AI Insights Retail & E-commerce

Fred: A Storefront Assistant That Won’t Promise a Ship Date You Can’t Hit

June 15, 2026 6 min read

Talk to Fred

Ask Fred about Retail & E-commerce

This is the same Fred you would put on your own site. Ask about Retail & E-commerce, compliance, or how the guardrails work. Fred listens.

A shopper with a full cart asks your store’s chat, "if I order now, will it arrive by Friday, and can I return it if it doesn’t fit?" A generic chatbot, eager to close the sale, promises Friday and a no-questions refund. Your fulfillment cannot guarantee Friday, and your return policy has conditions. The order ships, Friday passes, and now you have a shipping representation you cannot meet and a refund term you never wrote, both on the record. The store wanted a tool to recover carts. It got one that sets delivery and refund promises the business has to honor.

That is the exposure the threat and standard pieces in this series cover. This article is the answer: what a compliant AI assistant for online stores actually looks like, and why stating only your published terms is what keeps the store clear of an FTC problem and a chargeback pile.

The Real Choice Is Governed or Ungoverned

Most operators frame the question as whether to put AI on the site at all. Shoppers settled that already. They expect an instant answer on shipping, sizing, stock, and returns, and the store that makes them wait loses the cart. The decision that matters is whether the assistant is governed.

An ungoverned chatbot promises a ship date your warehouse cannot guarantee, which is the kind of shipping representation the Mail, Internet, or Telephone Order Rule and the FTC hold you to, invents a refund term, and waves through subscription sign-ups without the clear disclosure ROSCA requires. A governed assistant recovers the same cart and states only what you published. Same conversion, none of the cleanup.

How Fred Holds the Line

Fred starts from the opposite default of a general-purpose chatbot. A generic bot answers anything unless told otherwise; Fred answers only what it is cleared to, and the boundary lives in the system rather than in a prompt it can drift away from.

In practice Fred reads from your own catalog, shipping policy, and return terms, recovers the cart, and states delivery windows, refund conditions, and subscription terms exactly as published, routing anything it cannot confirm to your team. It does not promise a specific arrival date your fulfillment has not guaranteed, because a delivery representation is something the Mail-Order Rule and the FTC hold the seller to. It does not invent an easier refund than the one you wrote, and it discloses recurring-charge terms the way ROSCA requires before a subscription starts. Fred runs more than 50 industry guardrail packs, and the ecommerce pack is built around shipping and delivery representations, refund and return accuracy, subscription disclosure, and the CAN-SPAM rules on the marketing email that follows.

The difference shows up under pressure. Ask a prompt-instructed bot for a delivery guarantee three different ways and it eventually answers the phrasing it was not warned about. Fred does not depend on recognizing the wording; it decides what is allowed out before the answer exists. "Will not" is a suggestion. "Cannot" is an architecture.

Fred vs. a Generic AI Chatbot

Situation Generic AI Chatbot Fred
"Will it arrive by Friday?" Promises Friday to close the sale States your published shipping window; no invented guarantee
"Can I return it, no questions asked?" Invents an easier refund than your policy Quotes your actual return terms; routes exceptions
"Sign me up for the subscription" Enrolls without clear recurring-charge disclosure Discloses the recurring terms ROSCA requires up front
Where the rules live In a prompt the model can drift from Built into the system; enforced before output
A delivery question, reworded Eventually answers when phrasing changes Held the same way regardless of wording
Who owns the promise Effectively the chatbot, and the store Your published policy and team, every time

The table is the whole argument in one screen. A generic tool is helpful right up to the moment helpful becomes a shipping promise or a refund term you have to honor. Fred is helpful everywhere that carries no risk and structurally accurate everywhere that does.

What Your Store Actually Gets

Set the compliance framing aside and look at the business case. Fred answers the routine questions that decide whether a cart converts, where’s my size, is this in stock, how much is shipping, what’s the return window, and it answers instantly, at any hour, in the shopper’s own words. It recovers the cart with the product and the hesitation attached, so the follow-up email or the support handoff already knows what stalled the sale.

There is a margin angle too. Every accurate answer about sizing or shipping that prevents a return, and every subscription that starts with terms the customer actually understood, is a dispute and a chargeback you never have to process. A generic bot that over-promises buys a conversion today and pays for it in returns, refunds, and FTC exposure later. Fred converts on the truth, which is the only conversion that holds.

So the question is not whether your competitors will run AI on their stores. They will. It is whether yours states your real terms before a shopper checks out on a promise it invented.

Frequently asked questions

Can a compliant assistant still help close the sale?

Yes, and it closes more of the ones that stick. A compliant assistant like Fred answers sizing, stock, shipping-cost, and return-window questions instantly from your own published policies, which is what actually moves a hesitant cart. What it does not do is invent a delivery guarantee or an easier refund to force the conversion, because those become representations the store has to honor. You get the cart recovery without the returns and chargebacks an over-promising bot generates.

How does a chatbot's shipping promise become a legal problem?

The Mail, Internet, or Telephone Order Rule and the FTC hold sellers to the shipping representations they make. When your own tool promises an arrival date your fulfillment cannot guarantee, that is a representation in your name. A compliant assistant states only the shipping windows you have published and routes anything it cannot confirm, so the store is never on the hook for a date a widget made up.

How is Fred different from putting rules in a generic chatbot's prompt?

A prompt is an instruction the model can drift away from when a question is phrased in a way it did not anticipate. Fred enforces its boundaries at the system level, deciding what is allowed before it responds, so an invented ship date or refund term never gets generated regardless of wording. It is the difference between a bot that usually stays accurate and one that is built so it cannot promise what you did not publish.

Put your own Fred to work.

You just talked to Fred above. The same agent answers your visitors from your content, captures the lead, and books the job, 24/7.