A large share of visitors who reach the cart but leave without buying aren't actually undecided — they simply have a question that hasn't been answered: "Will this size fit me?", "Will it arrive tomorrow?", "What's the return policy?" If these questions don't get an instant answer, the visitor closes the tab and usually never comes back. Setting up chatbots and live chat correctly is the cheapest way to turn that exact moment of hesitation into a sale.
When does pre-purchase hesitation happen
Most visitors hesitate on the product page, in the cart, or right before the checkout step. These moments generally fall into three categories: product uncertainty (fit, material, usage), delivery uncertainty (time, cost, regional restrictions), and trust uncertainty (returns, warranty, payment security).
Even when these questions are already answered on the FAQ page, the visitor has no intention of looking for it — they want an answer right there, on the screen they're on. This is exactly where a chatbot's real value comes in: it brings the answer to where the visitor is, not to where they'd have to go. Unanswered hesitation usually results in cart abandonment; we covered other ways to recover that loss in our cart abandonment recovery article.
The bot's job and the human's job are different
Chatbots and live chat aren't rivals — they're two complementary layers. The bot resolves repetitive, structured questions instantly and around the clock; live chat takes over for cases the bot can't handle, whether emotional or complex.
- Bot territory: structured questions like delivery time, return conditions, stock status, size/fit charts, order tracking,
- Live chat territory: special discount requests, damaged-product complaints, corporate order negotiations, free-text questions the bot can't parse.
A bot deployed without this clear split either forwards every question to a human and loses its purpose, or answers complex questions incorrectly and erodes trust. The right setup is a bot that operates knowing its own limits. The table below summarizes where each layer is strongest:
| Criterion | Chatbot | Live Chat |
|---|---|---|
| Response speed | Instant, within seconds | Minutes, depends on agent availability |
| Cost | Low marginal cost after fixed setup | Depends on staffing, scales linearly with volume |
| Handling complex questions | Limited, can't go beyond structured scenarios | High, understands context and can craft custom solutions |
| 24/7 availability | Yes, uninterrupted | Usually limited to business hours |
Build bot scenarios from real questions
A good bot scenario isn't built on assumptions — it's built on real customer service history. Break down the last three to six months of support requests into categories; the top 15-20 most repeated questions form the bot's initial scenario set. When building scenarios, follow this order:
- Design each scenario with short, clear answers; the bot shouldn't write long paragraphs, it should give an answer that's understood at a glance,
- For questions tied to dynamic product-page data (stock, price, delivery time), make sure the bot pulls live data; static text goes stale fast,
- For product-specific questions like size/fit, match the bot to the product category; offer a guide specific to that product instead of a generic answer (we covered strengthening this area in our size guide article),
- When the bot doesn't understand a question, don't just say "I didn't understand" — suggest the relevant FAQ topic or a live chat option directly; don't leave the user at a dead end.
Make the bot-to-human handoff a rule, not an afterthought
The most common mistake is designing a bot without a clear fork in the road. If a user still hasn't found an answer after two or three attempts, they should automatically be routed to live chat or, at minimum, to a request form.
- Define rules that skip the bot and connect directly to live chat on certain keywords ("cancel", "damaged", "complaint", "urgent"),
- If live chat isn't available outside business hours, state this clearly and offer a way to leave a request through the bot; staying silent is the behavior that erodes trust the most,
- Make sure the conversation history is transferred to the human at the point of handoff; making the customer repeat the same question undoes the time the bot saved.
"A good chatbot isn't one that answers every question — it's one that hands off the question it can't answer to the right person at the right moment."
Track performance with the right metrics
The way to tell whether your bot setup is working isn't "how many messages came in" — it's the indicators that show its impact on sales:
- First response time: should be seconds for the bot, and an upper target of minutes for live chat,
- Resolution rate: the percentage of conversations the bot closes without handing off to a human; aim for a curve that rises over time,
- Post-chat conversion: the purchase rate of visitors who chatted with the bot or live chat, compared to those who never did,
- Abandoned chat count: conversations closed before an answer arrives; a high number here signals a problem with response speed or scenario coverage.
Review these metrics monthly and feed low-resolution-rate topics back into new scenarios; the bot should become a system that improves itself over time, not a fixed rule set. We covered how to institutionalize this tracking on the live chat side with SLA targets and ticket flow in our SLA and ticket management guide.
Integrate with order and stock data
The moment a chatbot loses the most trust is when it gives wrong or outdated information: a product it says is "in stock" is actually sold out, an order it says is "shipping tomorrow" hasn't even been prepared yet. That's why real-time access to stock, order status, and shipment tracking data is even more critical for the bot than the quality of its scenarios.
When a customer enters their order number and the bot instantly shows the current status, it both reduces the load on live chat and increases customer trust. A bot that runs on static text with no connection to any backend system may look easy to set up in the short term, but in the medium term it erodes trust through the risk of giving wrong information.
Live chat setup checklist
- Were the bot scenarios built from real support requests?
- Does the bot have real-time access to stock and order data?
- Are automatic live-chat handoff rules defined for specific keywords?
- Is there an option to leave a request outside business hours?
- Is the conversation history transferred to the human at handoff?
- Are resolution rate and post-chat conversion tracked monthly?
The hard part of building this integration is usually technical: when the bot, stock, and order systems are managed separately, access to current data gets delayed or never happens at all. Şimşek Software's structure, which keeps order and stock data in a single panel, makes it easy for the chat and live support layer to access this information instantly — turning pre-purchase hesitation into a sale with the right answer at the right moment.