How does an e-commerce business use inTusell? A daily operations guide

In the first article, you trained the inTusell AI: you uploaded your product catalog and FAQs to the knowledge base, marked your sales conversations, and configured the persona and response rules. Training was a one-time setup. Now the real work begins. How does an e-commerce business use inTusell — that is, how does an e-commerce business's day-to-day use of inTusell actually work? This article walks step by step through what the assistant you trained does in the field over a full workday, from the incoming DM to an order, and from shipping tracking to returns and handing off to a human.

This article is the second part of the e-commerce section of our sector-by-sector "how to train your AI" series. The first part, how to train the inTusell AI, covered the setup. This one covers turning that setup into daily operations.

Quick answer

A day in the flow

The typical day of an e-commerce team that has set up inTusell looks very different from before. The customer rep used to come in each morning and work through dozens of unread DMs and unanswered comments one by one. Now the messages that came in overnight are already answered, leads are categorized, and some carts are already recovered. The backbone of a day works like this:

The team no longer spends its time scanning messages. It spends it on conversations that genuinely need a decision.

Who is it for?

This usage model especially makes sense for e-commerce businesses that:

• Receive heavy product questions via Instagram and WhatsApp — those who can't keep up with dozens of "do you have this, how much is it?" messages a day.
• Lose after-hours demand — operations that miss sales because they respond late to questions coming in the evening and on weekends.
• Have a high volume of recurring questions — brands where size, color, shipping, and return questions arrive in the same pattern.
• Manage many channels with a small team — firms forced to track DMs, WhatsApp, comments, and email separately.

When demand outgrows what a single rep can handle, inTusell meets it like a senior salesperson.

Incoming request: the customer asks about a product, the AI recommends a suitable one

The flow starts when the customer sends a message. Imagine someone writes via Instagram DM: "I'm looking for navy sneakers, size 42." The AI reads the message, scans the product descriptions and catalog in your knowledge base, recommends a product that fits the customer's need, and directs them to the cart or product page.

One important point: the AI responds in whatever language the customer writes in. For brands selling abroad this is critical — a product question that comes in English gets an English reply, with no extra configuration.

If you added the "clarify the need before recommending" response rule during training, the AI first asks about size, color, and intended use. That rule keeps the recommendation from being random; it rests on the customer's real need. The Instagram and WhatsApp automation article explains separately how all these channels come together in a single inbox.

Order, shipping, and return questions: answered from the knowledge base

In e-commerce, most incoming messages aren't new sales but recurring operational questions: "How many days does shipping take?", "How do I make a return?", "Is there an exchange on this product?" They cycle in the same pattern all day and eat up the rep's time. inTusell takes on this load by answering from the knowledge base.

The most important rule here: the AI does not invent price or stock. It answers only from the source you uploaded to your knowledge base.

If someone asks about information that isn't in the knowledge base — the stock of a size you didn't upload, say, or the live shipping status of a specific order — the AI does not invent an answer. It says it doesn't know and offers to connect the customer to the live team. That keeps you from giving the customer wrong stock or price information and then having to correct it, and it is the foundation of trust.

The order and product questions automation article gives a detailed account of this engine — answering everything from orders to shipping and from returns to product questions automatically, out of the knowledge base.

How the knowledge base is fed and stays up to date

Whether the AI answers correctly depends entirely on how current your knowledge base is. inTusell feeds the knowledge base from several sources and indexes all of them for pgvector-based search (RAG):

• Product catalog: Uploaded as PDF, Excel, or CSV; product names, descriptions, size/color options, and policy texts come from here.
• Web URL: You provide the address of your returns page, shipping FAQ page, or product pages; the content is fetched and indexed.
• Free text: You type notes like "the X campaign applies this season" directly into the panel.

Stock and price change often, so refresh the knowledge base whenever you update your catalog. The AI always speaks from the latest source you uploaded — if the source isn't current, neither is the AI. That is why a fresh catalog is the backbone of operations.

Recovering the cart: proactive follow-up and CRM

In e-commerce, most sales dissolve among customers who add a product to the cart, say "let me think about it," and disappear. inTusell doesn't forget these leads. When a customer gets information or gets stuck at the cart and then goes quiet, a proactive follow-up message goes out and keeps the lead alive. Every conversation is logged in the CRM, so the customer's history, the product they were interested in, and the stage they reached all stay on record.

One boundary deserves a clear line: inTusell does not maintain a managed "waiting list." What it does is send an opportunistic re-offer to eligible WhatsApp customers in follow-up — a customer who left an item in the cart might get a reminder along the lines of "The item in your cart is still available, can I help?". When you sit down to operations in the morning, you can see straight away which lead is hot and which one is waiting for follow-up. Nobody has to keep a manual reminder list.

Appointment and reservation flow (if needed)

Most e-commerce flows revolve around orders, but some businesses also use appointments and reservations: a showroom visit, a product consultation call, an installation or assembly appointment, or a trial session. inTusell's universal appointment engine works in these cases too.

The AI doesn't offer times off the top of its head. It looks at the working hours and the busy slots you defined during training, and if it finds a suitable slot, it creates the appointment. An automatic reminder goes out for each appointment (1 day and 2 hours before, for example), and the customer manages their own appointment through a /manage-appointment/{token} link, cancelling or rescheduling in one click. Appointments are copied one-way to Google Calendar (inTusell → Calendar), so the team sees them in their own calendar. You can enable this module on any plan, and if you don't need it, you simply don't use it.

Handing off to a human: lock modes

You decide how autonomously the AI works. There are three lock modes:

In hybrid mode, the AI hands off a "where is my order?" query it can't answer, or any request that falls outside the knowledge base, to the live team. At the moment of handoff, the team can see the full conversation history, so the customer doesn't have to start over. You can change the mode at any time — ai_only on a campaign day, for example, or human_only in a sensitive complaint flow.

One more note: a customer sending a file — a product photo, an invoice, a shipping label — is not on its own a reason to hand off to a human. The AI receives the attachment, understands the context, and continues the flow. It escalates only when the situation genuinely needs a decision.

Reports: what's working, what to fix?

To keep operations visible, you track basic metrics in the panel: how many messages came from which channel, how many the AI answered, how many were handed off to a human, and which questions recur the most. Seeing these serves two purposes:

1. Finding knowledge-base gaps: If the AI often says "I don't know" about a topic and hands it off, that information is missing from the knowledge base. Add it, and the handoff rate drops.
2. Improving response rules: You review the tags and correct misclassified conversations, and the AI learns from those corrections. You can also compare two different response approaches with an A/B test.

Reports, then, aren't just a summary. They are the feedback loop that sharpens the assistant over time.

What it isn't

To place inTusell in the right category, let's clarify what it isn't:

• It's not a chatbot. It's not a flow bot that answers from preset templates; it's a fully autonomous AI sales assistant that works with the knowledge and sales techniques you trained into it.
• It's not an e-commerce platform or a payment provider. It doesn't replace your store; it directs customers to your catalog and your checkout page, and the business remains the seller.
• It's not a stock and price source. It invents nothing. It speaks only from the source you provide, says when it doesn't know, and hands off.
• It's not an e-commerce-specific priced product. The pricing model is based on messages and voice minutes and doesn't change by sector. You can enable the modules on any plan.

Knowing these boundaries from the outset keeps you from starting with the wrong expectation. For product and pricing details, see the solutions and pricing pages.

Frequently asked questions

Does the AI complete an order on its own?

The AI recommends the product, answers size, color and stock questions from the knowledge base, and takes the customer all the way to the payment step. In most setups, payment goes through the store's own checkout page or a cart link. In risky situations, or ones that fall outside the knowledge base, it hands the conversation off to the live team. Final approval can always stay with the business.

Does the AI invent stock and prices?

No. The AI gives price and stock status only from your knowledge base and the source you provide. It never invents a size, color, or stock detail it doesn't know; it says it doesn't know and hands off to the live team. This is the most important trust rule.

How does it answer shipping and return questions?

You upload shipping times, return conditions, and the exchange policy to your knowledge base, and the AI answers these recurring questions straight from that source. When something outside the system is needed — an order code, a carrier API, or live shipping status — it hands the conversation off to the live team.

Can a customer be connected to a human?

Yes. In hybrid mode the AI runs the normal flow and escalates the conversation to the live team when needed. In human_only mode all conversations go straight to the team, and in ai_only mode the AI handles everything. You can change the mode at any time.

What happens to a message that comes in at night?

The AI responds 24/7. A "do you have this size?" question that arrives at 02:00 gets an answer from the knowledge base within seconds, and where it fits, the AI directs the customer to the cart. When you open up in the morning, ready leads and answered questions are waiting for you.

Which channels does it bring into a single inbox?

WhatsApp, Instagram DM, Facebook Messenger, Telegram, web chat, and email come together in a single inbox. Instagram comment automation is rolled out gradually depending on Meta approval; the first step is WhatsApp and Instagram DM.

Next step

If you haven't trained your assistant yet, start with the how to train the inTusell AI article — that is where the foundation of this operation gets built. To go deeper on the channel side, move on to Instagram and WhatsApp automation, and for order and product questions, read order and product questions automation. The e-commerce solution page holds the full solution, and the all articles page lists the whole series.

To see live how this would work at your own store, request a demo or write directly to hello@intusell.com. In a 20-minute session we open your inbox together and test the first response in the system.