An AI Is Not a "Retail Employee" — Why the Language Matters

I've been seeing the phrase "AI retail employee" show up in vendor pitches and trade press, and I want to push back on it directly. An AI is not an employee. Calling it one sounds modern and it makes for catchy marketing, but it creates real liability problems and it misrepresents what the technology does. The language we use about these tools matters because customers, employees, and regulators are all listening.

This is a founder opinion piece. Treat it as such.

What an "employee" actually is

An employee is a person who has a contract with an employer, who is accountable for their actions, who can be trained and corrected, who is covered by labor law, and whose interests can be represented by them or by a union. An employee shows up, has agency, makes judgment calls, and is responsible for them.

An AI system has none of those properties. It has weights. It has a system prompt. It has guardrails set by the people who deployed it. It does not have agency, accountability in the legal sense, or interests of its own. When it says something wrong, no one fires it — they edit the prompt or retrain the model.

These are different categories of thing. Conflating them creates problems.

The liability problem

If you market a system to customers as your "AI employee," and the system says something incorrect — recommends a wine that triggers an allergy, gets a price wrong, suggests a product to someone underage — the customer's reasonable interpretation is that an authorized representative of the store told them this. The store is on the hook for the consequences in the same way it would be if a clerk had said it.

Now, the store may be on the hook regardless. If the kiosk gives bad advice, the store deployed the kiosk, and there's a path to liability there. But the framing matters in two ways:

Customer expectations. A customer who is told "this is our AI assistant" understands that it's a tool with limits. A customer who is told "this is our AI employee" expects the same level of authoritative answer they'd get from a human staff member. The expectation gap is where lawsuits come from.

Disclosure obligations. Some jurisdictions are moving toward requiring disclosure when AI is being used to communicate with consumers, especially in regulated categories like alcohol, finance, and healthcare. Calling the system an "employee" obscures the fact that it is AI, which works against you in any disclosure-based defense.

If your customer has to figure out from context that they're talking to an AI, you've already created the ambiguity that makes your legal life harder.

The trust problem

The trust problem is more interesting and more important. Customers, especially in indie retail, choose your store partly because of the people in it. The clerk who knows them. The owner who remembers what they bought last time. The chemistry of the place.

When a vendor tells a store owner that an AI is an "employee" — it suggests the AI is replacing those people. That's not what a good AI tool does, and it's not what good store owners want. The minute customers feel like the store has swapped a human relationship for a robot, you've damaged the thing they were paying you for.

The right framing is the opposite. The AI is a teammate. It's the tablet on the counter that helps the clerk handle more customers without losing the relationship. The clerk is still the heart of the store. The kiosk is one more tool the clerk uses, like the price gun or the POS.

What we call Remi

Internally and externally, we call Remi:

• A concierge. It greets customers, handles questions about the shelf, and points them in the right direction. Like a hotel concierge, it has a role and limits.
• A teammate. When we're talking to store staff, this is the word. Remi is a teammate that handles a class of conversations that the staff can't always get to.
• A kiosk. When we're being literal, this is the word. The hardware is a tablet on a mount.

What we don't call Remi:

• "Our AI employee." This is the framing this post is arguing against.
• "Your new staff member." Same problem, slightly softer.
• "An AI clerk / bartender / sommelier." These imply credentialed expertise that an AI does not have. A sommelier is a person with training and certification. Remi is a recommendation system trained on the store's catalog.
• "Smarter than a clerk." Even if you could measure this on some narrow benchmark, it sets up a comparison the technology will lose at the moments that matter — a customer in distress, an unusual situation, a regulatory edge case.

The vocabulary discipline isn't just brand polish. It's how we keep our heads straight about what we're building.

What this means for store owners

If you're a store owner deploying any AI tool, including Remi, here are the practical things I'd recommend:

Disclose clearly. "Talk to Remi, our AI assistant." Not "Talk to Remi, our newest team member." The first is true and helpful. The second is a marketing flourish that creates exposure.

Train your staff on the limits. Your staff should know what Remi can and can't do, and should be ready to step in when a customer wants a human. The AI doesn't replace your staff's authority — it extends their reach.

Keep humans in the loop on regulated decisions. Age verification, handling of sensitive customer information, anything legally binding — keep a human in the workflow. The kiosk recommends; the staff verify and complete.

Be ready to update or correct. If the kiosk gives a bad recommendation, the path to fix it is fast — update the prompt, update the catalog, set a guardrail. Treat this as a normal part of operating the tool, not a crisis.

Watch your marketing copy. If your in-store signage or your website says "AI employee," edit it. If a vendor's pitch hinges on that framing, push back.

The deeper point

There is a long tradition in tech of marketing tools as if they were people. Chatbots are "agents," automation is a "digital workforce," AI is an "employee." Each generation makes the language a little stronger. The more we lean into it, the more we lose the ability to talk clearly about what these tools are.

Tools have makers. Tools have users. Tools have failure modes. When we describe a tool as an employee, we're not just being whimsical — we're pretending the tool has the properties of a person, which means we're pretending it can be held accountable like a person, which means we're inventing a fictional accountable party that absorbs blame on behalf of the company that deployed it.

That's a useful fiction for the company. It's a bad fiction for everyone else.

I'd rather build a company where the language is honest. The kiosk is a kiosk. The recommendation engine is a recommendation engine. The staff are the staff, and the customers know who they're talking to, every time.

If you want to see this play out in practice — what the disclosure looks like, what the conversation feels like, how staff and kiosk interact — book a demo. And if you want our take on what Remi actually is and isn't, the Remi page is direct about it.

Read also: my piece on the great-grandfather test for the underlying philosophy on what we'll and won't ship, and the comparison piece on self-checkout vs AI concierge for how the kiosk fits alongside the rest of your store technology.

Frequently asked

Isn't "AI employee" just a marketing term?

It's marketing, but marketing has consequences. Customers form expectations from the words you use. Regulators read them. Lawyers in a dispute will quote your own marketing back at you. Treating "AI employee" as harmless is a mistake.

Are there any contexts where calling AI an employee is fine?

Internal jokes among your team, maybe. Customer-facing copy, no. Investor decks, also probably no — sophisticated investors prefer companies that describe what they actually do.

Do customers actually care about this distinction?

Some don't notice. Many do, especially older customers who are skeptical of being misled about whether they're talking to a person. Indie retail's customer base skews toward "actually does notice." Be plain about it.

What's the right way to introduce a kiosk to a customer?

Brief, plain, on-screen. "Hi, I'm Remi — I'm an AI assistant for this store. Ask me anything about what we carry." That's enough. The customer knows what they're talking to and gets on with the conversation.

Will this framing change as AI gets more capable?

The capability question is separate from the labeling question. A more capable AI is still a tool, not a person, until and unless our legal frameworks change to recognize it as something else. We'll adjust the language when the law does.