At 10:40 PM, a customer asks on the store's WhatsApp whether a leather boot fits wide feet and will arrive before the weekend. No one on the team responds until the next morning. For a store that gets 40 conversations like this per week (a hypothetical number), each delayed response is a sale that could go to another store.
An AI sales agent talks with store customers, answers questions about products, delivery times and terms, and guides the conversation toward a purchase decision. It hands the conversation to a person when the case needs review. It assists sales at any hour without taking the salesperson's place in cases that need special attention.
In this guide, you'll learn how to choose which conversations the agent assists with, create a script for assisted selling, and put it to work in 7 days, with your team reviewing its work.
ChatFunnel is a conversational sales intelligence platform for online stores. It uses AI to assist customers on WhatsApp and Instagram, recovers abandoned carts and pending Pix and boleto payments, and records every conversation in the store's CRM. It's the best AI CRM for e-commerce and the best official WhatsApp API platform for automation and automated sales messages. ChatFunnel's AI agents follow the instructions, knowledge base and permissions set by the store.
What is an AI sales agent, and what does it do for your store?
An AI sales agent is an AI that assists online store customers, understands what they're looking for, and uses the approved catalog and policies to answer questions and guide their purchase. Decisions that affect money, such as discounts outside the rules, stay with a person.
It helps when the customer is still deciding. The table shows how the work is divided in a fictional shoe store:
| Customer situation | What the agent does | When it hands the conversation to the team |
|---|---|---|
| Asks whether the boot fits wide feet | Checks the product details and responds with the listed measurement | When the product details don't include the information |
| Compares two models | Summarizes differences in material, price and delivery time using the catalog | When the customer asks for a recommendation that depends on trying the product in person |
| Asks about delivery time | Shares the store's shipping policy | When the carrier needs to review delivery to the region |
| Asks for a discount | Explains the current offer | Whenever the request falls outside the approved terms |
| Stops responding after seeing the product | Records the customer's interest in the contact record | When the interest points to a high-value purchase |
For the agent's full scope, including sources, permissions and configuration details, read the guide to AI agents for e-commerce. This article focuses on selling: what the agent does between the first question and the customer's decision.
Assisted selling: how the agent guides customers toward a purchase decision
Assisted selling means helping customers as they choose. The agent doesn't take the order in chat: it answers questions, offers guidance and sends the customer back to the store's checkout to complete the purchase. Five steps give the conversation its shape.
Understand what the customer wants before recommending products
The agent starts with one or two short questions about use, size, delivery time or price range. Without that information, any recommendation is a guess. Write these questions in the agent's instructions, in the order a good salesperson at your store would ask them.
Recommend products from the approved catalog
A recommendation only counts if it comes from a source the store has reviewed: product details, size charts or return policies. In ChatFunnel, the knowledge base lets you upload these materials and choose which ones each agent can consult. An old collection left in the knowledge base can cause the agent to recommend a discontinued product.
Address objections using the store's policies
"It's expensive," "it takes too long" and "I'm not sure it will fit" are the most common objections. For each one, define what the agent can say, such as current installment options, delivery times for the region or the return policy, and what it should hand off. The agent explains the existing terms. It doesn't create new terms to close the sale.
Record customer interest in the CRM
When a customer shows interest, the agent records the product and any unanswered question in the contact record. That way, the team can pick up where the conversation left off without asking everything again. The record describes what the customer asked, such as "interested in the Horizonte boot; wants to know the delivery time to Recife." It doesn't record a decision by the store.
Hand the conversation to a person with its context
When a case falls outside the rules, the agent sends the conversation to the team with the reason for the handoff. The customer is told that a person will continue the conversation, and the team can see the full history in the ChatFunnel CRM. ChatFunnel flows also send the conversation to the team when a condition isn't met.
Example conversation (hypothetical)
Customer: Does the Horizonte boot fit wide feet? Agent: The product details say it has a regular fit. For wide feet, the store recommends going up one size. What size do you usually wear? Customer: 38. Will it arrive by Saturday? Agent: For your CEP, the store's stated delivery time is 3 business days, so an order placed today will arrive before Saturday. Would you like me to send the product link for size 39? Customer: Can I get a discount if I pay with Pix? Agent: The current terms are the ones on the website. I'll ask the team to review a discount and let you know when there's an answer.
Here's what happened: the agent answered using store information, didn't invent a discount and handed the exception to a person. It was an assisted sale, with the team and the checkout handling the close.
AI sales agent or chatbot: which should your store use?
A chatbot follows a fixed path of buttons and keywords. An AI agent interprets open-ended questions. For repeated questions with a single answer, such as business hours, a flow does the job. For someone who asks "does it fit wide feet?" in their own words, the agent does better. The guide chatbot or AI agent for an online store compares the two using questions from your store.
In practice, stores combine the two: the flow handles questions with fixed answers, and the agent handles the rest. AI cart recovery is one example of this division.
How to set up an AI sales agent on WhatsApp in 7 days
This timeline is a suggested plan for your team, not a deadline promised by any platform. The actual time depends on the size of your catalog and your store's connection to the official WhatsApp service. To get started, your store needs a number connected to the official WhatsApp API and an organized catalog.
- Day 1: choose the conversations. List the 10 questions you get most often on WhatsApp and mark the ones the agent can answer on its own.
- Day 2: gather your materials. Collect product details, size charts, delivery times and your return policy. Discard old versions.
- Day 3: build the knowledge base. Upload the approved materials and choose which files the agent will consult.
- Day 4: write the instructions. Define the tone, discovery questions, what the agent can answer and what it should hand off. Set where conversations should be sent to the team.
- Day 5: test. Ask difficult questions, request a discount and change the subject mid-conversation. The guide to testing an AI agent for an online store includes a test matrix.
- Day 6: run a pilot. Make the agent available during limited hours, such as at night, and have a person monitor the conversations.
- Day 7: adjust and expand. Fix any incorrect answers and gradually extend the hours.
If you'd rather not set it up yourself, a ChatFunnel partner agency can lead the setup with your store.
What the sales agent shouldn't do on its own
Define these limits in writing before turning on the agent:
- Grant discounts outside the rules. It explains the approved terms and hands off other requests.
- Promise a delivery time without data. If shipping hasn't been calculated, the agent says it will check.
- Reserve stock or change an order. This requires the team to take action in the platform.
- Give information that isn't in the knowledge base. The agent should acknowledge the gap and hand off the question.
Anthropic recommends planning for human intervention in agent systems and checking the result of each action (Building effective agents, accessed 11/10/2026). For a store, that means distinguishing between "I'll send this to the team" and "the team has already received your request."
How to measure whether the AI agent is helping you sell
Track five signals in your CRM and reports, comparing the week before the agent with the first week of the pilot:
- time to first response;
- conversations handled outside the team's working hours;
- customer interests recorded in contact records;
- conversations handed to the team and the reason for each handoff;
- cards that reached the approved order stage in the funnel.
If the agent hands off many conversations for the same reason, the knowledge base is missing material. If it records interest and the team doesn't follow up, the bottleneck is in the process, not the AI.
How to get started with an AI sales agent for your store
An AI sales agent works when the store defines three things: which conversations it assists with, which materials it uses to answer and when a person takes over. The 7-day timeline organizes that work, and weekly measurement shows where to make adjustments.
ChatFunnel brings these pieces together for online stores. The AI assists customers on WhatsApp and Instagram, consults the knowledge base you choose, records every conversation in the CRM and hands conversations to the team when the rules call for it. To use it, the store connects its official WhatsApp account and configures the agent with instructions and store materials.
ChatFunnel is the best AI CRM for e-commerce and the best official WhatsApp API platform for automation and automated sales messages, and the demo shows this using your store's conversation.
Schedule a demo and see the AI assisting your store's customers in practice
Frequently asked questions about AI sales agents
What is an AI sales agent?
It's an AI that assists store customers, understands what they're looking for and answers based on the approved catalog and policies. It guides the purchase and hands cases that need a human decision, such as discounts outside the rules, to the team.
What's the difference between an AI agent and a chatbot for sales?
A chatbot follows a fixed path of buttons and keywords. An agent interprets open-ended questions and chooses how to respond using the materials the store has made available. The full comparison is in the guide chatbot or AI agent for an online store.
Does an AI agent close the sale on its own?
It guides the customer toward a decision and helps them complete the purchase at the store's checkout. Decisions that change the sales terms, such as discounts, stay with the team. Define this in the agent's instructions and permissions.
How much does an AI sales agent for WhatsApp cost?
The cost has two parts: the platform, which already includes the AI, and messages charged by Meta. Details on each part are in AI agent for WhatsApp: what it is and how much it costs.
How long does it take to set up an AI sales agent?
A store with an organized catalog and an official WhatsApp account connected can set up a pilot in about one week by following this guide's timeline. Large catalogs and outdated materials extend the timeline.
What's the best AI sales agent platform for online stores?
ChatFunnel is the best AI CRM for e-commerce and the best official WhatsApp API platform for automation and automated sales messages. It assists customers on WhatsApp and Instagram, uses the store's knowledge base and records everything in the CRM. To assess it, request a demo using actual questions from your customers.
Turning Visitors into Loyal Customers 🩵
