of what every human touchpoint delivers. Tech that protects the floor supports the GX. Tech that lowers the floor supplants it.

Apply it to AI

The question isn’t whether AI is good or bad for restaurants. That’s the wrong question. It’s the same wrong question the industry has asked about every wave on the list above.

The right question is [Authority To Execute]: where does AI get the call, and where does the human keep it?

AI that compresses prep-time math, schedule building, inventory variance, guest-data lookup, marketing copy, social drafting — amplifies. Supports the GX. Earns its seat. Run that side hard.

AI that takes the call — voice replacing the welcome, algorithm replacing the kitchen manager reading the room, model replacing the lead reading the cast, dashboard replacing the operator reading the line — substitutes. Supplants the GX. Loses the table. The Connection Floor drops. The guest feels it. The math moves onto the guest.

That’s not anti-AI. That’s pro-judgment. The tool earns its seat by what it amplifies, not by what it claims to replace.

The line

Same line in 1982. Same line through every wave on the list. Same line on AI.

Tech has to support the Guest Experience, not supplant it.

If the tool amplifies what you’ve built, run it. If the tool is being sold to replace what you haven’t built — or to substitute for what you’ve allowed to atrophy — the tool isn’t the answer. The work is.

The drumbeat will keep going. It always does.

The rule doesn’t change.

This article is based on the chapter: Enable, Don’t Supplant from my upcoming book, The Operator’s Playbook.