Answer product questions from the right version of the facts.

Customers get an answer from the current product facts, with the version named, and an honest “we don’t say that” when the facts stop.

Start with a support review How it works

Product facts · CAT-042Canvas field jacket

Fact	Source	What we say
Machine washable	Supplier spec, revision 2	Yes, on cold
Waterproof	Not claimed	No rating, never claimed
Colors	Catalog	Olive, Navy, Sand

Version 2 · owner: merchandiser

The approved facts for one product: where each came from, and what isn’t claimed.

Sound familiar?

The same three questions fill the chat, and the answer depends on who’s on.
Last season’s spec is still quoted for this season’s product.
The assistant answers anything, including things you’ve never claimed.

A confident wrong answer costs a return and the customer’s trust. An honest “we don’t say that” costs a day.

What changes.

TodayWith the engine

Source: PDFs, product pages and memoryOne set of product facts, with owners and versions
Version: Whichever spec someone foundThe version in force, named in the answer
Gaps: Filled with a confident guessSaid out loud, and sent to the owner
Claims: Anything the assistant can phraseOnly what you’ve approved, including what you don’t claim

How it works. 5 steps, 4 stages.

AI2
Rules and automation2
Your team1
Approval

1Ask
1. AI
  Understand the question
  Works out which product and which version the shopper means.
  - Gorgias
2Find
1. Rules and automation
  Find the facts in force
  Looks up the approved fact, its source and its version.
3Answer
1. AI
  Answer in the shopper’s words
  Says what’s claimed, and what isn’t.
2. Rules and automation
  Hold unsupported claims
  If no approved fact covers it, the answer says so and a person is asked.
4Fill gaps
1. Your teamApproval
  Add the fact, or decline it
  The owner adds a sourced fact, or decides the product won’t claim it.

Each gap becomes a fact, or a decision not to claim

What you get. Real work, every week.

Will the canvas jacket shrink in the wash?

Fernway doesn’t publish that, so I won’t guess. I’ve asked the product team and you’ll hear from us by tomorrow. What I can say is that it’s machine washable on cold.

No match in product facts, version 2 · sent to the merchandiser

When the facts stop, the shopper is told so.

Question not coveredOwner: merchandiser

Does the canvas field jacket shrink?

Asked: 6 times this week
In the facts: Nothing yet
Needed: A line from the supplier spec
Then: Add it, or decide not to claim it

The questions the facts can’t answer yet, in a queue for the person who owns them.

How you’ll know it’s working.

Answers that match approved facts
Checked against the source on a sample every week.
Questions that reached an owner, and how long they waited
A gap is only fixed when someone owns it.
Repeat questions on the same product
Shows which facts the page and the answers still don’t carry.

Baselines agreed before we build.

Works with what you have.

Support

Gorgias
Zendesk
Intercom

Commerce

Shopify

We configure first: Shopify’s AI staff member and your helpdesk’s AI already answer from your store and help center. We check what they cite and tidy the sources first.
We build when: Answers need facts your storefront copy doesn’t hold, such as supplier specs or versions, or you need to show which version an answer used.

What a first build includes.

Built in your accounts, around your team. Scope, fee and baseline agreed before we start.

1Design
2Build
3Run it, or hand it over

Start with a review Scope this build

We build

A register of approved product facts, with sources and owners
Answers that name the version and the source
A gap queue for questions the facts can’t answer
Test questions for the products that matter most

We start from

Product specs and supplier documents
A month of product questions
Helpdesk and store access

From your team

Owner: CX lead, with the merchandiser
Involves: Customer service
Merchandising
Catalog

Questions teams ask.

What if the facts don’t cover the question?

The shopper hears that, and the question goes to the fact’s owner. Questions that keep coming become approved facts, or a decision not to claim them.

Do customers need to know it’s AI?

For EU-facing shops, the EU AI Act’s transparency rules for AI that talks to people apply from August 2, 2026, with an exception where it’s obvious. We build disclosure into the first message, and you should confirm your own scope with counsel.

Who owns it once it’s running?

You do. It’s built in your accounts, documented, with training for the people who run it. Ownership usually sits with your CX lead, with the merchandiser. We can keep running it with you, or step back.

How do we start?

Usually with a support review: two to four weeks, a fixed fee credited to the build, and a plan that’s yours. If the job is already clear, we can scope the build directly.