Self-Checkout vs AI Concierge — Different Problems, Different Tools

A store owner asked me last month whether installing Remi meant they could skip self-checkout, or vice versa. The answer is neither. Self-checkout and an AI concierge solve completely different problems — one speeds up the transaction, the other improves the decision before the transaction. Most modern stores eventually want both. The interesting question is in what order, and which one to start with.

The two problems, named clearly

Transaction friction. A customer has chosen what they want. They are now standing in line waiting for a cashier to ring them up. The line is the problem. Self-checkout is the answer.

Decision friction. A customer is looking at a shelf. They aren't sure which product fits their need. They might leave with the wrong one, or leave with nothing, or wait twenty minutes for a clerk to be free. The shelf-side uncertainty is the problem. An AI concierge is the answer.

Self-checkout does not help the customer who can't decide. An AI concierge does not help the line. Anyone selling either as a universal solution is overselling.

What self-checkout actually does for indie retail

Self-checkout is mature technology at this point. It does one thing well: it lets customers who already know what they want move through the line faster, which means fewer customers walk out because the line was too long.

For indie retail specifically, self-checkout has tradeoffs that big-box has mostly accepted but indie has not:

• Theft / shrink. Self-checkout shrink rates are higher than staffed checkout. Big-box absorbs this; a margin-tight indie store often can't.
• Age verification. A liquor store cannot run a fully unattended self-checkout. Every alcohol transaction needs a staff override. The "self-checkout" turns into "self-scan, attendant approves," which is a smaller win.
• Cash handling. Many indie stores still take a meaningful share of cash. Self-checkout for cash is more expensive hardware and more failure modes.
• Customer relationship. A regular at an indie liquor store often wants the thirty-second chat with the owner at the register. That chat is part of why they don't shop at the big-box. Self-checkout removes the chat.

Self-checkout makes sense in indie retail when the line is genuinely a problem at peak (e.g., the Friday-night convenience store rush) and when the staff override workflow is acceptable to the team running it. Otherwise the gain is small.

What an AI concierge actually does

An AI concierge — the Remi model — handles the conversation that would otherwise require a knowledgeable clerk who isn't available because they're at the register or in the back room.

Customer asks the kiosk a question. Kiosk answers from the store's actual inventory, with a reason. Customer either takes the recommendation, asks a follow-up, or walks off having decided. The decision happens earlier in the visit and with better information.

The difference from self-checkout is that the AI concierge's value scales with how many decisions are open and unresolved on the floor. In a store where most customers walk in and grab the same six SKUs, the value is limited. In a store where customers ask "what should I bring," "what's like this but cheaper," "what pairs with this" — the value is significant and shows up in average ticket and repeat visits.

Worth reading: our comparison piece on AI upsell vs POS prompts walks through why a kiosk-based discovery conversation outperforms a register-based prompt, and the solutions page for liquor stores goes deeper on category-specific use cases.

Where they overlap (and don't)

Concern	Self-Checkout	AI Concierge
Reduces line at the register	Yes	No
Helps customer choose the right product	No	Yes
Helps in age-restricted categories	Limited (needs override)	Yes — Remi can recommend without checking ID, ID stays at register
Captures discovery questions for inventory planning	No	Yes — store sees what customers asked for
Replaces a clerk	Replaces transaction labor	Does not replace; supplements
Hardware cost (per lane)	Several thousand to tens of thousands	Hundreds (consumer tablet + mount)
Customer trust impact	Mixed — some customers prefer it, some resent it	Positive when the recommendation is good
PCI scope impact	High — handles payment	None — no payment on device

The overlap is essentially zero. They solve different problems and have different cost profiles.

When to deploy self-checkout

You probably want self-checkout if:

• You have a real line problem at predictable peaks (rush hour, weekend nights, lottery days).
• Your category is dominated by quick, transactional purchases (convenience, gas-and-go, grab-and-go food).
• You have the staffing to monitor the lanes and handle age verification overrides.
• Your shrink tolerance is workable for the category.

You probably do not want self-checkout, or want only a single lane, if:

• You're a wine shop, specialty foods store, or any store where the conversation at the register is part of the experience.
• Your average ticket is high and the cashier double-checks every transaction anyway.
• You're a liquor store where every transaction needs ID verification and the override workflow erases the speed gain.
• You're already managing tight shrink and don't want to add a vector.

When to deploy an AI concierge

You probably want an AI concierge if:

• Customers regularly ask staff "what should I get for X" — and staff are often unavailable to answer.
• You carry a deep catalog where most customers don't know your full range.
• You serve gift-giving occasions, pairing decisions, or recommendation-driven categories (wine, craft beer, spirits, specialty grocery).
• Your store has a quality differentiation story (knowledgeable staff, curated selection) that's currently bottlenecked by the staff being at the register.
• You want to capture discovery data — what customers asked for that you didn't have — to inform purchasing.

You probably do not need an AI concierge if:

• Your store is purely transactional (gas station with a small cooler, no curation).
• Your customers walk in, grab one of three SKUs, and leave.
• You have abundant floor staff already handling discovery.

When to deploy both

Most full-service stores eventually want both. The order I'd recommend:

Start with the AI concierge if your bottleneck is "customers leave without buying because they couldn't decide" or "average ticket is lower than the catalog supports." This is the more common indie problem. The hardware cost is low (under a thousand dollars per kiosk), the install is under an hour, and the impact shows up in the metrics within weeks.

Add self-checkout when your line at peak is genuinely costing you walkouts and the labor math justifies the hardware. This is usually a higher-volume store problem.

Run them together in larger stores. The kiosk handles discovery on the floor; self-checkout handles transactions at the front. The clerk handles the conversations and decisions that need a human.

If you want to see how Remi looks in the discovery role, book a short demo — it takes about 15 minutes and you'll see the actual store-specific catalog flow, not a generic mockup.

A note on hybrid models

There's a hybrid pattern that's tempting and that I'd caution against: combining the AI concierge and the payment terminal in the same device, so the kiosk does the conversation AND takes payment.

We deliberately don't do this. Two reasons:

PCI scope. Putting payment on a customer-facing internet device drags the kiosk into PCI compliance, which is not a problem you want for an under-$1,000 install.

Trust. A device that is both recommending products and asking for payment has a conflict of interest the customer can feel. Keeping recommendation and payment on separate devices keeps the recommendation honest.

So the right architecture, in our view: kiosk for discovery, existing POS or self-checkout for transaction. They don't have to be the same hardware.

Frequently asked

Will an AI concierge slow my line?

No, it operates on the floor before the register. If anything it shortens the register interaction because the customer arrives knowing what they want.

Can self-checkout in a liquor store actually be unattended?

No. State law in essentially every U.S. jurisdiction requires a human to verify ID for alcohol sales. Self-checkout in a liquor store always has an attendant on the lane.

Which one has a faster ROI?

Different shapes of ROI. An AI concierge tends to show up in average ticket and discovery data within weeks. Self-checkout shows up in labor savings and walkout reduction at peak. The former is usually faster to measure in indie retail because the baseline (no kiosk) is so different from the after state. We walk through worked examples in our ROI piece.

Can the kiosk help with age verification?

The kiosk does not verify ID directly — that stays at the register where the cashier is trained for it. The kiosk can recommend age-restricted products without checking ID, the same way a salesperson on the floor would. Verification happens at purchase.

What if I'm a small store with limited budget?

Start with the AI concierge. Hardware is cheap, install is fast, and the upside is on the discovery side, which is where most indie stores leak revenue. Self-checkout is a bigger commitment and worth it later if your line problem is real.